Vamp template track generation method
By attaching EVA strips to the bonding line between the upper and sole and performing three-dimensional difference and loss function optimization, a shoe upper template trajectory is generated, which solves the problems of shoe upper deformation and size difference in traditional shoe manufacturing methods and achieves efficient automation and high-quality bonding between the shoe upper and sole.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN YAOZU LIGHT TECHNOLOGY CO LTD
- Filing Date
- 2026-02-27
- Publication Date
- 2026-04-21
AI Technical Summary
In traditional shoe manufacturing, shoe uppers are mostly made of flexible materials, which leads to local deformation and size differences. Existing automated methods are complex and time-consuming, making it difficult to meet the needs of rapid style changes.
By attaching EVA strips to the bonding line between the upper and sole of the shoe, performing 3D photographic differential and thresholding processing, and combining likelihood and prior loss function iterative optimization, the shoe upper template trajectory is generated.
Significantly improves the automation level of the shoe manufacturing process, ensures the accuracy and consistency of the bonding between the upper and the sole, and enhances production efficiency and product quality.
Smart Images
Figure CN121890818A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of shoe processing technology, and more specifically, to a method for generating a shoe upper template trajectory. Background Technology
[0002] In the footwear industry, the traditional shoemaking process is mainly divided into three stages: cutting and sewing, forming, and packaging. Among them, the forming stage is the key link in bonding the sole and the upper, involving multiple steps such as feeding, roughing, applying treatment agent, drying with treatment agent in an oven, applying glue, drying with glue in an oven, and attaching the sole. For a long time, these processes have mainly relied on manual operation, which is not only inefficient but also makes it difficult to ensure the consistency of product quality. With the development of automation technology, the footwear industry has also begun to explore the path of automated production, especially in the upper processing process, such as the application of automated roughing, glue spraying, and treatment agent, in order to reduce reliance on manual labor and improve production efficiency and product quality. However, shoe uppers are mostly made of flexible materials, and the preceding processes are largely done manually, resulting in localized deformation and dimensional differences. This characteristic makes it difficult to directly apply traditional fixed-track automated methods to shoe upper processing steps, such as roughing, adhesive spraying, and treatment agents. Currently, the industry mainly uses 3D scanning technology to obtain 3D point cloud data of the shoe upper surface, and then calculates the difference between each shoe upper and the template point cloud in real time by creating a 3D template. This allows for adjustment of the bonding lines on the template to guide a six-axis robot to perform automated operations. However, this method is complex and time-consuming, making it difficult to meet the needs of rapid style changes. Therefore, we propose an improved method for generating shoe upper template trajectories. Summary of the Invention
[0003] This invention provides a method for generating a shoe upper template trajectory, including: S1. Data Acquisition: Prepare a shoe upper and attach EVA strips around the entire circle along the adhesive line between the upper and sole; then take a 3D photograph of the shoe upper with the EVA strips attached; after taking the 3D photograph, two sets of images are obtained: haveImages and pureImages; S2. Preliminary extraction of the bonding line between the upper and sole: For each surface image, extract its Z-axis image, and perform a difference analysis between the Z-axis image with the EVA strip and the Z-axis image of the pure upper, i.e., A. ; The differential image was thresholded with a threshold of T=0.35 to obtain the area where the EVA strip was located. Then, the thresholded image was opened to eliminate noise, and the largest independent bright area was selected. The skeletonization operation was then used to obtain the preliminary extracted adhesive lines of the bottom and top. S3. Fine localization of the upper and lower adhesive line: Construct two loss functions, likelihood and prior, and list the total loss function. By minimizing the total loss function, iterative optimization is performed to obtain the final upper and lower adhesive line.
[0004] As a preferred technical solution of this application, in S1, the width of the EVA strip is 3mm and the thickness is 1mm.
[0005] As a preferred technical solution of this application, in S1, before the EVA strip is pasted on the shoe upper, the adhesive line of the upper and sole is drawn with a pen of a different color than the shoe upper.
[0006] As a preferred technical solution of this application, in S1, the three-dimensional scanning method is as follows: the shoe upper is fixed on the scanning device, and three-dimensional photos of the four sides of the shoe, namely the toe, inner side, heel and outer side, are taken.
[0007] As a preferred technical solution of this application, when taking photos, the shoe upper with the EVA strip is first photographed four times around its perimeter; then the EVA strip is peeled off, and the clean shoe upper is photographed four times around its perimeter. Four XYZ images were obtained, with an image size of M. N 3, where M and N are the number of rows and columns of the image, and 3 means that the image has three channels. The gray value of each channel represents the X, Y and Z coordinates of the current pixel in the reference three-dimensional coordinate system. The two sets of images, haveImages and pureImages, each contain four images, each with a size of M. N 3; If we need to know the xyz coordinates of a point P(m,n) on the outer side, then the value is... , , .
