A shoe piece layout method based on an incomplete critical polygon algorithm

Abstract

The application discloses a shoe piece layout method based on an incomplete critical polygon algorithm, and utilizes a ResNet18 algorithm to realize automatic classification of a shoe piece shape, and is divided into a symmetric concave shape, a symmetric convex shape, a strip shape and a special shape; then, according to the category of the shoe piece, the entry and layout order and the entry and layout mode of the shoe piece are determined; then, an improved incomplete critical polygon algorithm is adopted to realize overlap detection of a shoe piece with a circular arc, and a layoutable area of the shoe piece is generated; finally, a positioning strategy is utilized to determine the placement mode, the placement position and the placement angle of the shoe piece, and the position of the shoe piece in the layoutable area is determined; after the shoe piece with a large W value is laid out, a shoe piece with a small W value is laid in to fill the gap generated by the layout of the large shoe piece, and finally, online automatic layout of the shoe piece on the leather is realized. After the layout is completed, a layout result graph, a layout utilization rate and the layout quantity of each shoe piece are output. The application reduces overlap detection of useless areas, and greatly improves the layout efficiency of the shoe piece.

Description

Technical Field

[0001] This invention relates to the field of clothing and footwear production technology, and in particular to a shoe pattern layout method based on an incomplete critical polygon algorithm. Background Technology

[0002] Safety shoe panels vary considerably in shape, and under the constraint of irregularly shaped leather, a single panel placement strategy is difficult to apply to all panels. The upper of a safety shoe consists of numerous panels; deciding which panels to arrange first and which to nest affects material utilization. Traditional critical polygon algorithms can be used for overlap detection when arranging polygonal panels, but they are not suitable for overlap detection of polygonal panels with curved edges, which are common in safety shoe panels.

[0003] The pattern layout method for safety shoes differs from that for everyday shoes. Everyday shoes prioritize aesthetics, and to avoid color variations, the important shoe panels must be laid out on the same piece of leather with consistent texture. Safety shoes, however, prioritize functionality, comfort, and durability, focusing more on maximizing material utilization than aesthetics. Therefore, safety shoes typically use only 2-3 different patterns cut from a single piece of leather, rather than multiple patterns. This makes the pattern layout methods used for everyday shoes unsuitable for safety shoes.

[0004] In the existing technology, (1) manual layout: workers visually observe the boundaries of the leather and the positions of the already arranged shoe pieces, and arrange and cut the shoe pieces according to their personal experience. The labor cost is high and the layout and cutting efficiency is low. Workers are affected by fatigue and lack of concentration, which reduces the utilization rate of layout. Manual cutting requires making a cutting knife for each shoe piece of each size. Safety shoes have a large number of shoe pieces, with sizes ranging from 35 to 48, which means that dozens or even hundreds of cutting knives need to be made for one style of safety shoe. (2) NFP layout algorithm based on angle rotation: On the basis of generating critical polygons by the "Minkowski vector sum" method, the centroid distance measurement is added to find the optimal rotation angle to generate critical polygons, solving the problem of overlapping detection of shoe piece layout. It is only suitable for the layout overlapping detection of relatively simple polygons and polygons without arcs. For polygons with many curves and high irregularity, it is difficult to layout in a low time. (3) Neural Network-Based Layout Algorithm: The part information is input into a neural network model, which predicts the placement order and angle of irregular parts during the layout process to obtain the optimal layout scheme. This method is currently not applicable to layout on irregular natural leather materials, as it focuses on finding the placement order and angle of parts on regular sheet metal or fabric. Summary of the Invention

[0005] The purpose of this invention is to provide a shoe piece layout method based on an incomplete critical polygon algorithm. By utilizing an improved incomplete critical polygon algorithm and shoe piece insertion and positioning strategies, automatic layout of safety shoe pieces can be achieved.

[0006] The technical solution adopted in this invention is:

[0007] A shoe piece layout method based on an incomplete critical polygon algorithm includes the following steps:

[0008] The ResNet18 algorithm is used to automatically classify shoe piece shapes into symmetrical concave, symmetrical convex, strip, and irregular shapes. Next, the layout weight W of the shoe pieces is determined based on their category and area, determining the placement order and method (single or nested layout). Then, an improved incomplete critical polygon algorithm is used to detect overlap and generate layoutable areas. Finally, a positioning strategy is used to determine the placement method, position, and angle of the shoe pieces, identifying their location within the layoutable area. After shoe pieces with higher weights are placed, shoe pieces with lower weights are placed to fill the gaps created by the larger shoe pieces, ultimately achieving online automatic layout of shoe pieces on the leather. After layout, the layout result image, layout utilization rate, and the number of shoe pieces placed are output.

[0009] Step 1: Obtain shoe piece images and automatically classify shoe piece shapes;

[0010] Step 2: Determine the layout weight W of the shoe pieces based on their classification and area, and determine the order and method of placing the shoe pieces.

[0011] Step 3: Perform overlap detection on the shoe pieces to generate a sortable area for the shoe pieces; the sortable area refers to the area where the shoe piece to be sorted will not overlap with the already sorted shoe pieces, nor will it exceed the boundary of the leather material.

[0012] Step 4: Use positioning strategies to determine the placement method, position, and angle of the shoe pieces, and determine the position of the shoe pieces in the layout area;

[0013] Step 5: After the shoe pieces with higher weights are arranged, shoe pieces with lower weights are then added to fill the gaps created by the arrangement of the larger shoe pieces, thus achieving online automatic arrangement of shoe pieces on the leather.

[0014] Step 6: After the layout is completed, output the layout result diagram, layout utilization rate and layout quantity of each shoe piece.

[0015] Furthermore, step 1 utilizes the ResNet18 algorithm to automatically classify the shape of the shoe piece, specifically including the following steps:

[0016] Step 1-1: Extract the shoe piece contour features through convolution operation, and use the ReLU activation function to increase the non-linearity of the neural network, increase the expressive power of the neural network, and improve the final classification ability of the model;

[0017] Steps 1-2: Reduce the size of the shoe piece feature map through pooling operations while preserving the main features of the shoe piece;

[0018] Steps 1-3 address the vanishing gradient problem during model training using four residual blocks, enabling the network to learn shoe features more deeply and effectively. Each residual block contains several convolutional layers and a ReLU activation function.

[0019] Steps 1-4: Global average pooling is used to average the values ​​of each channel of the feature map, thereby converting the shoe piece feature map into a fixed-length vector.

[0020] Steps 1-5 connect the output of global average pooling to a fully connected layer and use softmax to classify the shoe pieces.

[0021] Furthermore, in step 1, the shoe pieces are classified into irregular, symmetrical, and strip shapes, with the symmetrical shapes further divided into two types: symmetrical concave and symmetrical convex.

[0022] Furthermore, the sorting methods in step 2 include individual nested sorting and nested sorting; the specific sorting steps are as follows:

[0023] Step 2-1: Sort all shoe pieces in descending order of area, where N is the position of the shoe piece to be sorted in the sequence, and N is used as the area weight.

[0024] Step 2-2: Assign shape weight M to the shoe pieces according to the shoe piece category. Assign different shape weights to irregular shapes, symmetrical concave shapes, symmetrical convex shapes, and strip shapes. Among them, the shape weights of irregular shapes and symmetrical concave shapes are greater than the weights of other shapes.

