Five-axis raising machine upper polishing area segmentation method, device, medium and equipment
By optimizing the HSV color space and using the minimum surface method of multi-energy constrained soap film, automated high-precision segmentation of the napped area of the shoe upper is achieved, solving the problems of low precision and poor robustness in existing technologies. It is adaptable to various materials and meets the needs of mass production.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- DONGKE CNC (SUZHOU) CO LTD
- Filing Date
- 2026-03-04
- Publication Date
- 2026-06-12
AI Technical Summary
Existing technologies suffer from low precision and poor robustness when segmenting the napped area of the shoe upper, making it difficult to adapt to various materials and resulting in low processing efficiency, which cannot meet the needs of mass production.
The HSV color space optimization method is used to identify the features of the grinding line, and the non-uniform rational B-spline fitting is performed by combining the multi-energy constrained soap film minimum surface method to achieve automated and high-precision segmentation of the shoe upper processing area.
It significantly improves the robustness and fitting accuracy of ground joint line extraction, with segmentation error controlled within 0.05mm. It is compatible with a variety of materials, with a single sample processing time of ≤110s, meeting the mass production cycle and improving the adhesive qualification rate to over 98%.
Smart Images
Figure CN122199575A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of automated shoe manufacturing technology, and in particular to a method, device, medium, and electronic equipment for dividing the shoe upper grinding area of a five-axis napping machine. Background Technology
[0002] In the field of automated shoe manufacturing, upper napping is a crucial process before shoe material bonding, requiring the precise division of a 5-8mm wide annular processing area on a complex free-form surface. The precision of this area directly affects the adhesive bonding effect and the finished product yield. Currently, the industry mainly relies on manual marking or traditional image processing methods to define the processing area.
[0003] In existing technologies, the extraction of the grommets is mostly based on threshold segmentation in the RGB color space. However, the surface of the shoe upper is subject to reflections, stains, and complex lighting interference, resulting in poor robustness of the mark extraction. The extracted discrete point cloud is usually directly fitted with B-splines, but the point cloud has problems such as disorder, outliers, and uneven density, and the fitted curve is prone to local distortion and closure deviation.
[0004] Regarding the segmentation of processing areas, common methods include: the fixed threshold method, which extends equidistantly along the normal direction of the grinding line, but cannot adapt to the free-form surface characteristics of the shoe upper; the ruled surface method, which connects the upper and lower boundaries with straight lines, resulting in poor surface fit; and the traditional minimum surface method, which, although capable of generating smooth surfaces, lacks constraints on the fit of the shoe upper, easily leading to deviations or self-intersections. Actual measurements show that the segmentation error of these methods in complex curved areas often exceeds 0.1mm, causing overcutting and undercutting problems, making it difficult to meet mass production requirements for adhesive qualification.
[0005] In addition, existing solutions are mostly designed for single materials and are not adaptable to different shoe materials such as leather, mesh, and canvas. They also have low processing efficiency, with a single sample typically taking more than 300 seconds, which is difficult to match the pace requirements of mass production lines.
[0006] Therefore, there is an urgent need for an automated method for dividing the napped area of shoe uppers that is highly accurate, robust, adaptable to multiple materials, and meets mass production efficiency requirements. Summary of the Invention
[0007] In view of this, the present invention provides a method, device, storage medium and electronic device for dividing the sanding area of shoe uppers in a five-axis napping machine, the main purpose of which is to solve the problem of inaccurate division of the sanding area of shoe uppers in current napping machines.
[0008] To address the aforementioned problems, this application provides a method for dividing the sanding area of a shoe upper using a five-axis napping machine, comprising: Obtain the color point cloud data of the target shoe upper; The HSV color space optimization method is used to identify the grinding line features of the color point cloud data to obtain the grinding line fitting curve of the target shoe upper. Based on the ground joint line fitting curve, the multi-energy constrained soap film minimum surface method is used for fitting to obtain a non-uniform rational B-spline fitting surface. The processing area of the target shoe upper is segmented based on the non-uniform rational B-spline fitted surface to obtain the target processing area.
[0009] Optionally, the step of using the HSV color space optimization method to identify the grossing line features of the color point cloud data to obtain the grossing line fitting curve of the target shoe upper specifically includes: The HSV color space optimization method is used to extract features from the color point cloud data to obtain the ground line marking point cloud data of the target shoe upper; The grinding groove line marker point cloud data is smoothed to obtain the grinding groove line fitting curve.
[0010] Optionally, the step of using the HSV color space optimization method to extract features from the color point cloud data to obtain the point cloud data of the grossing line markings of the target shoe upper specifically includes: Extract color features from the color point cloud data, including color features of grinding marks and color features of the background; Based on the aforementioned color features, the Otsu algorithm is used to calculate and process the data with the goal of maximizing the inter-class variance, thereby obtaining the dynamic saturation threshold of the HSV color space. Based on the color features of the ground joint markings, the component extraction rules for different spatial dimensions of the HSV color space used for feature extraction are determined; The color point cloud data is used to extract features by employing the dynamic threshold of saturation and the component extraction rules of different spatial dimensions to obtain the ground line marking point cloud data of the target shoe upper.
[0011] Optionally, the smoothing process of the grinding groove mark point cloud data to obtain the grinding groove fitting curve specifically includes: The grinding groove line marker point cloud data is preprocessed to obtain preprocessed grinding groove line point cloud data; The preprocessed grinding line point cloud data was sorted using principal component analysis and angle sorting methods to obtain an annular ordered point cloud sequence. The annular ordered point cloud sequence was smoothed through multiple rounds of iterative processing using Gaussian smoothing and B-spline fitting methods to obtain the fitting curve of the grinding groove line.
