Protective shoe AI generation method based on physical process constraint and safety standard verification

Abstract

The application discloses a protective shoe AI generation method based on physical process constraints and safety standard verification, relates to the technical field of product design automation, and comprises the following steps: assembling the topological relationship of the protective component and the material process constraint based on the scale conversion benchmark and the determinable rule table, binding the rules according to the functional partition, forming a parameterized constraint set, generating a shoe scheme according to the parameterized constraint set and the appearance intention, performing semantic segmentation, key point positioning and edge extraction measurement and verifying the rules, when the verification result indicates that the shoe scheme does not conform to the determinable rule table and the parameterized constraint set, sequentially performing local shape correction, local redrawing generation and discarding regeneration, and obtaining a final scheme, rechecking the final scheme, outputting a multi-view design drawing, a partition annotation drawing, a vectorized contour line, a key dimension table, a material list and a process specification, and performing encryption and decryption, clearing the plaintext and recording the access log.

Description

Technical Field

[0001] This invention relates to the field of product design automation technology, and in particular to an AI-generated method for protective footwear based on physical process constraints and safety standard verification. Background Technology

[0002] In the field of labor protection equipment and footwear industrial design, the research and development of protective footwear is usually supported by computer-aided design. Through digital 3D models and parametric modeling, the collaborative design of shoe lasts, sole shapes, upper segmentation lines and protective components is realized. Combined with material databases and process information, material selection, structural manufacturability verification and design document output are completed. In recent years, data-driven and generative technology-based assisted design has been gradually applied to footwear appearance creation and rapid iteration of solutions. It can improve design efficiency and customization capabilities while maintaining the consistency of style expression, and supports multi-view display, outline and size table generation, which facilitates design review, prototyping communication and production connection.

[0003] However, in existing methods, safety standards are mostly incorporated into the design process in the form of textual clauses, which are difficult to directly transform into rules that can be automatically judged and form a consistent verification loop with the measurement results of the design drawings. At the same time, material and process constraints and the spatial relationship of protective components are often not sufficiently coupled with the appearance generation process, and the appearance changes are prone to being out of sync with manufacturability and structural relationships during scheme iteration. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an AI generation method for protective footwear based on physical process constraints and safety standard verification to solve the problems of safety standard clauses being difficult to transform into judgmentable rules and form a verification closed loop with the measurement results of design drawings, as well as the difficulty in synchronizing and maintaining consistency between material process constraints and the spatial relationship of protective components with appearance generation during scheme iteration.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention provides an AI generation method for protective footwear based on physical process constraints and safety standard verification, comprising: Collect design requirements for protective footwear, analyze the appearance intent and determine the applicable standard clauses, compile the applicable standard clauses into a judgment rule table, and establish a scale conversion benchmark.

[0007] Based on the scale conversion benchmark and the decisionable rule table, the topological relationship and material process constraints of the assembled protective components are combined, and the rules are bound according to the functional partition to form a set of parameterized constraints.

[0008] The footwear design is generated based on the parametric constraint set and appearance intent. During the generation process, semantic segmentation, key point localization, and edge extraction measurement are performed, and the rules are verified. When the verification result indicates that the footwear design does not conform to the decisionable rule table and parametric constraint set, local shape correction, local redrawing and generation, and discarding and regenerating are performed in sequence to obtain the final design.

[0009] The final design is reviewed, and outputs multi-view design drawings, partition annotation drawings, vectorized outlines, key dimension tables, bill of materials and process specifications. The design is then encrypted and decrypted, and plaintext is cleared and access logs are removed.

[0010] As a preferred embodiment of the protective footwear AI generation method based on physical process constraints and safety standard verification described in this invention, the specific steps of collecting protective footwear design requirements, parsing appearance intent and determining applicable standard clauses are as follows: collecting protective footwear design requirements through field forms, solidifying the fields into design requirement records, preprocessing the reference images in the design requirement records to obtain standard views, extracting the outer contour of the sole from the edges, and determining the toe based on curvature characteristics.

[0011] The text descriptions in the design requirements record are segmented and matched and merged according to the appearance intent dictionary. When the text descriptions do not cover the appearance intent, the main outline direction, texture density and dividing line distribution of the upper and sole are extracted. The expected protection items and usage scenarios in the design requirements record are mapped to applicable standard clauses according to the mapping table.

[0012] As a preferred embodiment of the protective footwear AI generation method based on physical process constraints and safety standard verification described in this invention, the specific steps of organizing the applicable standard clauses into a determinate rule table and establishing a scale conversion benchmark are as follows: the applicable standard clauses are split into determinate rule table row records, each row record consisting of the name of the measured item, comparison relationship, standard requirement value, and area of ​​effect, and only the measured items that can be obtained by semantic segmentation, key point positioning, and edge extraction are retained. The standard requirement values ​​are length diameter, area diameter, proportion, and material file attribute diameter, and the comparison relationship is not less than, not greater than, equal to, within an interval, and within a set.

[0013] The cumulative distance along the outer contour between the foremost contour point of the shoe toe and the last contour point of the heel is used as the pixel quantity of the sole length, and the ratio of the measured last length to the pixel quantity of the sole length is used as the scale conversion benchmark.

[0014] As a preferred embodiment of the AI ​​generation method for protective footwear based on physical process constraints and safety standard verification described in this invention, the assembly of protective components' topological relationships and material process constraints includes: reading the functional areas row by row from the determineable rule table and merging them into anti-impact areas, anti-puncture areas, electrical performance areas, and anti-slip areas; writing the merged results back to the functional area field of the row as functional partition names; writing the anti-impact areas, anti-puncture areas, electrical performance areas, and anti-slip areas into the design requirement record as a partition list; reading the expected protective items in the design requirement record; forming a set of anti-impact related protective components and a set of anti-puncture related protective components and writing them back to the design requirement record; assembling the topological relationships with the outer contour of the sole as a geometric reference; and establishing internal inclusion relationships, covering relationships, and non-intersecting relationships.

[0015] Read the material preferences and process preferences from the design requirements record, determine the unique material record, write it into the design requirements record as a material and process constraint field, and associate the material attribute fields according to the functional partition.

[0016] As a preferred embodiment of the protective footwear AI generation method based on physical process constraints and safety standard verification described in this invention, the step of binding rules according to functional partitions to form a parameterized constraint set includes: traversing each row of the determinate rule table, binding the row to the functional partition in the design requirement record according to the area of ​​action, establishing a correspondence between the row rules and the topology relationship, solidifying the measurement caliber description for the row rules and writing it back to the row.