[0025] Steps 2-3: Calculate the concave area of ​​all symmetrical concave shoe pieces, and take the largest concave area as the threshold. Shoe pieces with an area less than the threshold are classified as nestable shoe pieces.

[0026] Steps 2-4: Calculate the maximum length and width of the concave part of the symmetrical concave shoe piece, traverse the nestable shoe pieces to search for small shoe pieces used to embed the concave part, extract the searched small shoe pieces from the sequence, combine them with the corresponding symmetrical concave shoe pieces, and recalculate the area and area weight N of the combined shoe pieces.

[0027] Steps 2-5: Calculate the layout weight W = N × M for all shoe pieces, and finally determine the shoe piece placement order according to the layout weight W value from large to small; then select small shoe pieces to fill the gaps left by the large area of ​​shoe piece layout, starting from the smallest W value to the largest.

[0028] Furthermore, in steps 2-5, if the W values ​​of the two shoe pieces are the same, the principle of prioritizing area weight is adopted, and the piece with the larger area weight is included first.

[0029] Furthermore, in step 3, the incomplete critical polygon algorithm is used to perform overlap detection on the arc-shaped shoe pieces to generate a layoutable area for the shoe pieces. This specifically includes the following steps:

[0030] Step 3-1: Based on the traditional critical polygon algorithm based on trajectory lines, add a circular arc contact detection strategy, and treat the circular arc as a special angle to detect the contact between the shoe piece to be arranged and the boundary of the area to be arranged;

[0031] Step 3-2: Determine whether the tangent vector of the straight edge falls within the tangent vector interval at both ends of the arc; if so, determine that the arc and the straight line can contact each other; otherwise, the arc and the straight line cannot contact each other.

[0032] Step 3-3, Contact detection between arc and corner: Arcs include convex arcs and concave arcs. Determine whether the arc is convex. If it is, perform contact detection between the convex arc and the corner; otherwise, perform contact detection between the concave arc and the corner.

[0033] Steps 3-4: Merge the already laid-out shoe pieces at the bottom of the cowhide into a laid-out area, and fit the upper boundary of the laid-out area with the left contour of the cowhide part to form a new area to be laid out;

[0034] Steps 3-5: The shoe pieces are placed from the left or bottom of the area to be arranged. A trajectory line is generated by moving the shoe pieces relative to the left or bottom of the area to be arranged. The set of trajectory lines is obtained, the innermost trajectory line is extracted, and the generated incomplete critical polygon is used as the area to be arranged.

[0035] Furthermore, the contact detection steps between the convex arc and the corner in step 3-3 are as follows:

[0036] Step 3-3-11: For the convex circular arc, obtain the tangent vectors Ve', Vs, Ve and Vs' at the two endpoints to form the vector interval N; at the same time, obtain the side vectors Vp and Vc on the two opposite sides to form the side vector interval K.

[0037] Where Vs represents the tangent vector at one endpoint when one side of the angle is tangent to one endpoint of the arc; Vs' represents the tangent vector at one endpoint when the other side of the angle is tangent to one endpoint of the arc; Ve' represents the tangent vector at the other endpoint when one side of the angle is tangent to the other endpoint of the arc; Ve represents the tangent vector at the other endpoint when the other side of the angle is tangent to the other endpoint of the arc; the side vector of one side of the angle is Vp; the side vector of the other side of the angle is Vc.

[0038] Step 3-3-12: Determine whether there is an intersection between vector interval N and vector interval K; if so, determine that the convex arc and the corner can contact each other; otherwise, determine that the convex arc and the corner cannot contact each other.

[0039] Step 3-3-13: When Vc and Vp are both located between Vs and Ve, the convex arc does not intersect with either side of the angle, and the contact range between the convex arc and the angle is the entire convex arc. When Vc is located between Ve and Vs', during the sliding process of the convex arc relative to the angle, the CD side of the angle will intersect with part of the arc, and the closer Vc is to Vs', the smaller the contact range between the convex arc and the angle, and the closer the critical contact point Q is to point A, that is, the contact range between the convex arc and the angle is part of the convex arc AQ. When Vp is located between Ve' and Vs, during the sliding process of the convex arc relative to the angle, the CF side of the angle will intersect with part of the arc, and the closer Vp is to Ve', the smaller the contact range between the convex arc and the angle, and the closer the critical contact point Q is to point B, that is, the contact range between the convex arc and the angle is part of the convex arc QB.

[0040] Furthermore, the contact detection steps between the concave arc and the corner in step 3-3 are as follows:

[0041] Step 3-3-21: Calculate the tangent vectors Vs, Ve', Vs' and Ve at the two endpoints of the concave circular arc;

[0042] Step 3-3-22: When the two side vectors Vc and Vp of the angle are both located between Vs' and Ve', and the concave arc does not intersect with either side of the angle, the range that the concave arc can contact the angle is the entire concave arc.

[0043] Step 3-3-23: When the side vector Vc of the angle is located between Ve and Vs' or the side vector Vp is located between Ve' and Vs, there are 1 or 2 tangent points between the side of the angle and the arc. The range in which the angle can contact the concave arc is a partial concave arc. A partial concave arc refers to the part between the tangent points or the part between the endpoint and the tangent point.

[0044] Further, the specific method of steps 3-4 is as follows: when the corner or arc of the shoe piece to be arranged contacts a certain segment of the boundary of the area to be arranged, determine the contact point between the shoe piece to be arranged and the corresponding segment of the boundary of the area to be arranged, keep the contact point position of the shoe piece to be arranged unchanged, move the shoe piece to be arranged so that the contact point of the shoe piece to be arranged slides along the corresponding segment boundary from one end to the other end, and obtain the sliding trajectory of the reference point F of the shoe piece to be arranged during the movement as part of the trajectory line of the critical polygon;

[0045] Specifically, there are no limitations on the reference point, which can be determined as needed. It can be any point on the shoe piece to be arranged. As a feasible implementation method, the present invention uses the center of gravity of the shoe piece as the reference point.

[0046] Alternatively, when the corner or arc of a certain boundary of the area to be arranged can contact the straight edge or arc of the shoe piece to be arranged, keep the contact point position of that boundary of the area to be arranged unchanged, move the shoe piece to be arranged so that the contact point on the boundary of the area to be arranged slides along one side of the shoe piece from one end to the other end, and obtain another part of the sliding trajectory of the reference point F of the shoe piece to be arranged during the movement as another part of the trajectory line of the critical polygon.

[0047] Furthermore, the shoe piece positioning strategy in step 4 includes:

[0048] (1) Irregular shoe piece positioning strategy: Irregular shoe pieces are arranged in a normal single row, and the lowest point of the center of gravity in the area to be arranged is used as the entry point of the new shoe piece; the entry angle is tried in turn from 0°, 60°, 120°, 180°, 240° and 300°, and the waste area on the left and bottom sides of each angle is used as the selection criterion. The angle with the least waste is selected as the entry angle of the new shoe piece; after the shoe pieces are placed, the arranged shoe pieces are merged to generate a new area to be arranged and wait for the next shoe piece to be placed, until the leather can no longer accommodate the current shoe piece;

[0049] (2) Positioning strategy for symmetrical concave shoe pieces: Symmetrical concave shoe pieces adopt a combination of single-row and double-row arrangement, that is, shoe pieces in two adjacent rows are aligned and arranged, and two adjacent shoe pieces in the same row are arranged alternately, combining concave and gap into a large space to fit small shoe pieces, reducing material waste.