[0012] Optionally, the fitting curve based on the ground joint line is fitted using the multi-energy-constrained soap film minimum surface method to obtain a non-uniform rational B-spline fitting surface, specifically including: Discretize the fitting curve of the ground joint line to obtain discrete point cloud data; The extension vector of the ground joint fitting curve is constructed based on the tangent vector of the ground joint fitting curve and the average normal vector of the ground joint fitting curve obtained by KD-Tree neighborhood search. Based on the extended vector, the ground joint line fitting curve is used as the boundary constraint condition to perform surface fitting, and an initial ruled surface is obtained. Based on the energy function with the goal of minimizing the total energy, the gradient descent method is used to iteratively optimize the initial ruled surface to obtain a non-uniform rational B-spline fitted surface. The energy function includes an area energy term, a Laplace energy term, an endpoint constraint term, and a shoe upper fitting term.
[0013] Optionally, the segmentation of the processing area of the target shoe upper based on the non-uniform rational B-spline fitted surface to obtain the target processing area specifically includes: The control vertex mesh is obtained by interpolating the non-uniform rational B-spline fitted surface using cubic radial basis functions. The node vector of the control vertex mesh is generated based on a preset order using two-end heavy nodes; Based on the node vectors, the de Boer-Cox recursive method is used to construct the surface, and the target continuous surface point cloud data for processing is obtained. The target continuous surface point cloud data is segmented to obtain the target processing area.
[0014] Optionally, the non-uniform rational B-spline fitted surface includes a first non-uniform rational B-spline fitted surface and a second non-uniform rational B-spline fitted surface; the segmentation of the target continuous surface point cloud data to obtain the target processing region specifically includes: The L-BFGS-B optimization algorithm is used to solve the projection of the colored point cloud data onto the first non-uniform rational B-spline fitting surface and the second non-uniform rational B-spline fitting surface, respectively, to obtain the coordinates of the first projected point cloud and the second projected point cloud. The first normal dot product is calculated based on the color point cloud data, the coordinates of the first projected point cloud, and the first surface normal vector of the first non-uniform rational B-spline fitted surface. The second normal dot product is calculated based on the color point cloud data, the coordinates of the second projected point cloud, and the second surface normal vector of the second non-uniform rational B-spline fitted surface. Based on the first normal dot product and the second normal dot product, the position of the color point cloud data relative to the first non-uniform rational B-spline fitting surface and the second non-uniform rational B-spline fitting surface is determined, and the color point cloud region formed by the color point cloud data between the first non-uniform rational B-spline fitting surface and the second non-uniform rational B-spline fitting surface is determined as the target processing area.
[0015] To solve the above problems, this application provides a device for dividing the sanding area of a napping machine for shoe uppers, comprising: The acquisition module is used to acquire the color point cloud data of the target shoe upper; The recognition module is used to perform grinding line feature recognition on the color point cloud data using the HSV color space optimization method to obtain the grinding line fitting curve of the target shoe upper. The fitting module is used to fit the ground joint line fitting curve using the multi-energy constrained soap film minimum surface method to obtain a non-uniform rational B-spline fitting surface. The segmentation module is used to segment the processing area of the target shoe upper based on the non-uniform rational B-spline fitted surface to obtain the target processing area.
[0016] To solve the above problems, this application provides a storage medium storing a computer program, which, when executed by a processor, implements the steps of the five-axis napping machine shoe upper grinding area segmentation method described above.
[0017] To solve the above problems, this application provides an electronic device, which includes at least a memory and a processor. The memory stores a computer program, and the processor executes the computer program in the memory to implement the steps of the above-described five-axis napping machine shoe upper sanding area segmentation method.
[0018] The beneficial effects of this application are as follows: By integrating HSV color space optimization, multi-round iterative NURBS curve fitting, and multi-energy constrained minimum surface construction, this application achieves automated and high-precision segmentation of the napped area of the shoe upper, significantly improving the robustness and fitting accuracy of the grinding line extraction. Furthermore, it controls the segmentation error of the processing area within 0.05mm on complex free-form surfaces, achieving a surface fit of over 95%. Simultaneously, this method is adaptable to various materials such as leather and mesh, with a single sample processing time of ≤110s, meeting mass production cycle requirements. It can increase the adhesive qualification rate to over 98% and reduce the manual correction rate to below 1%, providing a high-precision operating benchmark for five-axis polishing equipment and realizing the standardization and digital control of the shoe napping process.
[0019] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description
[0020] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a method for dividing the sanding area of a shoe upper using a five-axis napping machine, as provided in an embodiment of this application, is shown. Figure 2 A flowchart illustrating a method for dividing the sanding area of a shoe upper using a five-axis napping machine, according to another embodiment of this application, is shown. Figure 3 This illustration shows a schematic diagram of the ground line marking point cloud data of the target shoe upper provided in an embodiment of this application; Figure 4 A schematic diagram of the ground joint line fitting curve provided in an embodiment of this application is shown; Figure 5 This illustration shows a schematic diagram of the initial ruled surface of the target shoe upper provided in an embodiment of this application; Figure 6 This illustration shows a schematic diagram of the minimum surface point set after point set processing provided in an embodiment of this application; Figure 7 This paper shows a schematic diagram of the non-uniform rational B-spline fitted surface after extended processing according to an embodiment of this application. Figure 8 This illustration shows an exploded view of the core processing area provided in an embodiment of this application; Figure 9 A structural block diagram of a five-axis napping machine shoe upper sanding area segmentation device according to another embodiment of this application is shown. Detailed Implementation
[0021] Various embodiments and features of this application are described herein with reference to the accompanying drawings.