[0017] The partition list, topology relationship, material and process constraint fields, and decision rule table are merged and stored into a parameterized constraint set, which is then bound to the design requirement record, and the coverage ratio is written into the parameterized constraint set.

[0018] As a preferred embodiment of the protective footwear AI generation method based on physical process constraints and safety standard verification described in this invention, the specific steps of generating a footwear scheme according to a set of parameterized constraints and appearance intent are as follows: the outer contour of the sole defines the outer boundary of the footwear scheme; the length direction is determined by the line connecting the toe end and the heel end, and a vertical dividing line is drawn at the midpoint of the line to form the corresponding area of ​​the toe direction and the corresponding range of the puncture-proof area; partitioning is generated according to the partitioning list, and anti-impact area, electrical performance area, puncture-proof area and anti-slip area are arranged; the set of anti-impact related protective components is placed in the anti-impact area; the set of puncture-proof related protective components is placed in the puncture-proof area; and the unique material record in the material process constraint field is associated with the footwear scheme according to the functional partition.

[0019] Generate the line style, texture density, segmentation complexity, and surface texture of the footwear design without changing the outer contour and topological relationship of the sole.

[0020] As a preferred embodiment of the protective footwear AI generation method based on physical process constraints and safety standard verification described in this invention, the generation process includes semantic segmentation, key point localization, and edge extraction measurement and verification rules. The specific steps are as follows: semantic segmentation is performed on the footwear design, outputting a partition mask consistent with the partition list and outputting masks for the sets of anti-impact related protective components and anti-puncture related protective components. The partition occupancy boundary and the component set occupancy boundary are used as the initial mask boundary for fitting correction. Key point localization is performed on the footwear design to obtain the endpoint positions of the length-class measured items and determine the two endpoints by projection along the length direction. Edge extraction is performed on the footwear design to obtain the outer contour edge of the sole, the partition boundary edge, and the protective component set boundary edge. Actual measurements are formed according to the measurement caliber description of the decisionable rule table, and the coverage ratio is calculated.

[0021] Traverse the rule table to verify the actual measurements according to the comparison relationship and generate a failure list. Simultaneously verify the topological relationship and determine whether the rule rows related to the coverage ratio and coverage are passed. The footwear solution is deemed to pass when the failure list is empty and all topological relationship verifications are passed.

[0022] As a preferred embodiment of the protective footwear AI generation method based on physical process constraints and safety standard verification described in this invention, the steps of sequentially performing local shape correction, local redrawing generation, and discarding and regenerating to obtain the final solution are as follows: local shape correction modifies the partition boundary and the protective component set boundary within the effective area while keeping the outer contour of the sole unchanged and satisfying the internal inclusion relationship; after local shape correction, the measurement is remeasured and re-verified; if it fails, local redrawing generation is performed within the effective area; after local redrawing generation, the measurement is remeasured and re-verified; if it still fails, the footwear solution is discarded and the footwear solution is regenerated until it passes; when all solutions pass, the final solution is determined and written into the design requirement record.

[0023] As a preferred embodiment of the protective footwear AI generation method based on physical process constraints and safety standard verification described in this invention, the following steps are taken to review the final solution and output multi-view design drawings, partition annotation drawings, vectorized outlines, key dimension tables, material lists, and process descriptions: if the list is empty, a consistency review is performed; if the consistency review is passed, the verification records are summarized in the order of the rows in the decisionable rule table, and the multi-view design drawings, partition annotation drawings, vectorized outlines, key dimension tables, material lists, and process descriptions are output and written into the design requirement record.

[0024] As a preferred embodiment of the protective footwear AI generation method based on physical process constraints and safety standard verification described in this invention, the steps of encryption / decryption, plaintext clearing, and access log recording include: summarizing multi-view design drawings, partition annotation drawings, vectorized outlines, key dimension tables, material lists, and process specifications into an output package and binding it as a plaintext output package with the design requirement record; encrypting the plaintext output package to generate an ciphertext output package and writing it into the design requirement record; performing decryption verification on the ciphertext output package; after successful decryption verification, clearing the plaintext output package and temporary files, retaining only the ciphertext output package; and writing the access log of the output package generation, encryption, decryption verification, and plaintext clearing process into the design requirement record.

[0025] The beneficial effects of this invention are as follows: by collecting the design requirements of protective footwear, analyzing the appearance intent and determining the applicable standard clauses, organizing the applicable standard clauses into a rule table that can be judged and establishing a scale conversion benchmark, the consistency and traceability of verification are improved. By generating footwear schemes based on the parameterized constraint set and appearance intent, and performing semantic segmentation, key point positioning and edge extraction measurement and verification of rules, the partition list, topological relationship field and material process constraint field are kept synchronously valid, and the final scheme is obtained. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of an AI-based method for generating protective footwear based on physical process constraints and safety standard verification.

[0028] Figure 2 A flowchart for splitting and organizing the decision rule table.

[0029] Figure 3 A flowchart for merging and binding functional areas.

[0030] Figure 4 The flowchart for performing verification and judgment.

[0031] Figure 5 A comparison chart showing how the traceability completeness rate changes with the number of rows in the decisionable rule table.

[0032] Figure 6 This is a comparison chart showing how the final pass rate changes with the maximum number of iterations. Detailed Implementation

[0033] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0034] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0035] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0036] Reference Figures 1-6 This is one embodiment of the present invention, which provides an AI generation method for protective footwear based on physical process constraints and safety standard verification, including the following steps: S1. Collect design requirements for protective footwear, analyze the appearance intent and determine the applicable standard clauses, compile the applicable standard clauses into a judgment rule table, and establish a scale conversion benchmark.

[0037] Design requirements for protective footwear are collected through a field form. The fields are, in order, shoe size and last information, expected protection items, usage scenarios, material preferences, process preferences, reference images, and text descriptions. After collection, the fields are recorded and fixed according to the same design requirement. The reference images are preprocessed, including unifying orientation, resolution, and background cropping range. Unifying orientation is achieved by comparing the curvature characteristics of the outer contour of the sole at both ends along the length direction, and the end with the sharper curvature is identified as the toe. Unifying resolution is achieved by scaling the long side of the standard view while maintaining the aspect ratio, so that all measurements within the same task are performed at a uniform pixel scale. Unifying the background cropping range is achieved by using the smallest bounding rectangle of the outer contour of the sole as the cropping frame and cropping to obtain the standard view.