[0050] The angle between the baseline and the horizontal line is selected for the layout angle: A straight line is fitted from the bottom edge contour of the leather to serve as the baseline for the layout, replacing the bottom contour of the cowhide; the angle between the baseline and the horizontal line is calculated to determine the layout angle of the symmetrical concave shoe piece; the internal no-fit polygon (INFP) between the shoe piece to be laid out and the left edge of the leather and the baseline is obtained, that is, the layout area on the left side and bottom of the leather is obtained, and the lower left side point of the layout area is taken as the layout position of the symmetrical concave shoe piece.

[0051] (3) Positioning strategy for symmetrical convex shoe pieces: Symmetrical convex shoe pieces adopt a head-to-head single-row method, that is, adjacent shoe pieces in the same row are arranged alternately in front and back, and adjacent rows do not need to be aligned; the strategy for determining the entry angle and entry point is the same as that for symmetrical concave shoe pieces, and the relative arrangement position of adjacent shoe pieces can be calculated offline when head-to-head single-row;

[0052] (4) Positioning strategy for strip shoe pieces: The gaps generated by the regular arrangement of strip shoe pieces in a single direction are minimal. The strategy for determining the entry point and entry angle is the same as that for concave shoe pieces.

[0053] This invention employs the above technical solution, utilizing the ResNet18 algorithm to automatically classify shoe piece shapes into symmetrical concave, symmetrical convex, strip, and irregular shapes. Next, based on the shoe piece's category and area, it determines the shoe piece's layout weight W, the shoe piece's placement order, and the placement method (whether individual or nested). Then, an improved incomplete critical polygon algorithm is used to detect overlaps in shoe pieces with curved surfaces, generating a layoutable area for the shoe pieces. Finally, a positioning strategy is used to determine the shoe piece's placement method, position, and angle, determining its location within the layoutable area. After shoe pieces with higher weights are placed, shoe pieces with lower weights are placed to fill the gaps created by the larger shoe pieces, ultimately achieving online automatic layout of shoe pieces on the leather. After layout is complete, a layout result diagram, layout utilization rate, and the number of shoe pieces placed are output.

[0054] This invention, based on the incomplete critical polygon algorithm, can not only solve the problem of overlapping detection of shoe pieces with rounded edges, but also reduce the overlapping detection of useless areas, greatly improving the efficiency of shoe piece layout and material utilization. Attached Figure Description

[0055] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments;

[0056] Figure 1 This is a flowchart illustrating a shoe piece layout method based on an incomplete critical polygon algorithm according to the present invention.

[0057] Figure 2a This is a schematic diagram showing the state where the arc and the straight edge are tangent when the arc and the straight edge can make contact during detection.

[0058] Figure 2b This is a schematic diagram of the tangent vector of the straight edge when the arc and the straight edge can be in contact for detection.

[0059] Figure 2c This is a schematic diagram of the tangent vector at the endpoint of the arc when the arc and the straight edge can be in contact.

[0060] Figure 2d This is a schematic diagram of the tangent vector interval of the arc when the arc and the straight edge can be in contact.

[0061] Figure 3a This is a schematic diagram showing two possible contact states between a convex arc and a corner;

[0062] Figure 3b This is a schematic diagram of a convex circular arc tangent to an angle;

[0063] Figure 4a A schematic diagram of the contact state between a convex circular arc and a corner when determining the contact vector range between the convex circular arc and the corner.

[0064] Figure 4bA schematic diagram illustrating the state of determining the tangent vector interval of a convex circular arc when determining the contact vector interval between the convex circular arc and the angle.

[0065] Figure 4c A schematic diagram illustrating the state of obtaining the angle and side vector interval when the contact vector interval between the convex arc and the angle is obtained;

[0066] Figure 4d A schematic diagram of the intersection of vector intervals when determining the contact vector interval between a convex circular arc and an angle;

[0067] Figure 5a A schematic diagram of the contact state between a concave circular arc and a corner when determining the vector range where the concave circular arc and the corner can contact each other;

[0068] Figure 5b A schematic diagram showing the state of determining the tangent vector interval of a concave circular arc when the vector interval between the concave circular arc and the angle can be determined.

[0069] Figure 5c A schematic diagram illustrating the state of obtaining the angle and side vector intervals when the contact vector intervals of a concave arc and an angle are determined.

[0070] Figure 5d A schematic diagram of the intersection of vector intervals when determining the contact vector interval between a concave arc and an angle;

[0071] Figure 6a A schematic diagram illustrating the state of obtaining the straight-edge sliding trajectory line when the trajectory line of the reference point F is obtained;

[0072] Figure 6b A schematic diagram illustrating the state of obtaining the angular sliding trajectory when the trajectory of the reference point F is obtained;

[0073] Figure 7 A schematic diagram of the trajectory lines of the critical polygon between the shoe piece to be arranged and the lower left part of the area to be arranged;

[0074] Figure 8 A schematic diagram showing the area where samples can be arranged according to the present invention;

[0075] Figure 9 This is a diagram of the ResNet18 network model of the present invention;

[0076] Figure 10a This is a schematic diagram of the irregularly shaped shoe piece of the present invention;

[0077] Figure 10b This is a schematic diagram of the symmetrical concave shoe piece of the present invention;

[0078] Figure 10c This is a schematic diagram of the symmetrical convex shoe piece of the present invention;

[0079] Figure 10d This is a schematic diagram of the strip-shaped shoe piece of the present invention;

[0080] Figure 11 This is a schematic diagram of the irregular shoe piece arrangement of the present invention;

[0081] Figure 12 This is a schematic diagram of the symmetrical concave shoe piece arrangement scheme of the present invention;

[0082] Figure 13 This is a schematic diagram of the symmetrical concave shoe piece arrangement of the present invention;

[0083] Figure 14 This is a schematic diagram of the symmetrical convex shoe piece arrangement of the present invention;

[0084] Figure 15 This is a schematic diagram of the arrangement of the strip-shaped shoe pieces according to the present invention;

[0085] Figure 16 This is a schematic diagram of the arrangement after filling the gaps in the present invention. Implementation

[0086] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0087] Safety shoes: Also known as work shoes, these are footwear designed to protect wearers from accidents and ensure the safety of the work area. Shoe pieces: Components that make up the upper of a shoe; each shoe's upper is composed of a certain number of shoe pieces. The shape of each shoe piece varies considerably depending on the shoe style and its position on the upper. Layout: Given a space and several objects, this involves rationally arranging these objects within a specified space under constraints to achieve an optimal goal. Layout can be one-dimensional, two-dimensional, or three-dimensional; this invention focuses on two-dimensional safety shoe piece layout. Critical polygon algorithm: Also known as the No-fit Polygon (NFP) algorithm, this is an algorithm that can solve for the placement of polygons, achieving the detection of non-overlapping and non-covering polygons.