[0008] As a preferred technical solution in this application, in S3, the total loss function is expressed as: ; in, Represents the three-dimensional coordinates of the i-th point. Represents all points; express The likelihood loss function at a point express Prior loss function for a point.
[0009] As a preferred technical solution in this application, the likelihood loss function The construction method is as follows: take the pixel coordinates on the difference Z-value map where the current point is located. The angle is the angle perpendicular to the line connecting the current point to the next point. Construct a line segment with the current point as the center and a length of 20 pixels. On the line segment, take points with a length of 1 unit, and obtain the gray value of each point through bilinear interpolation.
[0010] As a preferred technical solution of this application, the equation relating the difference Z-value and the likelihood loss value is as follows: .
[0011] As a preferred technical solution of this application, before constructing the prior loss function, two points before and after the current point are extracted, for a total of 5 points, and the 5 points are mapped into a two-dimensional space, where the Y-axis of the two-dimensional space is the Z coordinate of the point, and the X-axis of the two-dimensional space is the distance between the points; Let the coordinates of the current point be (0, The coordinates of the next point are... ,in, This represents the Euclidean distance from the next point to the current point in the three-dimensional coordinate system. Indicates the next point By applying the coordinate values to these five points, and substituting x=0 into the quadratic equation, we can obtain the expected value for the current point. value ; Calculate the difference between the expected Z-value and the actual Z-value: The prior loss of this distance point is set to 0, and the prior loss of other candidate points is the distance to this point. As a preferred technical solution of this application, after constructing the total loss function, since there are many independent variables, a simplified optimization method is used to obtain the final result: for each point, N candidate positions are established, the prior and likelihood losses of each candidate position are calculated respectively, and the candidate position with the minimum total loss is selected as the current point update position. One update for all points is one iteration. After N iterations, the position of the point tends to remain unchanged, that is, the iteration ends. At this time, the final bottom bonding line is obtained.
[0012] Compared with the prior art, the beneficial effects of the present invention are as follows: In the scheme of this application: 1. This application significantly reduces manual intervention and improves the automation level of the shoe manufacturing process by extracting and generating the bonding lines of the upper and sole. Compared with traditional three-dimensional scanning and template adjustment methods, the method of this invention is simpler and faster, and can quickly create shoe upper template information based on three-dimensional point clouds, which provides convenience for rapid style change in automated shoe manufacturing, thereby improving overall production efficiency.
[0013] 2. By precisely extracting and generating the bonding lines between the upper and the sole, this invention ensures the accuracy and consistency of the bonding between the upper and the sole, which helps to improve the quality of the final product. Attached Figure Description
[0014] Figure 1 A flowchart of the method for generating the shoe upper template trajectory provided in this application; Figure 2 A point cloud diagram composed of the xyz diagram of the outer side of the shoe upper provided in this application; Figure 3 This is a Z-difference image of the shoe toe provided in this application; Figure 4 The preliminary extraction of the adhesive line diagram of the upper and bottom is provided for this application using skeletalization operations; Figure 5 This application provides a pixel coordinate map on the difference Z-value map of the current point. Figure 6 The graph provided in this application shows the difference Z-value versus the likelihood loss value. Figure 7 A schematic diagram showing the mapping of the five points provided in this application into a two-dimensional space; Figure 8 Example diagram of the prior loss value of candidate points when ∆Zi=5 is provided for this application; Figure 9 A schematic diagram of the final upper-sole bonding line provided for this application. Detailed Implementation
[0015] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.
[0016] It should be noted that, unless otherwise specified, the embodiments and features and technical solutions in the present invention can be combined with each other.
[0017] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.