[0038] The outer contour of the sole refers to the closed contour obtained by edge extraction of the preprocessed reference image. The closed contour with the largest area is selected as the outer contour of the sole and written into the design requirements record. When the curvature features at both ends cannot uniquely determine the toe, the end with the greater outward convexity of the outer contour of the sole in the neighborhood of both ends is taken as the toe end. Local contour segments formed by the endpoints and the neighborhood contour points are taken in the neighborhood of both ends respectively. The maximum outward convex distance of the local contour segment relative to the endpoints is calculated. The end with the greater maximum outward convex distance is determined as the toe end.

[0039] Edge extraction refers to the process of applying edge detection algorithms (such as the Canny operator or the Sobel operator) to a preprocessed reference image to identify edge information in the image and generate an edge image. In the edge image, closed contours are extracted through morphological operations (such as closing operations) or contour detection algorithms (such as contour tracking algorithms).

[0040] After performing word segmentation on the text description, it is matched and merged according to the appearance intent dictionary. The matched words are merged into appearance intent description items and written into the design requirement record in order.

[0041] The appearance intent dictionary includes four categories of descriptions: line style, texture density, dividing line complexity, and surface texture. Based on the set of synonyms for each category of descriptions, different expressions are merged into the same appearance intent description.

[0042] When the text description does not cover the appearance intent, the main outline direction, texture density and dividing line distribution of the upper and sole are directly extracted from the reference image and written into the design requirement record as the same set of appearance intent description items. The main outline direction is obtained by statistically analyzing the main direction of the outer outline edge, the texture density is represented by the average spacing of the texture edge, and the dividing line distribution is represented by the skeleton line of the slender edge connected domain.

[0043] The main direction statistics are obtained by fitting the main direction of the outer contour point set to obtain the length direction, and the average spacing is calculated by the projection spacing of the texture edge in the direction perpendicular to the length direction.

[0044] Based on the expected protection items and usage scenarios in the design requirements record, determine the set of applicable standard clauses according to the mapping table, map each expected protection item to the corresponding clause, and map the usage scenario to the environmental application item of the clause. If there are parallel clauses for the same protection item, take the clause that covers the usage scenario and has more complete constraints as the applicable standard clause, and write the applicable standard clause into the design requirements record. When the parallel clauses are still indistinguishable in terms of the completeness of constraints, select the clause that contains a clear measurement object and can be obtained by semantic segmentation, key point localization and edge extraction measurement as the applicable standard clause.

[0045] A mapping table refers to a pre-defined format that includes the name of the expected protection item, the name of the clause, and the key points of the clause content (for example, for a protection item against impact, it may be mapped to a protection clause, which requires that the toe of the shoe must have at least 20mm thick protection material against impact, and the applicable scenarios of the clause include indoor operations and heavy object handling environments).

[0046] Semantic segmentation refers to the pixel-level classification of footwear design drawings using deep learning algorithms (such as U-Net, FCN, etc.), segmenting different regions in the image (such as soles, uppers, protective areas, etc.) and labeling them as different categories.

[0047] Keypoint localization refers to the precise location of key points in an image using computer vision algorithms (such as OpenPose).

[0048] The completeness of a constraint refers to the fact that, after breaking down parallel clauses into rows of determinate rule tables, the applicable standard clause with more rows is considered to have a more complete constraint.

[0049] The applicable standard clauses are broken down into rows of a decisionable rule table. Each row includes the name of the measurand, the comparison relationship, the standard requirement value, and the area of ​​application. During the compilation, only measurands that can be measured through semantic segmentation, key point localization, and edge extraction are retained. The standard requirement value in the decisionable rule table is represented by length for measurands of the length category, area for measurands of the area category, and ratio for measurands of the ratio category as a dimensionless ratio. The attribute category of measurands of the material attribute category follows the attribute category of the material file. The actual measurement and the standard requirement value of the same measurand are kept to use the same dimension category in the decisionable rule table. The measurement category correspondence is fixed for each category of measurands. The length category is obtained by converting the pixel length obtained by key point localization and edge extraction using the scale conversion benchmark. The area category is obtained by multiplying the pixel area of ​​the semantic segmentation mask by the square of the scale conversion benchmark. The ratio category is obtained by the ratio of the area or length of the same diameter. The material attribute category is obtained by reading the material file, which is a file table that records the material attributes by the material name.

[0050] Comparison relationships refer to the judgment relationships proposed by the standard clauses for the measured item. Comparison relationships are recorded in five categories: not less than, not greater than, equal to, within an interval, and within a set. "Not less than" indicates that the measured value meets or exceeds the standard requirement value; "not greater than" indicates that the measured value does not exceed the standard requirement value; "equal to" indicates that the measured value, after conversion according to the measurement caliber and processing with the same number of retained decimal places, is consistent with the standard requirement value; "within an interval" indicates that the measured value lies between the lower and upper limits of the standard requirement value and includes the boundary; and "within a set" indicates that the measured value belongs to the permissible set listed in the standard requirement value. When the standard clause states... When the standard clause states "at least," "must not be less than," or "should not be lower than," the comparison relationship is recorded as "not less than." When the standard clause states "at most," "must not be greater than," or "should not be higher than," the comparison relationship is recorded as "not greater than." When the standard clause states "should be" or "must be," the comparison relationship is recorded as "equal to." When the standard clause states "between," "within a range," or "within any interval," the comparison relationship is recorded as "within the interval," and the standard requirement value is written in the form of the lower limit sub-value and the upper limit sub-value on the same line. When the standard clause states "should adopt," "should belong to," or "should meet the material grade or category," the comparison relationship is recorded as "within the set," and the standard requirement value is written as the allowed set.

[0051] The standard requires that the value be a composite field, containing both a lower limit sub-value and an upper limit sub-value.

[0052] The cumulative distance along the outer contour between the foremost contour point of the shoe toe and the last contour point of the heel is used as the pixel quantity of the sole length. The pixel quantity of the sole length is obtained by accumulating the distance between adjacent contour points along the outer contour point sequence of the sole. The actual measured value of the last length corresponding to the shoe size and last information is read. The actual measured value of the last length is read from the last library. The last length is stored in the last library according to the shoe size and last information and bound to the design requirements record. The ratio of the actual measured value of the last length to the pixel quantity of the sole length is used as the scale conversion benchmark. The scale conversion benchmark represents the actual length corresponding to each pixel.