[0088] like Figures 1 to 16 As shown in the figure, this invention discloses a shoe piece layout method based on an incomplete critical polygon algorithm, which includes the following steps:

[0089] Step 1: Obtain shoe piece images and automatically classify shoe piece shapes;

[0090] Step 2: Determine the layout weight W of the shoe pieces based on their classification and area, and determine the order and method of placing the shoe pieces.

[0091] Step 3: Perform overlap detection on the shoe pieces to generate a sortable area for the shoe pieces; the sortable area refers to the area where the shoe piece to be sorted will not overlap with the already sorted shoe pieces, nor will it exceed the boundary of the leather material.

[0092] Step 4: Use positioning strategies to determine the placement method, position, and angle of the shoe pieces, and determine the position of the shoe pieces in the layout area;

[0093] Step 5: After the shoe pieces with higher weights are arranged, shoe pieces with lower weights are then added to fill the gaps created by the arrangement of the larger shoe pieces, thus achieving online automatic arrangement of shoe pieces on the leather.

[0094] Step 6: After the layout is completed, output the layout result diagram, layout utilization rate and layout quantity of each shoe piece.

[0095] Furthermore, shoe piece positioning is used to determine the placement method, position, and angle of the shoe pieces within the layable area. Common placement methods include standard single-row, standard double-row, head-to-head single-row, and head-to-head double-row. The placement position, i.e., the entry position, is commonly determined using methods such as the lowest center of gravity method, the BLF method, and the lowest horizontal line method. However, a single method is difficult to apply universally to all shoe piece shapes. Therefore, this invention employs different positioning strategies for different categories of shoe pieces based on the classification results to reduce the time spent on online layout.

[0096] Step 1 uses the ResNet18 algorithm to automatically classify the shape of the shoe piece. Specifically, it includes the following steps:

[0097] Step 1-1: Extract the shoe piece contour features through convolution operation, and use the ReLU activation function to increase the non-linearity of the neural network, increase the expressive power of the neural network, and improve the final classification ability of the model;

[0098] Steps 1-2: Reduce the size of the shoe piece feature map through pooling operations while preserving the main features of the shoe piece;

[0099] Steps 1-3, through four residual blocks (each residual block contains several convolutional layers and ReLU activation functions), solve the gradient vanishing problem during the training of this model, enabling the network to learn shoe features more deeply and effectively;

[0100] Steps 1-4: Global average pooling is used to average the values ​​of each channel of the feature map, thereby converting the shoe piece feature map into a fixed-length vector.

[0101] Steps 1-5 connect the output of global average pooling to a fully connected layer and use softmax to classify the shoe pieces.

[0102] Furthermore, in step 2, the shoe pieces are categorized into symmetrical concave, symmetrical convex, strip, and irregular shapes. Ultimately, the shoe pieces are divided into three categories: irregular, symmetrical, and strip, with the symmetrical type further divided into concave and convex shapes.

[0103] Specifically, a ResNet18 deep convolutional neural network is used to automatically classify different shoe pieces. This network takes shoe piece images as input and extracts and classifies shoe piece features through the following processes: ① Convolutional operations are used to extract the contour features of the shoe pieces, and the ReLU activation function is used to increase the non-linearity of the neural network, thereby increasing its expressive power and improving the model's final classification ability; ② Pooling operations are used to reduce the size of the shoe piece feature map while retaining the main features; ③ Four residual blocks (each containing several convolutional layers and ReLU activation functions) solve the gradient vanishing problem during model training, enabling the network to learn shoe piece features more deeply and effectively; ④ Global average pooling is used to average the values ​​of each channel of the feature map, thus converting the shoe piece feature map into a fixed-length vector; ⑤ The output of global average pooling is connected to a fully connected layer, and softmax is used to classify the shoe pieces. Finally, shoe pieces are classified into three categories: irregular, symmetrical, and striped. Symmetrical shapes are further divided into concave and convex shapes, such as... Figure 9 As shown.

[0104] Furthermore, the sorting methods in step 2 include individual nested nested nesting. The sorting steps in step 3 are as follows:

[0105] Step 2-1: Sort all shoe pieces in descending order of area, where N is the position of the shoe piece to be sorted in the sequence, and N is used as the area weight.

[0106] Step 2-2: Based on the shoe piece category, assign shape weight M to the shoe piece. Assign shape weights of 4.5, 4, 1.5, and 1 to irregular, symmetrical concave, symmetrical convex, and strip shapes, respectively. Among them, irregular and symmetrical concave shoe pieces have larger shape weights.

[0107] Steps 2-3: Calculate the concave area of ​​all symmetrical concave shoe pieces, and take the largest concave area as the threshold. Shoe pieces with an area less than the threshold are classified as nestable shoe pieces.

[0108] Steps 2-4: Calculate the maximum length and width of the concave portion of the symmetrical concave shoe piece, traverse the nestable shoe pieces, search for small shoe pieces that can be embedded in the concave portion, extract these shoe pieces from the sequence, combine them with the corresponding symmetrical concave shoe pieces, and recalculate the area and area weight N of the combined shoe pieces.

[0109] Steps 2-5: Calculate the layout weight W = N × M for all shoe pieces, and finally determine the layout order of the shoe pieces according to the size of the W value. Shoe pieces with larger W values ​​are laid out first, and the remaining gaps are filled with shoe pieces with smaller W values; this is called nesting. The selection of nested shoe pieces starts from the reverse order of W values, filling as many gaps as possible left by the layout of large shoe pieces to improve utilization. When two shoe pieces have the same W value, the area weight priority principle is adopted, and the one with the larger area weight is placed first.

[0110] Specifically, since many gaps are generated during the shoe piece layout process, waste will occur if they cannot be utilized, resulting in a decrease in utilization rate. In order to minimize gaps, this invention incorporates a weight-based sorting mechanism based on the classification results and area characteristics of shoe pieces, combined with the experience of manual layout, to determine which shoe pieces should be placed first and which shoe pieces can be nested.

[0111] The principle of the sorting mechanism proposed in this invention is to assign greater weight to shoe pieces with large areas and shoe pieces with large indentations. The specific sorting mechanism is as follows: ① Arrange all shoe pieces in descending order of area, where N is the position of the shoe piece to be sorted in the sequence, and N is used as the area weight. ② Assign shape weight M to shoe pieces according to their category, assigning shape weights of 4.5, 4, 1.5, and 1 to irregular, symmetrical concave, symmetrical convex, and striped shoe pieces respectively, with irregular and symmetrical concave shoe pieces having larger shape weights. ③ Calculate the indentation area of ​​all symmetrical concave shoe pieces, and take the largest indentation area as a threshold, classifying shoe pieces with areas smaller than the threshold as nestable shoe pieces. ④ Calculate the maximum length and width of the indented portion of the symmetrical concave shoe piece, traverse the nestable shoe pieces, search for small shoe pieces that can be embedded in the indented portion of the shoe piece, extract these shoe pieces from the sequence, combine them with the corresponding symmetrical concave shoe pieces, and recalculate the area and area weight N of the combined shoe piece. ⑤ Calculate the layout weight W = N × M for all shoe pieces, and finally determine the layout order of the shoe pieces according to the size of the W value. Shoe pieces with larger W values ​​are laid out first, and the remaining gaps are filled with shoe pieces with smaller W values. This is called nesting. The selection of nested shoe pieces starts from the reverse order of W values, in order to fill the gaps left by the layout of large shoe pieces as much as possible and improve the utilization rate.