[0022] It should be understood that various modifications can be made to the embodiments described herein. Therefore, the above description should not be considered as limiting, but merely as an example of embodiments. Other modifications within the scope and spirit of this application will be apparent to those skilled in the art.
[0023] The accompanying drawings, which are included in and form part of this specification, illustrate embodiments of the present application and, together with the general description of the present application given above and the detailed description of the embodiments given below, serve to explain the principles of the present application.
[0024] These and other features of this application will become apparent from the following description of preferred forms of embodiments given as non-limiting examples, with reference to the accompanying drawings.
[0025] It should also be understood that although this application has been described with reference to some specific examples, those skilled in the art can certainly implement many other equivalent forms of this application.
[0026] The above and other aspects, features and advantages of this application will become more apparent when taken in conjunction with the accompanying drawings and in view of the following detailed description.
[0027] Specific embodiments of this application are described thereafter with reference to the accompanying drawings; however, it should be understood that the claimed embodiments are merely examples of this application, which can be implemented in various ways. Well-known and / or repeated functions and structures are not described in detail to avoid unnecessary or redundant details that could obscure the application. Therefore, the specific structural and functional details claimed herein are not intended to be limiting, but merely to teach those skilled in the art to use this application in a variety of substantially any suitable detailed structures.
[0028] This specification may use the phrases “in one embodiment,” “in another embodiment,” “in yet another embodiment,” or “in other embodiments,” all of which may refer to one or more of the same or different embodiments according to this application.
[0029] This application provides a method for dividing the shoe upper sanding area using a five-axis napping machine, such as... Figure 1 As shown, it includes: Step S101: Obtain the color point cloud data of the target shoe upper; In this step, a line laser 3D camera can be used to acquire color point cloud data of the target shoe upper using texture mapping technology. While scanning the shoe upper to obtain its 3D contour, a built-in binocular camera captures the surface texture, and an algorithm maps color information to each spatial point to generate a color point cloud. Alternatively, an RGB-D camera can be used to simultaneously acquire depth and color maps in a single shot. The two images are aligned using camera calibration parameters to calculate the coordinates and color of each pixel in 3D space, generating a color point cloud.
[0030] Step S102: The HSV color space optimization method is used to identify the grinding line features of the colored point cloud data to obtain the grinding line fitting curve of the target shoe upper. In this step, the HSV color space optimization method is used to extract features from the color point cloud data to obtain the point cloud data of the grossing line markers on the target shoe upper. The grossing line marker point cloud data is then smoothed to obtain the grossing line fitting curve. A full-format adaptation algorithm is used to accurately parse the shoe upper point cloud data under different storage modes (ASCII, binary, and compressed), bitwise operations are used to restore the packaged color information, and further HSV space transformation and clustering denoising are employed to overcome complex lighting interference, achieving robust extraction of the grossing line markers, thus providing high-quality data support for subsequent fitting.
[0031] Step S103: Based on the ground joint line fitting curve, the multi-energy constrained soap film minimum surface method is used for fitting to obtain a non-uniform rational B-spline fitting surface; In this step, the fitted curve of the grinding groove line is discretized to obtain discrete point cloud data. An extension vector of the fitted curve is constructed based on the tangent vector of the fitted curve and the average normal vector of the fitted curve obtained using KD-Tree neighborhood search. A surface is fitted using the fitted curve as a boundary constraint based on the extension vector to obtain an initial ruled surface. The initial ruled surface is iteratively optimized using gradient descent based on an energy function that minimizes the total energy, resulting in a non-uniform rational B-spline fitted surface. The energy function includes an area energy term, a Laplace energy term, an endpoint constraint term, and a shoe upper fitting term.
[0032] Step S104: Based on the non-uniform rational B-spline fitted surface, the processing area of the target shoe upper is segmented to obtain the target processing area.
[0033] In this step, a cubic radial basis function is used to interpolate the non-uniform rational B-spline fitted surface to obtain a control vertex mesh. Based on a preset order, node vectors of the control vertex mesh are generated using double nodes at both ends. Based on the node vectors, the de Boer-Cox recursive method is used to construct the surface to obtain the target continuous surface point cloud data for processing. The target continuous surface point cloud data is segmented to obtain the target processing area.
[0034] This application achieves automated, high-precision segmentation of the napped area in shoe uppers by integrating HSV color space optimization, multi-round iterative NURBS curve fitting, and multi-energy-constrained minimum surface construction. This significantly improves the robustness and fitting accuracy of the ground line extraction, and controls the segmentation error of the processing area within 0.05mm on complex free-form surfaces, with a surface fit of over 95%. Furthermore, this method is adaptable to various materials such as leather and mesh, with a single sample processing time of ≤110s, meeting mass production cycle requirements. It can increase the adhesive qualification rate to over 98% and reduce the manual correction rate to below 1%, providing a high-precision operating benchmark for five-axis polishing equipment and realizing the standardization and digital control of the shoe napping process.
[0035] Another embodiment of this application provides a different method for dividing the shoe upper sanding area using a five-axis napping machine, such as... Figure 2 As shown, it includes: Step S201: Obtain the color point cloud data of the target shoe upper; In the specific implementation of this step, a line laser 3D camera can be used to acquire color point cloud data of the target shoe upper using texture mapping technology. While scanning the shoe upper to obtain its 3D contour, a built-in binocular camera captures the surface texture, and an algorithm maps color information to each spatial point to generate a color point cloud. Alternatively, an RGB-D camera can be used to simultaneously acquire depth and color images in a single shot. The two images are aligned using camera calibration parameters to calculate the coordinates and color of each pixel in 3D space, generating a color point cloud. This application does not limit the acquisition of color point cloud data.