[0053] S2. Based on the scale conversion benchmark and the decisionable rule table, assemble the topological relationship and material process constraints of the protective components, bind the rules according to the functional partition, and form a set of parameterized constraints.

[0054] The effective areas are read row by row from the rule table. Each effective area is the original description of the area corresponding to the rule. The effective areas are merged into anti-impact areas, puncture-resistant areas, electrical performance areas, and anti-slip areas. The merged results are written back to the effective area field of the corresponding row as the functional partition name. For rows where the effective area is missing or cannot be merged, deterministic keywords are used to merge them according to the name of the measured item. If the name of the measured item contains "toe" or "front", it is merged into the anti-impact area; if the name of the measured item contains "puncture" or "coverage", it is merged into the puncture-resistant area; and if the name of the measured item contains "electrical", "insulation", or "antistatic", it is merged into the electrical performance area. The measured item name containing friction, anti-slip, or texture is merged into the anti-slip zone, and the merging result is written back to the current row's area field. When the measured item name matches multiple keywords, the first matching partition is taken as the unique merging result in the order of impact protection zone, puncture protection zone, electrical performance zone, and anti-slip zone. When the measured item name does not match any keywords, the functional partition corresponding to the first expected protection item in the design requirements record is taken as the unique merging result, and the impact protection zone, puncture protection zone, electrical performance zone, and anti-slip zone are written into the design requirements record as a partition list.

[0055] Read the expected protection items from the design requirements record, forming sets of impact-resistant and puncture-resistant protective components, and write them back to the design requirements record. The formation of the impact-resistant protective component set is based on rows in the determineable rule table where the effective areas are merged into impact-resistant zones, and the formation of the puncture-resistant protective component set is based on rows in the determineable rule table where the effective areas are merged into puncture-resistant zones. If no corresponding row exists, no corresponding set is formed. Using the outer contour of the sole written in the design requirements record as a geometric reference, assemble the topological relationships, establish the correspondence between the impact-resistant and impact-resistant protective component sets and the impact-resistant zones, and establish the internal inclusion relationships relative to the outer contour of the sole, so that... The set of anti-impact protective components is located inside the outer contour of the sole and in the corresponding area in the direction of the toe. For the set of anti-puncture protective components, a correspondence is established with the anti-puncture area, and a coverage relationship is established with the anti-puncture area, so that the set of anti-puncture protective components covers the corresponding range of the anti-puncture area inside the outer contour of the sole. For the sets of anti-impact protective components and the set of anti-puncture protective components that exist at the same time, a non-intersecting relationship is established, and it is determined that the projection areas of the sets of anti-impact protective components and the set of anti-puncture protective components in the outer contour of the sole do not overlap. For the electrical performance area and the anti-slip area, respectively, an internal inclusion relationship is established between the electrical performance area and the anti-slip area relative to the outer contour of the sole.

[0056] The projection region refers to the corresponding area within the plane where the preprocessed reference image is located.

[0057] The corresponding area in the toe direction and the corresponding range of the puncture-proof zone are defined by taking the line connecting the toe end and the heel end as the length direction, taking the midpoint of the line as the dividing line perpendicular to the length direction, and the dividing line intersecting with the inner part of the outer contour of the sole to form two parts. The side where the toe end is located is defined as the corresponding area in the toe direction, and the other side inside the outer contour of the sole is defined as the corresponding range of the puncture-proof zone.

[0058] Write the topology relationship into the topology relationship field of the design requirement record in the form of a quadruple of object 1, relationship type, object 2 and scope. Object 1 and object 2 are sets of anti-smashing protective components, sets of anti-puncture protective components, outer contour of shoe sole or functional area. The relationship type is internal containment relationship, covering relationship or disjoint relationship.

[0059] Read the material preferences and process preferences from the design requirements record, filter material records that match the material preferences from the material archive, and perform consistency screening on the process descriptions in the material archive based on the process preferences. If the process description field of the material archive contains the original text of the process preference field, it is considered consistent. When the process description field of the material archive contains the original text of multiple process preference fields, it is still considered consistent and the material record is retained. Extract the measurable items of the material attribute class from the decision rule table line by line to form a material attribute requirement list. Only retain the material records that can provide all the material attribute fields in the material attribute requirement list. If there are still multiple material records, select the material record with the most complete fields in the material archive as the unique material record and write it into the design requirements record as a material process constraint field. When there are ties for the most complete fields, take the first material record in the material archive record order as the unique material record and associate the material attribute fields of the material record according to the functional partition.

[0060] The most complete fields refer to the material file having the largest number of non-empty items in the material attribute field and process description field.

[0061] Iterate through each row of the rule table, bind the row to the corresponding functional partition in the design requirement record according to the area of ​​action, and establish a correspondence between the row rules and the topology relationship field. The anti-smashing zone rule is bound to the topology relationship quadruple of the anti-smashing related protective component set, the anti-puncture zone rule is bound to the topology relationship quadruple of the anti-puncture related protective component set, the electrical performance zone rule is bound to the internal inclusion relationship quadruple of the electrical performance zone relative to the outer contour of the sole, and the anti-slip zone rule is bound to the internal inclusion relationship quadruple of the anti-slip zone relative to the outer contour of the sole. The measurement diameter description is solidified for the row rules and written back to the row. The length class is obtained by converting the pixel length obtained by key point positioning and edge extraction through the scale conversion benchmark. The area class is obtained by multiplying the pixel area of ​​the semantic segmentation mask by the square of the scale conversion benchmark. The ratio class is obtained by the ratio of the area or length of the same diameter. The material attribute class is obtained by reading the material file.

[0062] The partition list, topology relationship field, material and process constraint field, and decision rule table are merged and stored into a parameterized constraint set, which is then bound to the design requirement record. Each rule includes the area of ​​effect, topology relationship, and measurement caliber description. The coverage ratio is written into the parameterized constraint set. The coverage ratio is the ratio of the projected area of ​​the puncture-resistant protective component set in the outer contour of the sole to the projected area of ​​the puncture-resistant area in the outer contour of the sole. Both types of projected areas are obtained by multiplying the pixel area of ​​the semantic segmentation mask by the square of the scale conversion benchmark.