[0112] In special cases, two shoe pieces may have the same W value. This invention adopts an area-weighted priority principle, placing the piece with the larger area weight first. This is because the smaller shoe piece may be used as a nested or inlaid shoe piece.

[0113] Furthermore, in step 3, an incomplete critical polygon algorithm is used to detect overlap of the shoe pieces. To determine the final placement position of the shoe pieces, the possible placement area must first be determined. The so-called possible placement area refers to the area where the shoe piece to be placed will neither overlap with the already placed shoe pieces nor exceed the boundary of the leather. This invention proposes to use an improved incomplete critical polygon algorithm to generate the possible placement area.

[0114] Among the methods for generating critical polygons, commonly used ones include the "sliding collision method," the "convex segmentation method," the "Minkowski vector sum method," and the "trajectory line method." Of these, the trajectory line-based critical polygon algorithm can easily handle overlap detection of polygons with concavities and is relatively time-efficient, making it suitable for online nesting. However, generating a nestable region using the trajectory line method first requires determining whether the boundaries of the shoe piece to be nested and the region to be nested are contactable. Only when they are in contact can a sliding trajectory be generated. The boundary of the region to be nested is composed of certain edges of the already nested shoe pieces and parts of the leather boundary; that is, the boundary of the region to be nested is composed of corners, straight edges, and arcs. Similarly, the boundary of the shoe piece to be nested is also composed of a certain number of corners, straight edges, and arcs. Therefore, determining whether the boundaries of the shoe piece to be nested and the region to be nested are contactable is essentially detecting whether the corners, straight edges, and arcs constituting their boundaries are in contact. While corner and edge contact detection has been addressed in the original trajectory line-based critical polygon algorithm, a good method is still lacking for detecting contact between arcs and lines, arcs and corners, and arcs and arcs. The contact detection algorithm proposed in this invention adds a circular arc contact detection strategy to the traditional trajectory-based critical polygon algorithm, treating the circular arc as a special angle to solve the above problems; specifically, it includes the following steps:

[0115] Step 3-1: Based on the traditional critical polygon algorithm based on trajectory lines, add a circular arc contact detection strategy, and treat the circular arc as a special angle for contact detection of the shoe pieces to be arranged;

[0116] Step 3-2: Determine whether the tangent vector of the straight edge falls within the tangent vector interval of the arc; if so, determine that the arc is in contact with the straight line; otherwise, the arc and the straight line are not in contact.

[0117] Specifically, the contact detection of an arc and a straight line: when the arc edge of one shoe piece can contact the straight edge of another shoe piece, it must be because the arc and the straight edge are tangent. Figure 2a As shown in the diagram. If the shoe pieces are already arranged at the following angles... Figure 2c As shown in the figure, at this placement angle, the tangent vectors of the endpoints A and B of the arc are Vs and Ve, respectively. The tangent vectors of other points on the arc must fall between Vs and Ve, as shown in the figure. Figure 2d As shown in the diagram. When the circular arc AB is tangent to the straight side CD at point G, the tangent vector of the straight side CD is Va, and Va is also the tangent vector of the circular arc AB at point G, as shown in the diagram. Figure 2b Therefore, by determining whether the tangent vector of the straight edge falls within the tangent vector range of the arc, it can be determined whether the arc and the straight edge can contact each other.

[0118] Step 3-3, Contact detection between arc and corner: Arcs include convex arcs and concave arcs. Determine whether the arc is convex. If it is, perform contact detection between the convex arc and the corner; otherwise, perform contact detection between the concave arc and the corner.

[0119] For the convex circular arc, the tangent vectors Ve', Vs, Ve, and Vs' at the two endpoints are obtained to form the vector interval N; at the same time, the side vectors Vp and Vc on the two opposite sides are obtained to form the side vector interval K.

[0120] Where Vs represents the tangent vector at one endpoint when one side of the angle is tangent to one endpoint of the arc; Vs' represents the tangent vector at one endpoint when the other side of the angle is tangent to one endpoint of the arc; Ve' represents the tangent vector at the other endpoint when one side of the angle is tangent to the other endpoint of the arc; Ve represents the tangent vector at the other endpoint when the other side of the angle is tangent to the other endpoint of the arc; the side vector of one side of the angle is Vp; the side vector of the other side of the angle is Vc.

[0121] Determine whether there is an intersection between vector interval N and vector interval K; if so, determine whether the convex arc and the corner can touch; otherwise, determine whether the convex arc and the corner cannot touch.

[0122] For a concave circular arc, find the tangent vectors Vs, Ve', Vs' and Ve at the two endpoints;

[0123] When the two side vectors Vc and Vp of the angle are both located between Vs' and Ve', and the concave arc does not intersect with either side of the angle, the range that the concave arc can contact the angle is the entire concave arc.

[0124] When the side vector Vc of the angle is between Ve and Vs' or the side vector Vp is between Ve' and Vs, the area where the angle can contact the concave arc is a portion of the concave arc.

[0125] Specifically, there are two types of angles: convex angles and concave angles. However, only convex angles can come into contact with arcs. Similarly, there are two types of arcs: convex arcs and concave arcs. Therefore, there are two types of contact between an angle and an arc: convex arc contacting the angle and concave arc contacting the angle.

[0126] Contact detection between a convex circular arc and a corner: When a convex circular arc and a corner are in contact, both sides forming the corner must be located outside the convex circular arc, and neither side of the corner intersects the arc. Figure 3a The angle MLN is a critical case where one side of the angle lies outside the arc and the other side is tangent to the arc. For example... Figure 3a (DCF in the middle angle)

[0127] The two endpoints of a circular arc are the critical points where an angle can be tangent to a convex circular arc. For example... Figure 3bAs shown in the diagram, when the CF side of the angle DCF is tangent to the endpoint A of the arc, Vc is the tangent vector Vs of the endpoint A; when the CF side of the angle DCF is tangent to the endpoint B of the arc, the side vector Vp of CF is the tangent vector Ve' of the endpoint B. Similarly, when the CD side of the angle DCF is tangent to the endpoint B of the arc, the side vector Vc of CD is the tangent vector Ve of the endpoint B; when the CD side of the angle DCF is tangent to the endpoint A of the arc, the side vector Vc of CD is the tangent vector Vs' of the endpoint A.

[0128] In summary, to determine whether a convex circular arc and an angle can contact each other, we can calculate the tangent vectors Ve', Vs, Ve, and Vs' at the two endpoints of the convex circular arc, forming a vector interval N, as shown below. Figure 4b As shown in the diagram. The side vectors are calculated from the two opposite sides, forming the side vector interval K, as shown in the diagram. Figure 4c As shown in the diagram. When there is an intersection between N and K, the convex arc and the corner can make contact. (See diagram below.) Figure 4d As shown in the diagram, if both Vc and Vp are located between Vs and Ve, the convex arc does not intersect with either side of the angle, and the contact range between the convex arc and the angle is the entire convex arc. If Vc is located between Ve and Vs', during the sliding process of the convex arc relative to the angle, the CD side of the angle will intersect with a portion of the arc, and the closer Vc is to Vs', the smaller the contact range between the convex arc and the angle, and the closer the critical contact point Q is to point A, that is, the contact range between the convex arc and the angle is a portion of the convex arc AQ. If Vp is located between Ve' and Vs, during the sliding process of the convex arc relative to the angle, the CF side of the angle will intersect with a portion of the arc, and the closer Vp is to Ve', the smaller the contact range between the convex arc and the angle, and the closer the critical contact point Q is to point B, that is, the contact range between the convex arc and the angle is a portion of the convex arc QB.