[0036] Step S202: Use the HSV color space optimization method to extract features from the color point cloud data to obtain the ground line marking point cloud data of the target shoe upper; In this step, the color features of the color point cloud data are extracted. These color features include color features of the grinding mark class and color features of the background class. The color features of the grinding mark class include red color features and blue color features. Based on these color features, the Otsu algorithm is used to calculate and process the data with the goal of maximizing the inter-class variance, thereby obtaining the dynamic saturation threshold of the HSV color space. The mathematical formula for calculating the inter-class variance is as follows:
[0037] in, For class proportions. Traverse. The normalized range of the H / S components in the HSV color space is denoted as T, which is selected as the dynamic saturation threshold to maximize T. Based on the color features of the ground metal marking class, component extraction rules for different spatial dimensions of the HSV color space are determined for feature extraction. When the ground metal marking class color feature is red, the component extraction rules for different spatial dimensions of the HSV color space are as follows: H component range is... The S component has a predetermined lower threshold of 0.3; the V component range is [0.2, 0.9]. When the ground joint marking color feature is blue; the H component range is... The S component has a predetermined lower threshold of 0.25; the V component range is [0.3, 0.85]. Feature extraction is performed on the color point cloud data using the dynamic saturation threshold and the component extraction rules for different spatial dimensions to obtain the point cloud data of the grossing line markings of the target shoe upper. First, based on a comparison between the predetermined lower threshold of the S component and the dynamic saturation threshold, the larger value of the two is determined as the dynamic segmentation threshold of the S component; the color point cloud data that simultaneously satisfies the color features of the grossing line markings, including the H component range, the V component range, and the dynamic segmentation threshold of the S component, is determined as the point cloud data of the grossing line markings of the target shoe upper. Figure 3 The diagram shows the point cloud data of the grinding line markings of the target shoe upper obtained by extracting features from the color point cloud data. The grinding line marking point cloud data includes a first grinding line point cloud and a second grinding line point cloud. The first grinding line point cloud is located on the upper side of the image and is called the upper grinding line point cloud. The second grinding line point cloud is located on the lower side of the image and is called the lower grinding line point cloud.
[0038] Step S203: Smooth the point cloud data of the ground joint line marker to obtain the ground joint line fitting curve; In this step, the grinding groove mark point cloud data is preprocessed to obtain preprocessed grinding groove point cloud data; KD-Tree nearest neighbor search is used to calculate the distance from the grinding groove mark point cloud data to its nearest neighbor. k The average distance of the nearest neighbor grinding line marker point cloud data Based on global average distance with standard deviation Determine the adaptive threshold .in, Average distance; Standard deviation; retain After identifying the points, a second standard deviation is used for screening to remove those exceeding the X, Y, and Z directions. The points within the range are used to obtain a clean point set composed of preprocessed grinding line point cloud data. This step can reduce the noise residual rate to below 0.3%. Principal component analysis (PCA) and angle sorting methods are used to sort the preprocessed grinding line point cloud data, obtaining a ring-shaped ordered point cloud sequence; based on PCA and angle sorting, the ring-shaped ordered arrangement of the point cloud is achieved. Calculation center point Where M is Find the number of midpoints; construct vectors for each point relative to the center point. Based on its angle with the X-axis Pair set according to In ascending order sorting yields an ordered set of points. The annular ordered point cloud sequence is iteratively smoothed using Gaussian smoothing and B-spline fitting methods to obtain the fitting curve for the grinding joint line. The process includes the following steps: Step 1: Use a one-dimensional Gaussian function to weight and fuse the coordinates of ordered points to eliminate local fluctuations. The mathematical expression is as follows:
[0039] in, is the smoothing coefficient, and x is the x-coordinate of a point on the fitted curve of the ground joint line. The x-axis is the weighted and merged coordinate; Step 2: The smoothed point set is then extended at both ends to ensure closure. Specifically, the B-spline fitting order k=5. The fitting process minimizes the error. 2 Solve for the control vertices; where, This is the i-th calculation point on the fitted curve; Let be the coordinates of the i-th ordered point cloud after sorting.
[0040] Step 3: Repeat steps 1 to 2 until the preset iteration round is reached. The preset iteration round can be 3, and the smoothing parameter is gradually increased by 500, 1000, and 1500.
[0041] Step 4: After the iteration is complete, within the parameter range A set of 5000 uniformly sampled points is used to generate the final fitting curve point set Pfinal, corresponding to the final grinding joint line fitting curve. This point set possesses the characteristics of uniform density and continuous fourth-order derivatives, ensuring an ultra-smooth curve that is suitable for five-axis machining path planning. Figure 4 The figure shown is a schematic diagram of the fitting curve for the ground joint line.
[0042] Step S204: Discretize the fitting curve of the ground joint line to obtain discrete point cloud data; In the specific implementation process of this step, the fitted curve of the ground joint line is used. Discrete sampling is performed. The sampling density is based on the total arc length L and the minimum feature size of the shoe upper. Determine the sampling interval. Must meet For example, when L = 500 mm, the number of sampling points N ≥ 2000.