[0018] For an example, please refer to... Figures 1-9 A method for generating a shoe upper template trajectory, comprising: S1. Data Acquisition: Prepare a shoe upper and apply EVA strips around the entire perimeter of the upper-sole adhesive line. Then, take a 3D photograph of the shoe upper with the EVA strips. Two sets of images are obtained after the 3D photograph: haveImages and pureImages. The EVA strips are 3mm wide and 1mm thick. Before applying the EVA strips, the upper-sole adhesive line is marked with a pen of a different color than the shoe upper. Pre-marking the upper-sole adhesive line with a different colored pen provides a preliminary target trajectory, offering a visual reference for subsequent EVA strip positioning and preventing EVA strip misalignment. Using 3mm wide and 1mm thick EVA strips ensures a good fit between the strip and the adhesive line (width matches standard adhesive line size, thickness is moderate and does not obstruct other areas of the shoe upper), and creates a clear Z-axis height difference in the 3D scan, laying the foundation for subsequent differential calculations to extract the adhesive line. Simultaneously, the EVA strips are lightweight, easy to remove, and will not damage the shoe upper or leave any residue, ensuring the integrity and accuracy of the pureImages (clean shoe upper images). The 3D scanning method is as follows: the shoe upper is fixed on the scanning device and 3D photos are taken of its four sides: toe, inner side, heel, and outer side; the partitioned photos of the toe, inner side, heel, and outer side can capture the detailed features of each area (such as the curvature of the toe and the three-dimensionality of the heel), ensuring that the extracted upper and sole bonding lines can accurately correspond to the actual trajectory in each area. When taking photos, first take four photos around the shoe upper with the EVA strip attached; then peel off the EVA strip and take four photos around the clean shoe upper. Four XYZ images were obtained, with an image size of M. N 3, where M and N are the number of rows and columns of the image, and 3 means that the image has three channels. The gray value of each channel represents the X, Y and Z coordinates of the current pixel in the reference three-dimensional coordinate system. The two sets of images, haveImages and pureImages, each contain four images (in the order of toe, outer side, heel, and inner side), and each image has a size of M. N 3; If we need to know the xyz coordinates of a point P(m,n) on the outer side, then the value is... , , , that is Figure 2 As shown; S2. Preliminary extraction of the bonding line between the upper and sole: For each surface image, extract its Z-axis image, and perform a difference analysis between the Z-axis image of the image with EVA strips and the Z-axis image of the pure upper, i.e.: A ; that is, Figure 3As shown, differential operations can directly highlight the grayscale value changes at the location of the EVA strip (i.e., the area of the adhesive line between the upper and lower parts), forming a clear differential image, which provides a high-quality image foundation for subsequent thresholding processing. For the difference image, thresholding is performed with a threshold T=0.35 (>0.35 is 255, <=0.35 is 0), obtaining the region where the EVA strip is located. Then, an opening operation is performed on the thresholded image to eliminate noise (noise is the small bright area after thresholding). The largest independent bright area is selected, and a skeletonization operation is used to obtain the initially extracted adhesive line, as shown below. Figure 4 As shown; S3. Fine positioning of the bonding line between the upper and lower parts: Since the bonding line between the upper and lower parts is generally smooth, in addition to adjusting point by point based on likelihood information such as the difference Z value, the smooth prior information between the current point position and the surrounding points should also be considered. Therefore, this application constructs two loss functions, likelihood and prior, and lists the total loss function. By minimizing the total loss function, iterative optimization is performed to obtain the final bonding line between the upper and lower parts. The loss function is expressed as: ; in, Represents the three-dimensional coordinates of the i-th point. Represents all points; express The likelihood loss function at a point express The prior loss function of the point; Furthermore, the likelihood loss function The construction method is as follows: take the pixel coordinates on the difference Z-value map where the current point is located. The angle is the angle perpendicular to the line connecting the current point and the next point. A line segment with a length of 20 pixels is constructed, centered at the current point. Points are selected along this line segment in units of 1, and the grayscale value (i.e., the difference Z-value) of each point is obtained using bilinear interpolation. Figure 5 As shown; Furthermore, refer to Figure 6 The equations relating the difference Z-value and the likelihood loss value are as follows: ; Furthermore, before constructing the prior loss function, two points before and two points after the current point are extracted, for a total of five points. These five points are then mapped into a two-dimensional space, where the Y-axis represents the Z-coordinate of the point, and the X-axis represents the distance between the points. Specifically, as shown below... Figure 7 As shown; Let the coordinates of the current point be (0, The coordinates of the next point are... ,in, This represents the Euclidean distance from the next point to the current point in the three-dimensional coordinate system. Indicates the next point By applying the coordinate values to these five points, and substituting x=0 into the quadratic equation, we can obtain the expected value for the current point. value ; Calculate the difference between the expected Z-value and the actual Z-value: (With positive and negative directions), the prior loss of this distance point is set to 0, and the prior loss of other candidate points is the distance to this point. Figure 8 An explanation was given, assuming (Example of prior loss value for candidate points). Furthermore, after constructing the overall loss function, due to its large number of independent variables, a simplified optimization method is used to obtain the final result: For each point, N candidate positions are established, the prior and likelihood losses for each candidate position are calculated, and the candidate position with the minimum total loss is selected as the current point's update position. One update for all points constitutes one iteration. After N iterations, the point positions tend to remain unchanged, thus ending the iteration. At this point, the final bottom bonding line is obtained, as shown below. Figure 9 As shown.
[0019] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection, an electrical connection, or a connection that allows communication between them; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components, unless otherwise explicitly limited. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.