[0129] Contact detection for concave arcs and corners: Similarly, for the concave arc, obtain the tangent vectors Vs, Ve', Vs', and Ve at both endpoints, as follows: Figure 5b As shown in the diagram. If both side vectors Vc and Vp of the angle are located between Vs' and Ve', the concave arc does not intersect with either side of the angle, and the contact range between the concave arc and the angle is the entire concave arc. If Vc is located between Ve and Vs' or Vp is located between Ve' and Vs, the contact range between the angle and the concave arc is a portion of the concave arc.

[0130] When an arc is considered as a special type of angle, the methods for "contact detection between arcs" and "contact detection between arc and angle" are similar, so the detection principle will not be explained here.

[0131] Steps 3-4: The trajectory lines are generated by moving the shoe pieces to be arranged relative to the partial boundary of the area to be arranged. The innermost polygon of the trajectory line set is taken to obtain an incomplete critical polygon.

[0132] Specifically, when the corner or arc of the shoe piece to be arranged contacts a certain boundary of the area to be arranged, the contact point between the shoe piece to be arranged and the corresponding boundary of the area to be arranged is determined. Keeping the position of the contact point of the shoe piece to be arranged unchanged, the shoe piece to be arranged is moved so that the contact point of the shoe piece to be arranged slides along the corresponding boundary from one end to the other end. The sliding trajectory of the reference point F of the shoe piece to be arranged during the movement is obtained as part of the trajectory line of the critical polygon. There are no limitations on the reference point, and it can be determined as needed. It can be any point on the shoe piece to be arranged. In this embodiment, the center of gravity of the shoe piece is used.

[0133] Alternatively, when the corner or arc of a certain boundary of the area to be arranged can contact the straight edge or arc of the shoe piece to be arranged, keep the contact point position of that boundary of the area to be arranged unchanged, move the shoe piece to be arranged so that the contact point on the boundary of the area to be arranged slides along one side of the shoe piece from one end to the other end, and obtain another part of the sliding trajectory of the reference point F of the shoe piece to be arranged during the movement as another part of the trajectory line of the critical polygon.

[0134] Specifically, after determining that the arc can contact a straight edge or corner, a trajectory line is generated by moving the shoe piece to be arranged between the boundary of the area to be arranged. The following uses an arc and a straight edge as examples to illustrate how to generate the trajectory line. Figure 6a As shown in the diagram, assume shoe piece I is the already arranged shoe piece, with its arc edge forming a boundary segment of the area to be arranged, and shoe piece II is the shoe piece to be arranged. During the arrangement process, the already arranged shoe pieces are fixed in position and cannot be rotated. At a certain angle, the arc and the straight edge have only one fixed contact point, namely the tangent point between the right-angled side and the arc. Using the tangent point as the contact point, the arc slides from the starting point C of the straight edge to the ending point D. During this process, the reference point F also slides accordingly. The sliding trajectory FF' of point F is a portion of the trajectory line required to generate the critical polygon.

[0135] like Figure 6b As shown in the diagram, FF' is the trajectory line generated by the sliding of angle CDK relative to the arc AB. During the sliding process, the vertex D of angle CDK starts from the endpoint A of the arc and slides along the arc to the endpoint B. The trajectory FF' traversed by the reference point F is a portion of the trajectory line we need.

[0136] After generating the sliding trajectory lines of all edges of the shoe pieces to be arranged and the lower left boundary of the area to be arranged, a set of trajectory lines can be obtained, such as... Figure 7 As shown. Taking the innermost trajectory line from the set of trajectory lines yields an incomplete critical polygon, such as... Figure 8 As shown.

[0137] Since the upper region of the incomplete critical polygon can be used as the layout area for shoe pieces, in order to maximize the utilization rate, this patent selects the outline of the critical polygon for layout.

[0138] Steps 3-5: The shoe pieces are placed from the left and bottom of the area to be arranged. The trajectory line set is obtained from the lower left side of the area to be arranged, and the incomplete critical polygon is used as the area to be arranged.

[0139] Specifically, since shoe pieces are arranged from the bottom of the leather upwards during the actual nesting process, only the left and bottom sides of the area to be nested will actually be included in the shoe pieces. This patent improves upon the traditional critical polygon algorithm by calculating the trajectory line set only for the areas where shoe pieces may be nested, generating an incomplete critical polygon as the nestable area, such as... Figure 8 The area above the green curve in the diagram. This improvement can enhance sorting efficiency.

[0140] Furthermore, the shoe piece positioning strategy in step 4 includes:

[0141] (1) Irregular Shoe Piece Positioning Strategy: Due to the high degree of irregularity in the shape of irregular shoe pieces, a standard single-row arrangement method is adopted. The lowest point of the center of gravity in the area that can be arranged is used as the entry point for the new shoe piece. This allows some small gaps to be transferred to the far right to accommodate smaller shoe pieces. The entry angle is tried sequentially from 0°, 60°, 120°, 180°, 240°, and 300°. The area of ​​the wasted area on the left and bottom sides of each angle is used as the selection criterion, and the angle with the least waste is selected as the entry angle for the new shoe piece. After placement, the already arranged shoe pieces are merged to generate a new area to be placed, waiting for the next shoe piece to be placed, until the leather can no longer accommodate the current shoe piece; Figure 11 As shown.

[0142] (2) Positioning Strategy for Symmetrical Concave Shoe Pieces: To make good use of the shape characteristics of symmetrical concave shoe pieces and avoid excessive waste, a combination of single-row and double-row arrangement is adopted. That is, shoe pieces in two adjacent rows are aligned, and adjacent shoe pieces in the same row are arranged alternately, combining concave and gap into large spaces to accommodate smaller shoe pieces, thus reducing material waste. Furthermore, because the angle between adjacent shoe pieces is fixed, the relative arrangement positions of two adjacent shoe pieces can be calculated offline, reducing the time spent on online calculations.

[0143] Specifically, the above arrangement is a regular arrangement, but the shape of natural leather is irregular. To achieve a regular arrangement of shoe pieces, this invention fits a straight line from the bottom edge contour of the leather as the baseline for the layout, replacing the bottom contour of the cowhide; calculates the angle between the baseline and the horizontal line, thus determining the entry angle of the symmetrical concave shoe piece; and determines the internal no-fit polygon (INFP) between the shoe piece to be arranged and the left edge of the leather and the baseline, that is, the arrangeable area on the left side and bottom of the leather, and uses the lower left point of the arrangeable area as the entry position of the symmetrical concave shoe piece; for example... Figure 13 As shown.