[0043] Step S205: Based on the tangent vector of the ground joint fitting curve, and using the average normal vector of the ground joint fitting curve calculated based on KD-Tree neighborhood search, the extension vector of the ground joint fitting curve is constructed. In this step, to prevent surface deviation, the extension direction of the ruled surface directly determines the relative position of the initial curved surface and the upper. If the direction is incorrect (e.g., pointing towards the outside of the upper instead of the processing area), subsequent optimization will fail to achieve a proper fit to the upper. The extension direction needs to be determined by combining the tangent of the ground line and the normal of the upper surface: the tangent vector of the ground line at point Pi. The first derivative of the NURBS curve is:
[0044] After normalization, we have:
[0045] Shoe upper normal correction: For each sampling point Pi, select its neighborhood point set in the shoe upper point cloud. (Including 50 points around Pi, obtained based on KD-Tree nearest neighbor search), calculate the average normal vector of the neighborhood point set, where Let be the unit normal vector of point Q (obtained by PCA principal component analysis, with the minimum eigenvalue corresponding to the normal direction). The direction must point towards the inside of the shoe upper (the side where the processing area is located). If the direction is opposite, take the negative value, denoted as . .
[0046] Combined with tangent vector The average normal vector calculated based on KD-Tree neighborhood search Construct extended vectors The mathematical expression is as follows:
[0047] in, for The vertical vector (take the counterclockwise direction to ensure that the extension direction of all sampling points is consistent). The normal correction weights (determined through 10 sets of experiments) (If it's too large, the extension direction will deviate excessively from the tangent; if it's too small, it won't fit the shoe upper properly.) Ultimately, this affects... Normalization yields .
[0048] Step S206: Based on the extended vector, use the ground joint line fitting curve as the boundary constraint to perform surface fitting and obtain the initial ruled surface; In the specific implementation process, this step is based on the extended vector. Using the ground joint line fitting curve as a boundary constraint, a surface fitting is performed using a predetermined surface equation to obtain an initial ruled surface; as shown below. Figure 5 The diagram shown is a schematic of the initial ruled surface; the mathematical expression for the predetermined surface equation is as follows:
[0049] in, The initial ruled surface is represented by colored point cloud data. Fitting curves for the already fitted ground joint lines; represents the linear parameters of the ruled surface; The initial extension length can be 0.1 mm; Let be the extended vector corresponding to u.
[0050] The fit of the initial ruled surface was tested to ensure the average distance from the verification point to the curved surface. Only then can we proceed to the next stage.
[0051] Step S207: Based on the energy function with the goal of minimizing the total energy, the initial ruled surface is iteratively optimized using the gradient descent method to obtain a non-uniform rational B-spline fitted surface; In the specific implementation process, this step involves minimizing the total energy function. This causes the initial surface to evolve into a true soap film equilibrium state. The mathematical expression for the total energy function is as follows:
[0052] in, These are weighting coefficients, determined using a controlled variable method to ensure a balanced effect of each energy term. The multiple energy terms include: area energy term. Its physical meaning is the degree of stretching of the soap film, which is achieved by minimizing the area by calculating the sum of the cross product moduli of the triangular facets; Laplace energy term. This measures the deviation of a vertex from its neighboring vertices; for internal vertices, it uses neighborhood mean constraints, and for boundary vertices, it constructs virtual neighborhoods through "mirror expansion," ensuring smoothness and no self-intersections. Endpoint constraint energy term. Limiting the endpoint extension within the process-allowed value Lmax; and the upper bonding energy term. The shoe upper dot cloud is used as implicit support, and distance penalty is used to ensure that the surface does not deviate from the real shape. Iterative optimization: Gradient descent is used, with a piecewise adaptive step size strategy. After 250-300 iterations, when the energy changes... Convergence is determined at the specified time, and the target ruled surface is obtained, such as... Figure 5 The diagram shows a schematic of the target ruled surface. The iteratively optimized triangular mesh is converted into a point cloud file suitable for processing: specifically, point set processing is performed on the target ruled surface to obtain the minimum surface point set; specifically, mesh vertices are extracted and "distance threshold deduplication" is performed (threshold d < 0.001 mm) to ensure coordinate accuracy retains 6 decimal places. Simultaneously, linear interpolation ensures the maximum distance between adjacent points is ≤ 0.3 mm. For example... Figure 6 The figure shows the minimum surface point set after point set processing; the minimum surface point set is then adaptively enlarged to obtain the non-uniform rational B-spline fitted surface; specifically, to avoid missing edge processing, the enlargement amount is calculated based on the bounding box size as follows:
[0053] in, The maximum x-axis coordinate among all points before surface expansion; The minimum x-axis coordinate among all points before surface expansion; The maximum y-axis coordinate among all points before surface expansion; The minimum y-axis coordinate among all points before surface expansion. A local planar model is constructed within the expanded region, and the sampling points are completed, ensuring that the normal deviation between the completed points and the original surface is ≤0.02mm, resulting in a non-uniform rational B-spline fitted surface after boundary expansion. For example... Figure 7 The image shows the non-uniform rational B-spline fitted surface after the extended processing. Step S208: Interpolate the non-uniform rational B-spline fitted surface using cubic radial basis functions to obtain the control vertex mesh; In this step, cubic radial basis function (RBF) interpolation is used to complete the missing Z coordinates and generate a control vertex mesh.
[0054] Step S209: Generate the node vector of the control vertex mesh using two-end heavy nodes based on a preset order; In this step, the order is set to p=q=3, and the node vector is generated using the "double-node" method. Here, p is the order parameter for calculating the node vector in the u direction, and q is the order parameter for calculating the node vector in the v direction.