[0063] S3. Generate a footwear design based on the parametric constraint set and appearance intent. During the generation process, semantic segmentation, key point localization, and edge extraction measurement are performed, and the rules are verified. When the verification result indicates that the footwear design does not conform to the decisionable rule table and parametric constraint set, local shape correction, local redrawing and generation, and discarding and regenerating are performed in sequence to obtain the final design.

[0064] The outer contour of the shoe sole defines the outer boundary of the footwear design. The footwear design is generated in the plane of the preprocessed reference image, and no visible contour or partition boundary exceeds the outer contour of the shoe sole.

[0065] Based on the corresponding area in the toe direction and the corresponding range of the puncture-resistant area, partitions are generated. The line connecting the toe end and the heel end is taken as the length direction, and the midpoint of the line is taken as the vertical dividing line. The dividing line intersects with the inner part of the outer contour of the sole to form two parts, front and back. The side where the toe end is located is the corresponding area in the toe direction, and the other side is the corresponding range of the puncture-resistant area. According to the partition list, the impact-resistant area and the electrical performance area are arranged in the corresponding area in the toe direction, and the puncture-resistant area and the anti-slip area are arranged in the corresponding range of the puncture-resistant area.

[0066] The structure placeholders are generated according to the topological relationship field while maintaining the relationship type. The set of anti-smashing related protective components is placed in the anti-smashing area and satisfies the internal inclusion relationship relative to the outer contour of the sole. The set of anti-puncture related protective components is placed in the anti-puncture area and satisfies the coverage relationship of the anti-puncture area. If both the set of anti-smashing related protective components and the set of anti-puncture related protective components exist, the projection areas of the set of anti-smashing related protective components and the set of anti-puncture related protective components are guaranteed not to overlap in the plane of the preprocessed reference image. The internal inclusion relationship relative to the outer contour of the sole is maintained for the electrical performance area and the anti-slip area, respectively.

[0067] Associate the unique material record in the material process constraint field with the footwear solution according to the functional partition.

[0068] Without altering the outer contour and topological relationship fields of the sole, the line style, texture density, segmentation line complexity, and surface texture of the footwear design are generated based on the appearance intent description. The footwear design is then written into the design requirement record. When writing into the design requirement record, the partition placeholder boundaries and the placeholder boundaries of the anti-impact related protective component set and the puncture-resistant related protective component set are also written.

[0069] Semantic segmentation is performed on the footwear design, and the output is a partition mask that is consistent with the partition list. The output is also a mask for the set of anti-impact protective components and the set of anti-puncture protective components. All masks are limited to the outer contour of the sole. The semantic segmentation uses the partition placeholder boundary and the component set placeholder boundary written in the design requirement record as the initial mask boundary. The initial mask boundary is only fitted and corrected near the edge of the partition boundary and the edge of the protective component set obtained by edge extraction. During the fitting and correction process, the partition name and the component set name remain unchanged.

[0070] Key point localization is performed on the footwear solution to obtain the endpoint positions corresponding to the length-class measurable items in the decisionable rule table. For any length-class measurable item, the boundary point set of the mask corresponding to the effective area is taken. The length direction is taken as the line connecting the toe endpoint and the heel endpoint. The boundary point set is projected in the length direction. The boundary point with the smallest projection and the boundary point with the largest projection are taken as the two endpoints of the measurable item. When the measurable item needs to be limited to the anti-smashing zone or the anti-puncture zone, the endpoints are only taken from the boundary point set of the corresponding partition mask. When the boundary point with the smallest projection or the boundary point with the largest projection is not unique, the boundary point closer to the toe endpoint is taken as the minimum projection endpoint and the boundary point closer to the heel endpoint is taken as the maximum projection endpoint.

[0071] Edge extraction is performed on the footwear design to obtain the outer contour edge of the sole, the boundary edge of the partition, and the boundary edge of the protective component set.

[0072] Actual measurements are generated one by one according to the measurement diameter description in the rule table and stored in correspondence with the rule rows. Length-type actual measurements are obtained by taking the cumulative distance between the two endpoints obtained from key point positioning on the corresponding boundary edge point sequence as the pixel length, and converting it into the actual length diameter according to the scale conversion benchmark. Area-type actual measurements are obtained by multiplying the pixel area of ​​the semantic segmentation mask by the square of the scale conversion benchmark. Proportion-type actual measurements are obtained by the ratio of the area of ​​the same diameter or the length of the same diameter. Material attribute-type actual measurements are obtained by reading the unique material record pointed to by the material process constraint field according to the material file.

[0073] If the effective area mask for any rule row fails to be generated, the endpoints cannot be uniquely determined, or the edge extraction fails to obtain the corresponding boundary, then the current rule row is marked as failed.

[0074] The coverage ratio is calculated based on the parametric constraint set, and the ratio of the projected area of ​​the puncture-resistant protective components set within the outer contour of the sole to the projected area of ​​the puncture-resistant zone within the outer contour of the sole is taken.

[0075] Traverse each row of the decision rule table, read the name of the measured item, comparison relationship, standard requirement value, and scope of application. Verify the actual measurement against the standard requirement value according to the comparison relationship. During verification, perform a decision on each rule row according to the comparison relationship. When the comparison relationship is not less than, the actual measurement reaches or exceeds the standard requirement value and is judged as pass; otherwise, it is judged as fail. When the comparison relationship is not greater than, the actual measurement does not exceed the standard requirement value and is judged as pass; otherwise, it is judged as fail. When the comparison relationship is equal to, the actual measurement and the standard requirement value are processed according to the same measurement caliber and judged as consistent and pass; otherwise, it is judged as fail. When the comparison relationship is within an interval, the actual measurement is not lower than the lower limit sub-value and not higher than the upper limit sub-value and is judged as pass; otherwise, it is judged as fail. When the comparison relationship is within a set, the actual measurement belongs to the allowed set listed by the standard requirement value and is judged as pass; otherwise, it is judged as fail. Record the rule rows that fail as a fail list.

[0076] Simultaneously verify the topological relationship fields in the parameterized constraint set. For internal inclusion relationships, check whether all corresponding objects are located inside the outer contour of the sole. For non-intersecting relationships, check whether the projection areas of the anti-impact related protective component set and the anti-puncture related protective component set overlap in the plane of the preprocessed reference image. Check whether the coverage ratio can be calculated and whether the projection area of ​​the anti-puncture zone in the outer contour of the sole is positive in the coverage ratio calculation. Then, determine whether the coverage ratio and the rule rows related to coverage in the decisionable rule table pass together.