[0144] (3) Positioning Strategy for Symmetrical Convex Shoe Pieces: Convex shoe pieces only have symmetrical features and no large concave areas that can be reused. In this invention, the symmetrical convex shoe pieces adopt an opposite single-row arrangement, meaning adjacent shoe pieces in the same row are arranged alternately, and adjacent rows do not need to be aligned. The strategy for determining the entry angle and entry point is consistent with that for symmetrical concave shoe pieces, and the relative arrangement positions of adjacent shoe pieces in the opposite single-row arrangement can be calculated offline, such as... Figure 14 As shown.

[0145] (4) Positioning Strategy for Strip-Shaped Shoe Pieces: Strip-shaped shoe pieces are relatively regular in shape, close to or exactly rectangular. Arranging strip-shaped shoe pieces in a single direction using a standard single-row method minimizes gaps. The strategy for determining the entry point and entry angle is the same as for concave shoe pieces, such as... Figure 15 As shown.

[0146] Furthermore, in step 5, after arranging the shoe pieces with large W values, due to the irregular shape of the leather, gaps will remain after arranging all types of shoe pieces. To minimize material waste, we use shoe pieces with small W values ​​to fill the remaining gaps, such as... Figure 16 As shown.

[0147] Shoe Piece Layout Results Test: Table 1 lists the test data for various shoe pieces laid out manually and automatically using the algorithm of this patent. The comparison shows that the algorithm of this patent significantly improves utilization compared to manual layout. The utilization rate of strip shoe pieces is increased by as much as 16.01%, and the utilization rates of irregular, symmetrical-concave, and symmetrical-convex shoe pieces are also significantly improved, reaching 7.72%, 4.96%, and 9.89% respectively. Secondly, this patent also has advantages in layout speed. Symmetrical strip shoe pieces are the most efficient, requiring only 0.91 seconds to lay out one piece. Irregular shoe pieces take the longest time to lay out, but each piece still only requires 1.62 seconds.

[0148] Table 1 Comparison of Manual and Algorithmic Nesting Data

[0149]

[0150] This invention employs the above technical solution, utilizing the ResNet18 algorithm to automatically classify shoe piece shapes into symmetrical concave, symmetrical convex, strip, and irregular shapes. Next, based on the shoe piece category, the insertion order and method (whether to arrange them individually or nested) are determined. Then, an improved incomplete critical polygon algorithm is used to detect overlaps in shoe pieces with curved surfaces, generating a suitable area for arrangement. Finally, a positioning strategy is used to determine the placement method, position, and angle of the shoe pieces, thus determining their location within the suitable area. After shoe pieces with large W values ​​are arranged, shoe pieces with smaller W values ​​are inserted to fill the gaps created by the large shoe pieces, ultimately achieving online automatic arrangement of shoe pieces on the leather. After arrangement, the arrangement result diagram, arrangement utilization rate, and the number of shoe pieces arranged are output.

[0151] This invention, based on the incomplete critical polygon algorithm, can not only solve the problem of overlapping detection of shoe pieces with rounded edges, but also reduce the overlapping detection of useless areas, thus greatly improving the efficiency of shoe piece layout.

[0152] Obviously, the described embodiments are only a part of the embodiments of this application, not all of them. Without conflict, the embodiments and features in the embodiments of this application can be combined with each other. The components of the embodiments of this application described and illustrated herein can generally be arranged and designed in various different configurations. Therefore, the detailed description of the embodiments of this application is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

Claims

1. A shoe piece layout method based on an incomplete critical polygon algorithm, characterized in that: It includes the following steps: Step 1: Obtain the shoe piece image and use the ResNet18 algorithm to automatically classify the shoe piece shape. This includes the following steps: Step 1-1: Extract the contour features of the shoe piece through convolution operation, and use the ReLU activation function to increase the non-linearity of the neural network, increase the expressive power of the neural network, and improve the final classification ability of the model. Steps 1-2: Reduce the size of the shoe piece feature map through pooling operations while preserving the main features of the shoe piece; Steps 1-3 address the vanishing gradient problem during model training using four residual blocks, enabling the network to learn shoe features more deeply and effectively; each residual block contains several convolutional layers and a ReLU activation function. Steps 1-4: Global average pooling is used to average the values ​​of each channel of the feature map, thereby converting the shoe piece feature map into a fixed-length vector. Steps 1-5: Connect the output of global average pooling to a fully connected layer and use softmax to classify the shoe pieces; Step 2: Determine the layout weight W of the shoe pieces based on their classification and area, and determine the placement order and method. Placement methods include individual placement and nested placement. Specific sorting steps are as follows: Step 2-1: Sort all shoe pieces in descending order of area, where N is the position of the shoe piece to be sorted in the sequence, and N is used as the area weight. Step 2-2: Assign shape weight M to the shoe pieces according to the shoe piece category. Assign different shape weights to irregular shapes, symmetrical concave shapes, symmetrical convex shapes, and strip shapes. Among them, the shape weights of irregular shapes and symmetrical concave shapes are greater than the weights of other shapes. Steps 2-3: Calculate the concave area of ​​all symmetrical concave shoe pieces, and take the largest concave area as the threshold. Shoe pieces with an area less than the threshold are classified as nestable shoe pieces. Steps 2-4: Calculate the maximum length and width of the concave part of the symmetrical concave shoe piece, traverse the nestable shoe pieces to search for small shoe pieces used to embed the concave part, extract the searched small shoe pieces from the sequence, combine them with the corresponding symmetrical concave shoe pieces, and recalculate the area and area weight N of the combined shoe pieces. Steps 2-5: Calculate the layout weight W=N×M for all shoe pieces, and finally determine the shoe piece placement order according to the layout weight W value from large to small; then select small shoe pieces to fill the gaps left by the large area of ​​shoe piece layout, starting from the smallest W value to the largest. Step 3: Use the incomplete critical polygon algorithm to perform overlap detection on the shoe pieces and generate a layable area for the shoe pieces. A layable area refers to a region where the shoe piece to be laid out will not overlap with already laid-out shoe pieces, nor will it exceed the boundaries of the leather. Specifically, this includes the following steps: Step 3-1: Based on the traditional critical polygon algorithm based on trajectory lines, add a circular arc contact detection strategy, and treat the circular arc as a special angle to detect the contact between the shoe piece to be arranged and the boundary of the area to be arranged; Step 3-2: Determine whether the tangent vector of the straight edge falls within the tangent vector interval at both ends of the arc; if so, determine that the arc and the straight line can contact each other; otherwise, the arc and the straight line cannot contact each other. Step 3-3, Contact detection between arc and corner: Arcs include convex arcs and concave arcs. Determine whether the arc is convex. If it is, perform contact detection between the convex arc and the corner; otherwise, perform contact detection between the concave arc and the corner. Steps 3-4: Merge the already laid-out shoe pieces at the bottom of the cowhide into a laid-out area, and fit the upper boundary of the laid-out area with the left contour of the cowhide part to form a new area to be laid out; Steps 3-5: The shoe pieces are placed from the left or bottom of the area to be arranged. The shoe pieces are moved relative to the left or bottom of the area to be arranged to generate trajectory lines. The set of trajectory lines is obtained, the innermost trajectory line is extracted, and the generated incomplete critical polygon is used as the area to be arranged. Step 4: Use positioning strategies to determine the placement method, position, and angle of the shoe pieces, and determine the position of the shoe pieces in the layout area; Step 5: After the shoe pieces with higher weights are arranged, shoe pieces with lower weights are then inserted to fill the gaps created by the arrangement of the larger shoe pieces, thus achieving online automatic arrangement of shoe pieces on the leather. Step 6: After the layout is completed, output the layout result diagram, layout utilization rate and layout quantity of each shoe piece.