[0055] Step S210: Based on the node vectors, the de Boer-Cox recursive method is used to construct the surface to obtain the target continuous surface point cloud data for processing; In the specific implementation process of this step, the de Boer-Cox recursive method is used to construct the surface based on the node vector to obtain the target continuous surface point cloud data for processing.
[0056] Step S211: Segment the target continuous surface point cloud data to obtain the target processing area.
[0057] In this step, the L-BFGS-B optimization algorithm is used to solve for the projections of the colored point cloud data onto the first and second non-uniform rational B-spline fitting surfaces, respectively, to obtain the coordinates of the first and second projected point clouds. L-BFGS-B (Limited-memory Broyden–Fletcher–Goldfarb–Shanno with Bounds) is an algorithm based on the colored point cloud data P, the coordinates of the first projected point cloud, and the first surface normal vector of the first non-uniform rational B-spline fitting surface. The calculation yields the first normal dot product; the mathematical expression is as follows: Based on the color point cloud data, the coordinates of the second projected point cloud, and the normal vector of the second surface of the second non-uniform rational B-spline fitting surface, a second normal dot product is calculated. Based on the first and second normal dot products, the position of the color point cloud data relative to the first and second non-uniform rational B-spline fitting surfaces is determined. The color point cloud region formed by the color point cloud data located between the first and second non-uniform rational B-spline fitting surfaces is determined as the target processing region. Figure 8 The diagram shown is a breakdown of the core processing area.
[0058] This application achieves automated, high-precision segmentation of the napped area in shoe uppers by integrating HSV color space optimization, multi-round iterative NURBS curve fitting, and multi-energy-constrained minimum surface construction. This significantly improves the robustness and fitting accuracy of the ground line extraction, and controls the segmentation error of the processing area within 0.05mm on complex free-form surfaces, with a surface fit of over 95%. Furthermore, this method is adaptable to various materials such as leather and mesh, with a single sample processing time of ≤110s, meeting mass production cycle requirements. It can increase the adhesive qualification rate to over 98% and reduce the manual correction rate to below 1%, providing a high-precision operating benchmark for five-axis polishing equipment and realizing the standardization and digital control of the shoe napping process.
[0059] Another embodiment of this application provides a five-axis napping machine shoe upper sanding area segmentation device 900, such as... Figure 9 As shown, it includes: The acquisition module 901 is used to acquire the color point cloud data of the target shoe upper; The recognition module 902 is used to perform grinding line feature recognition on the color point cloud data using the HSV color space optimization method to obtain the grinding line fitting curve of the target shoe upper. The fitting module 903 is used to fit the ground joint line fitting curve using the multi-energy constrained soap film minimum surface method to obtain a non-uniform rational B-spline fitting surface. The segmentation module 904 is used to segment the processing area of the target shoe upper based on the non-uniform rational B-spline fitted surface to obtain the target processing area.
[0060] In the specific implementation process, the recognition module 902 is specifically used to extract features from the color point cloud data using the HSV color space optimization method to obtain the grinding line mark point cloud data of the target shoe upper; and to smooth the grinding line mark point cloud data to obtain the grinding line fitting curve.
[0061] In the specific implementation process, the recognition module 902 is further used to extract color features from the color point cloud data, the color features including grossing mark color features and background color features; based on the color features, the Otsu algorithm is used to calculate and process the data with the goal of maximizing the inter-class variance, to obtain the dynamic threshold of saturation in the HSV color space; according to the grossing mark color features, the component extraction rules for different spatial dimensions of the HSV color space used for feature extraction are determined; the saturation dynamic threshold and the component extraction rules for different spatial dimensions are used to extract features from the color point cloud data to obtain the grossing mark point cloud data of the target shoe upper.
[0062] In the specific implementation process, the identification module 902 is further used to preprocess the grinding groove line marker point cloud data to obtain preprocessed grinding groove line point cloud data; sort the preprocessed grinding groove line point cloud data using principal component analysis and angle sorting methods to obtain an annular ordered point cloud sequence; and perform multi-round iterative smoothing processing on the annular ordered point cloud sequence using Gaussian smoothing and B-spline fitting methods to obtain the grinding groove line fitting curve.
[0063] In the specific implementation process, the fitting module 903 is specifically used to discretize the fitting curve of the grinding groove line to obtain discrete point cloud data; based on the tangent vector of the fitting curve of the grinding groove line and the average normal vector of the fitting curve of the grinding groove line calculated based on KD-Tree neighborhood search, the extension vector of the fitting curve of the grinding groove line is constructed; based on the extension vector and using the fitting curve of the grinding groove line as the boundary constraint, the surface is fitted to obtain an initial ruled surface; based on the energy function with the goal of minimizing the total energy, the gradient descent method is used to iteratively optimize the initial ruled surface to obtain a non-uniform rational B-spline fitted surface.
[0064] In the specific implementation process, the segmentation module 904 is specifically used to: perform interpolation calculation on the non-uniform rational B-spline fitted surface using cubic radial basis functions to obtain a control vertex mesh; generate node vectors of the control vertex mesh using double nodes at both ends based on a preset order; construct the surface using the de Boer-Cox recursive method based on the node vectors to obtain target continuous surface point cloud data for processing; and segment the target continuous surface point cloud data to obtain the target processing area.