[0077] If the failure list is empty and all topology relationship fields pass the verification, the footwear solution is deemed to pass. If any one of them fails, then local shape correction, local redraw generation, and discard and regenerate are executed in sequence.

[0078] Based on the scope of the area of ​​effect not specified in the list, modification of the partition boundary or the boundary of the protective component set is only allowed within the area of ​​effect. The outer contour of the sole remains unchanged, and the modification must still satisfy the corresponding internal inclusion relationship in the topology relationship field.

[0079] If the measured item does not pass through the length category, the difference between the actual measurement of the rule row and the standard requirement value is taken and converted into a pixel difference according to the scale conversion benchmark. Within the effective area, the corresponding boundary is moved along the perpendicular direction of the endpoint connection, causing the pixel length to change in the direction of the difference, until the rule row satisfies the comparison relationship. When moving the boundary, only the boundary point set of the effective area belonging to the rule row that has not passed through is allowed to be changed, while the boundary point set of the corresponding area of ​​the rule row that has passed through is kept unmoved. The boundary movement direction is determined by the comparison relationship. When the comparison relationship is not less than and the actual measurement is less than the standard requirement value, a boundary movement is executed to increase the corresponding actual measurement. When the comparison relationship is not greater than and the actual measurement is greater than the standard requirement value... When the comparison relationship is within the interval and the actual measurement is less than the lower limit sub-value, the movement is executed in the increasing direction. When the actual measurement is greater than the upper limit sub-value, the movement is executed in the decreasing direction. When the comparison relationship is equal, the movement is executed in the increasing direction when the actual measurement is less than the standard requirement value. When the actual measurement is greater than the standard requirement value, the movement is executed in the decreasing direction. When the comparison relationship is within the set, only when the rule behavior is a material attribute type measurand, the unique material record pointed to by the material process constraint field is replaced and re-verified. When the rule behavior is a non-material attribute type measurand, continuous movement is not used. The rule row is directly kept as failing and locally redrawn.

[0080] If the mask boundary is not expanded or contracted layer by layer within the effective area without the measurement item from the area class, the area of ​​the mask pixels will change monotonically. After each change, the actual area diameter will be converted to the square of the scale conversion reference for verification until the comparison relationship is met. The expanded or contracted area does not exceed the inner contour of the shoe sole. Expanding layer by layer means adding a ring of adjacent pixels outward along the mask boundary in the plane where the preprocessed reference image is located. Contracting layer by layer means reducing a ring of adjacent pixels inward along the mask boundary in the plane where the preprocessed reference image is located.

[0081] If not through disjoint relations, pruning is performed in the overlapping region according to the row order of the decision rule table, preserving the set region with the earlier row order, pruning the projection of the set with the later row order into the overlapping region, and keeping the pruned set still within its own partition.

[0082] If the rules from the coverage-related rules are not approved, the mask boundary of the set of puncture-resistant protective components is expanded within the corresponding range of the puncture-resistant zone to increase the projected area until the comparison relationship with the rule row related to the coverage is satisfied.

[0083] After completing the local shape correction, remeasure and re-verify.

[0084] If there are still failures after local shape correction, local redrawing will be performed in the area corresponding to the failure list. The outer contour of the sole, the verified partition boundaries and the verified topological relationship fields will remain unchanged. Only the line style, texture density, dividing line complexity and surface texture in the area will be regenerated according to the appearance intention description, and the unique material record pointed to by the material process constraint field will remain unchanged.

[0085] After the local redraw is completed, remeasure and re-verify.

[0086] If the design still fails after partial redrawing, the current footwear design is discarded, but the appearance intent description and parametric constraint set in the design requirements record are kept unchanged. A new footwear design is then generated and measured and verified again until it passes. If a successful design still cannot be obtained while keeping the parametric constraint set unchanged, the list of failed designs and the affected area obtained from the last verification are written into the design requirements record.

[0087] When the footwear design passes all the checks in the rule table, all the topological relationship fields, and all the measured items in the material attribute class are consistent with the unique material record pointed to by the material process constraint field, the current footwear design is determined as the final design, and the final design, along with the semantic segmentation results, key point localization results, and edge extraction results, is written into the design requirement record.

[0088] Figure 6 The paper presents a comparison of the final pass rates under different maximum iteration counts, where the horizontal axis represents the maximum iteration count and the vertical axis represents the final pass rate. The three curves correspond to only discarding and regenerating (A), local shape correction and discarding and regenerating (B), and local shape correction, local redrawing and regenerating, and discarding and regenerating (C), respectively. It can be seen that as the maximum iteration count increases, the final pass rate improvement using invention C is significantly higher than that of A and B. Furthermore, under the same iteration budget, it is easier to converge to the final solution that satisfies the decisionable rule table and parameterized constraint set. This indicates that when verification fails, this invention first performs local shape correction within the effective region based on the failure list to directly eliminate length-related, area-related, and coverage-related rules. For violations such as non-intersecting relationships that can be resolved through geometric adjustments, a local redraw generation is performed on the affected area that still fails. This redraws the line style, texture density, segmentation line complexity, and surface texture without changing the outer contour of the sole, the verified partition boundaries and topological relationship fields, and keeping the unique material record pointed to by the material and process constraint fields unchanged. This reduces the number of invalid regenerations while keeping the partition list, topological relationship fields, and material and process constraint fields synchronously valid. In contrast, strategies that lack local redraw generation or rely solely on discarding regenerations are less targeted at the reasons for failure, resulting in a slow increase in the pass rate as the number of iterations increases, making it difficult to stably obtain the final solution under a limited iteration budget.

[0089] S4. Review the final scheme, output multi-view design drawings, partition annotation drawings, vectorized outlines, key dimension tables, material lists and process descriptions, and perform encryption / decryption, plaintext clearing and access log recording.

[0090] The system reads the rejection list from the design requirements record. If the rejection list is not empty, it rolls back to discard and regenerates the solution to obtain the final solution again. After rewriting the verification results, it performs a consistency review. If the rejection list is empty, it performs a consistency review, comparing the final solution in the design requirements record with the semantic segmentation results, key point localization results, and edge extraction results to see if they correspond. It also checks if all masks are still limited to the outer contour of the sole. The system compares the number of rows in the rule table with the number of actual measurement entries written. It compares whether the material process constraint field referenced by the measured item in the material attribute class still points to the same unique material record. It compares whether the internal inclusion relationship, non-intersection relationship, and the mask of the effective area required for the coverage ratio calculation in the topology relationship field can still be obtained from the semantic segmentation results. Finally, it compares whether the projected area of ​​the puncture-proof zone within the outer contour of the sole in the coverage ratio calculation is positive.