2. The shoe piece layout method based on the incomplete critical polygon algorithm according to claim 1, characterized in that: In step 1, the shoe pieces are categorized into irregular, symmetrical, and striped shapes. The symmetrical shapes are further divided into two types: symmetrical concave and symmetrical convex.

3. The shoe piece layout method based on the incomplete critical polygon algorithm according to claim 1, characterized in that: In steps 2-5, if the W values ​​of two shoe pieces are the same, the principle of prioritizing area weight is adopted, and the piece with the larger area weight is included first.

4. The shoe piece layout method based on the incomplete critical polygon algorithm according to claim 1, characterized in that: The contact detection steps between the convex arc and the corner in step 3-3 are as follows: Step 3-3-11: For the convex circular arc, obtain the tangent vectors Ve', Vs, Ve and Vs' at the two endpoints to form the vector interval N; at the same time, obtain the side vectors Vp and Vc on the two opposite sides to form the side vector interval K. Where Vs represents the tangent vector at one endpoint when one side of the angle is tangent to one endpoint of the arc; Vs' represents the tangent vector at one endpoint when the other side of the angle is tangent to one endpoint of the arc; Ve' represents the tangent vector at the other endpoint when one side of the angle is tangent to the other endpoint of the arc; Ve represents the tangent vector at the other endpoint when the other side of the angle is tangent to the other endpoint of the arc; the side vector of one side of the angle is Vp; the side vector of the other side of the angle is Vc. Step 3-3-12: Determine whether there is an intersection between vector interval N and vector interval K; if so, determine that the convex arc and the corner can contact each other; otherwise, determine that the convex arc and the corner cannot contact each other. Step 3-3-13: When Vc and Vp are both located between Vs and Ve, the convex arc does not intersect with either side of the angle, and the contact range between the convex arc and the angle is the entire convex arc. When Vc is located between Ve and Vs', during the sliding process of the convex arc relative to the angle, one side CD of the angle will intersect with a portion of the arc. The closer Vc is to Vs', the smaller the contact range between the convex arc and the angle. The position of the critical contact point Q is closer to point A, that is, the contact range between the convex arc and the angle is a portion of the convex arc AQ. When Vp is located between Ve' and Vs, during the sliding process of the convex arc relative to the angle, the other side CF of the angle will intersect with a portion of the arc. The closer Vp is to Ve', the smaller the contact range between the convex arc and the angle. The position of the critical contact point Q is closer to point B, that is, the contact range between the convex arc and the angle is a portion of the convex arc QB.

5. The shoe piece layout method based on the incomplete critical polygon algorithm according to claim 1, characterized in that: The steps for detecting the contact between the concave arc and the corner in step 3-3 are as follows: Step 3-3-21: Calculate the tangent vectors Vs, Ve', Vs' and Ve at the two endpoints of the concave circular arc; Step 3-3-22: When the two side vectors Vc and Vp of the angle are both located between Vs' and Ve', and the concave arc does not intersect with either side of the angle, the range that the concave arc can contact the angle is the entire concave arc. Step 3-3-23: When the side vector Vc of the angle is located between Ve and Vs' or the side vector Vp is located between Ve' and Vs, there are 1 or 2 tangent points between the side of the angle and the arc. The range in which the angle can contact the concave arc is a partial concave arc. A partial concave arc refers to the part between the tangent points or the part between the endpoint and the tangent point.

6. The shoe piece layout method based on the incomplete critical polygon algorithm according to claim 1, characterized in that: The specific method for steps 3-4 is as follows: When the corner or arc of the shoe piece to be arranged contacts a certain segment of the boundary of the area to be arranged, determine the contact point between the shoe piece to be arranged and the corresponding segment of the boundary of the area to be arranged. Keep the position of the contact point of the shoe piece to be arranged unchanged, move the shoe piece to be arranged so that the contact point of the shoe piece to be arranged slides along the corresponding segment boundary from one end to the other end, and obtain the sliding trajectory of the reference point F of the shoe piece to be arranged during the movement as part of the trajectory line of the critical polygon. Alternatively, when the corner or arc of a certain boundary of the area to be arranged can contact the straight edge or arc of the shoe piece to be arranged, keep the contact point position of that boundary of the area to be arranged unchanged, move the shoe piece to be arranged so that the contact point on the boundary of the area to be arranged slides along one side of the shoe piece from one end to the other end, and obtain another part of the sliding trajectory of the reference point F of the shoe piece to be arranged during the movement as another part of the trajectory line of the critical polygon.

7. The shoe piece layout method based on the incomplete critical polygon algorithm according to claim 1, characterized in that: Step 4, the shoe piece positioning strategy, includes: (1) Irregular shoe piece positioning strategy: Irregular shoe pieces are arranged in a normal single row, and the lowest point of the center of gravity in the area that can be arranged is used as the entry point of the new shoe piece; the entry angle is tried in turn from 0°, 60°, 120°, 180°, 240° and 300°, and the waste area on the left and bottom sides of each angle is used as the selection criterion. The angle with the least waste is selected as the entry angle of the new shoe piece; after the shoe piece is placed, the arranged shoe pieces are merged to generate a new area to be arranged and wait for the next shoe piece to be placed, until the current shoe piece cannot be placed inside the leather. (2) Positioning strategy for symmetrical concave shoe pieces: Symmetrical concave shoe pieces adopt a combination of single-row and double-row arrangement, that is, shoe pieces in two adjacent rows are aligned and arranged, and two adjacent shoe pieces in the same row are arranged alternately, combining concave and gap into a large space to fit small shoe pieces, reducing material waste; The strategy for determining the entry point and entry angle is as follows: fit a straight line from the bottom edge contour of the leather as the baseline for the layout, replacing the bottom contour of the cowhide; calculate the angle between the baseline and the horizontal line, which determines the entry angle of the symmetrical concave shoe piece; find the internal critical polygon between the shoe piece to be laid out and the left edge of the leather and the baseline, which means finding the layout area on the left side and bottom of the leather, and take the lower left side point of the layout area as the entry position of the symmetrical concave shoe piece. (3) Positioning strategy for symmetrical convex shoe pieces: Symmetrical convex shoe pieces adopt a head-to-head single-row method, that is, adjacent shoe pieces in the same row are arranged alternately in front and back, and adjacent rows do not need to be aligned; the strategy for determining the entry angle and entry point is the same as that for symmetrical concave shoe pieces, and the relative arrangement position of adjacent shoe pieces in head-to-head single-row is calculated offline; (4) Positioning strategy for strip shoe pieces: The gaps generated by the regular arrangement of strip shoe pieces in a single direction are minimal. The strategy for determining the entry point and entry angle is the same as that for symmetrical concave shoe pieces.

Images