[0065] In the specific implementation process, the segmentation module 904 is further used to: use the L-BFGS-B optimization algorithm to solve the projection of the color point cloud data onto the first non-uniform rational B-spline fitting surface and the second non-uniform rational B-spline fitting surface, respectively, to obtain the first projected point cloud coordinates and the second projected point cloud coordinates; calculate the first normal dot product based on the color point cloud data, the first projected point cloud coordinates, and the first surface normal vector of the first non-uniform rational B-spline fitting surface; calculate the second normal dot product based on the color point cloud data, the second projected point cloud coordinates, and the second surface normal vector of the second non-uniform rational B-spline fitting surface; determine the position of the color point cloud data relative to the first non-uniform rational B-spline fitting surface and the second non-uniform rational B-spline fitting surface based on the first normal dot product and the second normal dot product; and determine the color point cloud region formed by the color point cloud data between the first non-uniform rational B-spline fitting surface and the second non-uniform rational B-spline fitting surface as the target processing area.
[0066] This application achieves automated, high-precision segmentation of the napped area in shoe uppers by integrating HSV color space optimization, multi-round iterative NURBS curve fitting, and multi-energy-constrained minimum surface construction. This significantly improves the robustness and fitting accuracy of the ground line extraction, and controls the segmentation error of the processing area within 0.05mm on complex free-form surfaces, with a surface fit of over 95%. Furthermore, this method is adaptable to various materials such as leather and mesh, with a single sample processing time of ≤110s, meeting mass production cycle requirements. It can increase the adhesive qualification rate to over 98% and reduce the manual correction rate to below 1%, providing a high-precision operating benchmark for five-axis polishing equipment and realizing the standardization and digital control of the shoe napping process.
[0067] Another embodiment of this application provides a storage medium storing a computer program, which, when executed by a processor, implements the following method steps: Step 1: Obtain the color point cloud data of the target shoe upper; Step 2: Use the HSV color space optimization method to identify the grinding line features of the color point cloud data to obtain the grinding line fitting curve of the target shoe upper. Step 3: Based on the ground joint line fitting curve, the multi-energy constrained soap film minimum surface method is used for fitting to obtain a non-uniform rational B-spline fitting surface; Step 4: Based on the non-uniform rational B-spline fitted surface, segment the processing area of the target shoe upper to obtain the target processing area.
[0068] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
[0069] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.
[0070] The specific implementation process of the above method steps can be found in the embodiment of the above-mentioned method for dividing the shoe upper grinding area of any five-axis napping machine. This embodiment will not be repeated here.
[0071] This application achieves automated, high-precision segmentation of the napped area in shoe uppers by integrating HSV color space optimization, multi-round iterative NURBS curve fitting, and multi-energy-constrained minimum surface construction. This significantly improves the robustness and fitting accuracy of the ground line extraction, and controls the segmentation error of the processing area within 0.05mm on complex free-form surfaces, with a surface fit of over 95%. Furthermore, this method is adaptable to various materials such as leather and mesh, with a single sample processing time of ≤110s, meeting mass production cycle requirements. It can increase the adhesive qualification rate to over 98% and reduce the manual correction rate to below 1%, providing a high-precision operating benchmark for five-axis polishing equipment and realizing the standardization and digital control of the shoe napping process.
[0072] Another embodiment of this application provides an electronic device, which can be a server. The electronic device includes a processor, a memory, a network interface, and a database connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile and / or volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The network interface is used to communicate with external clients via a network connection. When the program is executed by the processor, it implements the functions or steps of a server-side method for dividing the sanding area of a five-axis napping machine for shoe uppers.
[0073] In one embodiment, an electronic device is provided, which can be a client. The electronic device includes a processor, memory, a network interface, a display screen, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The network interface is used to communicate with an external server via a network connection. When the program of the electronic device is executed by the processor, it implements the functions or steps of a client-side method for dividing the sanding area of a five-axis napping machine for shoe uppers.
[0074] Another embodiment of this application provides an electronic device, including at least a memory and a processor. The memory stores a computer program, and the processor, when executing the computer program in the memory, performs the following method steps: Step 1: Obtain the color point cloud data of the target shoe upper; Step 2: Use the HSV color space optimization method to identify the grinding line features of the color point cloud data to obtain the grinding line fitting curve of the target shoe upper. Step 3: Based on the ground joint line fitting curve, the multi-energy constrained soap film minimum surface method is used for fitting to obtain a non-uniform rational B-spline fitting surface; Step 4: Based on the non-uniform rational B-spline fitted surface, segment the processing area of the target shoe upper to obtain the target processing area.
[0075] The specific implementation process of the above method steps can be found in the embodiment of the above-mentioned method for dividing the shoe upper grinding area of any five-axis napping machine. This embodiment will not be repeated here.
[0076] This application achieves automated, high-precision segmentation of the napped area in shoe uppers by integrating HSV color space optimization, multi-round iterative NURBS curve fitting, and multi-energy-constrained minimum surface construction. This significantly improves the robustness and fitting accuracy of the ground line extraction, and controls the segmentation error of the processing area within 0.05mm on complex free-form surfaces, with a surface fit of over 95%. Furthermore, this method is adaptable to various materials such as leather and mesh, with a single sample processing time of ≤110s, meeting mass production cycle requirements. It can increase the adhesive qualification rate to over 98% and reduce the manual correction rate to below 1%, providing a high-precision operating benchmark for five-axis polishing equipment and realizing the standardization and digital control of the shoe napping process.
[0077] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. Those skilled in the art can make various modifications or equivalent substitutions to this application within the scope and nature of this application, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.
Claims
1. A method for dividing the sanding area of a shoe upper using a five-axis napping machine, characterized in that, include: Obtain the color point cloud data of the target shoe upper; The HSV color space optimization method is used to identify the grinding line features of the color point cloud data to obtain the grinding line fitting curve of the target shoe upper. Based on the ground joint line fitting curve, the multi-energy constrained soap film minimum surface method is used for fitting to obtain a non-uniform rational B-spline fitting surface. The processing area of the target shoe upper is segmented based on the non-uniform rational B-spline fitted surface to obtain the target processing area.