[0091] After the consistency review is passed, the actual measurement will not be recalculated. The name of the measurand, comparison relationship, standard requirement value, area of ​​effect, actual measurement and verification result of each rule row will be summarized into a verification record in the order of the rule table rows that can be judged. The coverage ratio and the corresponding judgment result of the rule row related to coverage will also be written into the verification record. If the consistency review fails, it will be rolled back to discard and regenerate, and the final solution will be regenerated and rewritten into the verification result.

[0092] Using the final design's appearance in the plane of the preprocessed reference image as the main view, a multi-view design drawing is output. The main view includes the outer contour edge of the sole, the boundary edges of the partitions, and the boundary edges of the protective component assembly. The line connecting the toe end and the heel end is used as the length direction reference. The heel end is also labeled as the coordinate starting point and the length direction indicator. This multi-view design drawing is output alongside the main view. Figure 1 The auxiliary view is defined by the perpendicular dividing line between the length direction and the midpoint of the length direction, so that multiple views share the same reference system. The multi-view design drawing is written into the design requirement record and bound to the final solution.

[0093] Based on the semantic segmentation results, a partition annotation map is output. The partition masks of the anti-impact zone, anti-puncture zone, electrical performance zone, and anti-slip zone are outlined along the partition boundary edges, and the partition name is labeled inside each partition. At the same time, the masks of the anti-impact related protective component set and the anti-puncture related protective component set are outlined along the boundary edges of the protective component set and labeled with the set name. Then, according to the order of the rows of the determineable rule table, the name of the measured item and the rule row identifier are labeled near the corresponding action area, so that any rule row can be directly located to the action area from the annotation map. The partition annotation map is written into the design requirement record and bound to the final solution.

[0094] The rule row identifier is determined by the row order of the rule table and written to the verification record.

[0095] Based on the edge extraction results, a vectorized contour line is output. The outer contour edge of the sole, the boundary edge of the partition, and the boundary edge of the protective component set are connected in sequence according to the edge point sequence to form a vectorized contour line. A closure consistency check is performed on the outer contour of the sole to ensure that a closed curve is formed. An inclusion relationship check is performed on the boundary of the partition and the boundary of the protective component set to ensure that they both fall inside the outer contour of the sole. The vectorized contour line and the key point positioning results are written into the design requirement record.

[0096] The critical dimension table is output in the order of the rule table rows. The name of the measured item, the area of ​​effect, the comparison relationship, the standard requirement value, the actual measurement and the verification result are written into the critical dimension table row by row. The actual measurement of length and area is the actual diameter after conversion according to the scale conversion benchmark. The actual measurement of ratio is the dimensionless diameter. The actual measurement of material property is the value read from the material file. For the rule row with the standard requirement value being a composite field, the lower limit sub-value and the upper limit sub-value are presented in the same row in the form of field. The rule row identifier is written into the critical dimension table and is consistent with the same rule row identifier in the verification record, so that the critical dimension table and the verification record are corresponding. The critical dimension table is written into the design requirement record and bound to the verification record.

[0097] Figure 5The results of the traceability integrity rate comparison under different numbers of rows in the decisionable rule table are presented. The horizontal axis represents the number of rows in the decisionable rule table, and the vertical axis represents the average traceability integrity rate. The two curves correspond to the present invention and the control scheme, respectively. It can be seen that the present invention maintains a nearly stable high traceability integrity rate as the number of rows in the decisionable rule table expands from few to many. This indicates that the present invention organizes the applicable standard clauses into a decisionable rule table and does not recalculate the actual measurement after the consistency review is passed. Instead, it summarizes the name of the measurand, the comparison relationship, the standard requirement value, the area of ​​effect, the actual measurement, and the verification result into a verification record and outputs a key dimension table according to the row order of the decisionable rule table. This makes the rule row, the actual measurement item, and the verification result stably correspond under the same row identifier, which significantly improves the verification traceability and reduces the alignment deviation risk when the rule scale expands. The curve of the control scheme decreases significantly with the increase of the number of rule rows, indicating that when there is a lack of row-level alignment and review mechanism of equal strength, the more rules there are, the more likely it is that verification record fields will be missing or row alignment will be inconsistent, making it difficult to form a replayable traceability link.

[0098] The material list is output by reading the unique material record pointed to by the material process constraint field. The material list contains material name and material attribute fields. The correspondence between the material attribute fields and the measurable item names of the material attribute classes is written in the material list. The process preference in the design requirement record is read and combined with the process description field in the material process constraint field to output the process description. The process description references the material name and keeps the original content of the process description field unchanged. The material list and process description are written into the design requirement record and bound to the final solution.

[0099] The multi-view design drawings, zoning annotation drawings, vectorized outlines, critical dimension tables, bill of materials, and process specifications are compiled into a single output package and bound to the design requirements record as a plaintext output package. The plaintext output package is then encrypted to generate a ciphertext output package, which is written into the design requirements record as a delivery result. A one-time key is generated during encryption and is only retained in runtime memory during encryption and decryption verification; it is not written into the design requirements record in plaintext form. A decryption verification is performed on the ciphertext output package, verifying that the decrypted file list matches the plaintext output package file list and that the number of rows in the critical dimension table matches the verifiable data. If the number of rows in the rule table is consistent, and the decryption verification fails, the ciphertext output packet is marked as invalid and written to the access log. The ciphertext output packet is deleted, and the encryption and decryption verification is re-executed. After the verification passes, the plaintext output packet and temporary files are cleared, and only the ciphertext output packet is retained. At the same time, the one-time key and plaintext cache in the running memory are cleared, and the access log of the plaintext clearing is written. The access logs for the output packet generation, encryption, decryption verification, and plaintext clearing processes are recorded separately. The access logs include the design requirement record identifier, event type, occurrence time, output packet identifier, and access initiator information, and the access logs are written to the design requirement record.

[0100] In summary, this invention improves verification consistency and traceability by: collecting protective footwear design requirements, analyzing appearance intent and determining applicable standard clauses, organizing the applicable standard clauses into a rule table and establishing a scale conversion benchmark; generating footwear solutions based on parameterized constraint sets and appearance intent, performing semantic segmentation, key point localization, and edge extraction measurement and verifying the rules, and keeping the partition list, topological relationship fields, and material and process constraint fields synchronously and effectively.