[0020] Obviously, the embodiments described above are merely some embodiments of the present invention, not all embodiments. The accompanying drawings show preferred embodiments of the present invention, but do not limit the patent scope of the present invention. The present invention can be implemented in many different forms; rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or make equivalent substitutions for some of the technical features. Any equivalent structures made using the content of this specification and drawings, directly or indirectly applied to other related technical fields, are similarly within the patent protection scope of this invention.
Claims
1. A method for generating a shoe upper template trajectory, characterized in that, include: S1. Data Acquisition: Prepare a shoe upper and attach EVA strips around the entire circle along the adhesive line between the upper and sole; then take a 3D photograph of the shoe upper with the EVA strips attached; after taking the 3D photograph, two sets of images are obtained: haveImages and pureImages; S2. Preliminary extraction of the bonding line between the upper and sole: For each surface image, extract its Z-axis image, and perform a difference analysis between the Z-axis image with the EVA strip and the Z-axis image of the pure upper, i.e., A. ; The differential image was thresholded with a threshold of T=0.35 to obtain the area where the EVA strip was located. Then, the thresholded image was opened to eliminate noise, and the largest independent bright area was selected. The skeletonization operation was used to obtain the preliminary extracted adhesive line of the bottom. S3. Fine localization of the upper and lower adhesive line: Construct two loss functions, likelihood and prior, and list the total loss function. By minimizing the total loss function, iterative optimization is performed to obtain the final upper and lower adhesive line.
2. The method for generating shoe upper template trajectory according to claim 1, characterized in that, In S1, the width of the EVA strip is 3mm and the thickness is 1mm.
3. The method for generating shoe upper template trajectory according to claim 1, characterized in that, In S1, before attaching the EVA strips to the upper, the adhesive line on the sole is drawn with a pen of a different color than the upper.
4. The method for generating a shoe upper template trajectory according to claim 1, characterized in that, In S1, the three-dimensional scanning method is as follows: the shoe upper is fixed on the scanning device, and three-dimensional photos of the four sides of the shoe, namely the toe, inner side, heel, and outer side, are taken.
5. The method for generating a shoe upper template trajectory according to claim 4, characterized in that, When taking photos, first take four photos around the shoe upper with the EVA strip attached; then peel off the EVA strip and take four photos around the clean shoe upper. Four XYZ images were obtained, with an image size of M. N 3, where M and N are the number of rows and columns of the image, and 3 means that the image has three channels. The gray value of each channel represents the X, Y and Z coordinates of the current pixel in the reference three-dimensional coordinate system. The two sets of images, haveImages and pureImages, each contain four images, each with a size of M. N 3; If we need to know the xyz coordinates of a point P(m,n) on the outer side, then the value is... , , .
6. The method for generating a shoe upper template trajectory according to claim 1, characterized in that, In S3, the total loss function is expressed as: ; in, Represents the three-dimensional coordinates of the i-th point. Represents all points; express The likelihood loss function at a point express Prior loss function for a point.
7. The method for generating a shoe upper template trajectory according to claim 6, characterized in that, Likelihood loss function The construction method is as follows: take the pixel coordinates on the difference Z-value map where the current point is located. The angle is the angle perpendicular to the line connecting the current point to the next point. Construct a line segment with the current point as the center and a length of 20 pixels. On the line segment, take points with a length of 1 unit, and obtain the gray value of each point through bilinear interpolation.
8. The method for generating a shoe upper template trajectory according to claim 7, characterized in that, The equation relating the difference Z-value and the likelihood loss value is: 。 9. The method for generating a shoe upper template trajectory according to claim 8, characterized in that, Before constructing the prior loss function, extract two points before and after the current point, for a total of 5 points. Map these 5 points into a two-dimensional space, where the Y-axis of the two-dimensional space is the Z-coordinate of the point, and the X-axis of the two-dimensional space is the distance between the points. Let the coordinates of the current point be (0, The coordinates of the next point are... ,in, This represents the Euclidean distance from the next point to the current point in the three-dimensional coordinate system. Indicates the next point By applying the coordinate values to these five points, and substituting x=0 into the quadratic equation, we can obtain the expected value for the current point. value ; Calculate the difference between the expected Z-value and the actual Z-value: The prior loss of this distance point is set to 0, and the prior loss of other candidate points is the distance to this point.
10. The method for generating a shoe upper template trajectory according to claim 9, characterized in that, After constructing the total loss function, since it has many independent variables, a simplified optimization method is used to obtain the final result: for each point, N candidate positions are established, the prior and likelihood losses of each candidate position are calculated, and the candidate position with the minimum total loss is selected as the current point update position. One update for all points is one iteration. After N iterations, the position of the point tends to remain unchanged, and the iteration ends. At this time, the final bottom bonding line is obtained.