2. The method as described in claim 1, characterized in that, The step of using the HSV color space optimization method to identify the grossing line features of the color point cloud data to obtain the grossing line fitting curve of the target shoe upper specifically includes: The HSV color space optimization method is used to extract features from the color point cloud data to obtain the ground line marking point cloud data of the target shoe upper; The grinding groove line marker point cloud data is smoothed to obtain the grinding groove line fitting curve.
3. The method as described in claim 2, characterized in that, The step of using the HSV color space optimization method to extract features from the color point cloud data to obtain the point cloud data of the grossing line markings on the target shoe upper specifically includes: Extract color features from the color point cloud data, including color features of grinding marks and color features of the background; Based on the aforementioned color features, the Otsu algorithm is used to calculate and process the data with the goal of maximizing the inter-class variance, thereby obtaining the dynamic saturation threshold of the HSV color space. Based on the color features of the ground joint markings, the component extraction rules for different spatial dimensions of the HSV color space used for feature extraction are determined; The color point cloud data is used to extract features by employing the dynamic threshold of saturation and the component extraction rules of different spatial dimensions to obtain the ground line marking point cloud data of the target shoe upper.
4. The method as described in claim 2, characterized in that, The smoothing process of the grinding groove mark point cloud data to obtain the grinding groove fitting curve specifically includes: The grinding groove line marker point cloud data is preprocessed to obtain preprocessed grinding groove line point cloud data; The preprocessed grinding line point cloud data was sorted using principal component analysis and angle sorting methods to obtain an annular ordered point cloud sequence. The annular ordered point cloud sequence was smoothed through multiple rounds of iterative processing using Gaussian smoothing and B-spline fitting methods to obtain the fitting curve of the grinding groove line.
5. The method as described in claim 1, characterized in that, The fitting curve based on the ground joint line is fitted using the multi-energy constrained soap film minimum surface method to obtain a non-uniform rational B-spline fitting surface, specifically including: Discretize the fitting curve of the ground joint line to obtain discrete point cloud data; The extension vector of the ground joint fitting curve is constructed based on the tangent vector of the ground joint fitting curve and the average normal vector of the ground joint fitting curve obtained by KD-Tree neighborhood search. Based on the extended vector, the ground joint line fitting curve is used as the boundary constraint condition to perform surface fitting, and an initial ruled surface is obtained. Based on the energy function with the goal of minimizing the total energy, the initial ruled surface is iteratively optimized using the gradient descent method to obtain a non-uniform rational B-spline fitted surface. The energy function includes an area energy term, a Laplace energy term, an endpoint constraint term, and a shoe upper fitting term.
6. The method as described in claim 1, characterized in that, The segmentation of the processing area of the target shoe upper based on the non-uniform rational B-spline fitted surface to obtain the target processing area specifically includes: The control vertex mesh is obtained by interpolating the non-uniform rational B-spline fitted surface using cubic radial basis functions. The node vector of the control vertex mesh is generated based on a preset order using two-end heavy nodes; Based on the node vectors, the de Boer-Cox recursive method is used to construct the surface, and the target continuous surface point cloud data for processing is obtained. The target continuous curved surface point cloud data is segmented to obtain the target processing area.
7. The method as described in claim 6, characterized in that, The non-uniform rational B-spline fitted surface includes a first non-uniform rational B-spline fitted surface and a second non-uniform rational B-spline fitted surface; the segmentation of the target continuous surface point cloud data to obtain the target processing region specifically includes: The L-BFGS-B optimization algorithm is used to solve the projection of the colored point cloud data onto the first non-uniform rational B-spline fitting surface and the second non-uniform rational B-spline fitting surface, respectively, to obtain the coordinates of the first projected point cloud and the second projected point cloud. The first normal dot product is calculated based on the color point cloud data, the coordinates of the first projected point cloud, and the first surface normal vector of the first non-uniform rational B-spline fitted surface. The second normal dot product is calculated based on the color point cloud data, the coordinates of the second projected point cloud, and the second surface normal vector of the second non-uniform rational B-spline fitted surface. Based on the first normal dot product and the second normal dot product, the position of the color point cloud data relative to the first non-uniform rational B-spline fitting surface and the second non-uniform rational B-spline fitting surface is determined, and the color point cloud region formed by the color point cloud data between the first non-uniform rational B-spline fitting surface and the second non-uniform rational B-spline fitting surface is determined as the target processing area.
8. A five-axis napping machine shoe upper sanding area segmentation device, characterized in that, include: The acquisition module is used to acquire the color point cloud data of the target shoe upper; The recognition module is used to perform grinding line feature recognition on the color point cloud data using the HSV color space optimization method to obtain the grinding line fitting curve of the target shoe upper. The fitting module is used to fit the ground joint line fitting curve using the multi-energy constrained soap film minimum surface method to obtain a non-uniform rational B-spline fitting surface. The segmentation module is used to segment the processing area of the target shoe upper based on the non-uniform rational B-spline fitted surface to obtain the target processing area.
9. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the steps of the five-axis napping machine shoe upper grinding area segmentation method according to any one of claims 1-7.
10. An electronic device, characterized in that, It includes at least a memory and a processor, wherein the memory stores a computer program, and the processor, when executing the computer program in the memory, implements the steps of the five-axis napping machine shoe upper grinding area segmentation method according to any one of claims 1-7.