[0101] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for generating protective footwear using AI based on physical process constraints and safety standard verification, characterized in that, include: Collect design requirements for protective footwear, analyze the appearance intent and determine the applicable standard clauses, compile the applicable standard clauses into a judgment rule table, and establish a scale conversion benchmark. Based on the scale conversion benchmark and the decisionable rule table, the topological relationship and material process constraints of the assembled protective components are combined, and the rules are bound according to the functional partition to form a set of parameterized constraints. The footwear design is generated based on the parametric constraint set and appearance intent. During the generation process, semantic segmentation, key point localization, and edge extraction measurement are performed and the rules are verified. When the verification result indicates that the footwear design does not conform to the decisionable rule table and parametric constraint set, local shape correction, local redrawing and generation, and discarding and regenerating are performed in sequence to obtain the final design. The final design is reviewed, and outputs multi-view design drawings, partition annotation drawings, vectorized outlines, key dimension tables, bill of materials and process specifications. The design is then encrypted and decrypted, and plaintext is cleared and access logs are removed.

2. The AI ​​generation method for protective footwear based on physical process constraints and safety standard verification as described in claim 1, characterized in that: The specific steps for collecting design requirements for protective footwear, analyzing the appearance intent, and determining applicable standard clauses are as follows: Protective footwear design requirements are collected through field forms, the fields are solidified into design requirement records, the reference images in the design requirement records are preprocessed to obtain standard views, the outer contour of the sole is obtained by edge extraction, and the toe is determined based on curvature characteristics; The text descriptions in the design requirements record are segmented and matched and merged according to the appearance intent dictionary. When the text descriptions do not cover the appearance intent, the main outline direction, texture density and dividing line distribution of the upper and sole are extracted. The expected protection items and usage scenarios in the design requirements record are mapped to applicable standard clauses according to the mapping table.

3. The AI ​​generation method for protective footwear based on physical process constraints and safety standard verification as described in claim 2, characterized in that: The specific steps for compiling applicable standard clauses into a decidable rule table and establishing a scale conversion benchmark are as follows: The applicable standard clauses are broken down into rows of determinate rule tables. Each row consists of the name of the measurand, the comparison relationship, the standard requirement value, and the area of ​​application. Only measurands that can be obtained by semantic segmentation, key point localization, and edge extraction are retained. The standard requirement values ​​are the length, area, proportion, and material file attribute values. The comparison relationships are not less than, not greater than, equal to, within an interval, and within a set. The cumulative distance along the outer contour between the foremost contour point of the shoe toe and the last contour point of the heel is used as the pixel quantity of the sole length, and the ratio of the measured last length to the pixel quantity of the sole length is used as the scale conversion benchmark.

4. The AI ​​generation method for protective footwear based on physical process constraints and safety standard verification as described in claim 1, characterized in that: The topological relationships and material and process constraints of the assembled protective components include: Read the effective areas row by row from the rule table and merge them into anti-impact area, anti-puncture area, electrical performance area and anti-slip area. Write the merged results back to the effective area field of the row as the functional partition name. Write the anti-impact area, anti-puncture area, electrical performance area and anti-slip area into the design requirement record as a partition list. Read the expected protection items in the design requirement record, form the anti-impact related protection component set and the anti-puncture related protection component set and write them back to the design requirement record. Assemble the topological relationship with the outer contour of the sole as the geometric reference and establish internal inclusion relationship, covering relationship and non-intersection relationship. Read the material preferences and process preferences from the design requirements record, determine the unique material record, write it into the design requirements record as a material and process constraint field, and associate the material attribute fields according to the functional partition.

5. The AI ​​generation method for protective footwear based on physical process constraints and safety standard verification as described in claim 4, characterized in that: The binding rules by functional partition form a set of parameterized constraints, including: Traverse each row of the rule table, bind the row to the functional partition in the design requirement record according to the area of ​​effect, establish the correspondence between the row rule and the topology relationship, solidify the measurement caliber description for the row rule and write it back to the row; The partition list, topology relationship, material and process constraint fields, and decision rule table are merged and stored into a parameterized constraint set, which is then bound to the design requirement record, and the coverage ratio is written into the parameterized constraint set.

6. The AI ​​generation method for protective footwear based on physical process constraints and safety standard verification as described in claim 1, characterized in that: The specific steps for generating a footwear design based on the set of parametric constraints and the appearance intent are as follows: The outer contour of the sole defines the outer boundary of the footwear design. The length direction is determined by the line connecting the toe end and the heel end. A vertical dividing line is drawn at the midpoint of the line to form the corresponding area of ​​the toe direction and the corresponding range of the puncture-proof area. The partition space is generated according to the partition list and the anti-impact area, electrical performance area, puncture-proof area and anti-slip area are arranged. The anti-impact related protective components are placed in the anti-impact area and the puncture-proof related protective components are placed in the puncture-proof area. The unique material record in the material process constraint field is associated with the footwear design according to the functional partition. Generate the line style, texture density, segmentation complexity, and surface texture of the footwear design without changing the outer contour and topological relationship of the sole.

7. The AI ​​generation method for protective footwear based on physical process constraints and safety standard verification as described in claim 6, characterized in that: The generation process involves semantic segmentation, key point localization, edge extraction measurement, and rule verification. The specific steps are as follows: Semantic segmentation is performed on the footwear solution, outputting a partition mask consistent with the partition list, and outputting masks for the sets of anti-impact and anti-puncture protective components. The partition occupancy boundary and the component set occupancy boundary are used as the initial mask boundary for fitting correction. Key point localization is performed on the footwear solution to obtain the endpoint positions of the length-class measured items and determine the two endpoints by projection along the length direction. Edge extraction is performed on the footwear solution to obtain the outer contour edge of the sole, the partition boundary edge, and the protective component set boundary edge. Actual measurements are formed according to the measurement caliber description of the decision rule table, and the coverage ratio is calculated. The system iterates through the rule table to verify the actual measurements according to the comparison relationship and generates a failure list. It also verifies the topological relationship and determines whether the rule rows related to the coverage ratio and coverage pass or fail. The footwear solution is considered to pass when the failure list is empty and all topological relationship verifications pass.

8. The AI ​​generation method for protective footwear based on physical process constraints and safety standard verification as described in claim 7, characterized in that: The process of sequentially performing local shape correction, local redrawing and generation, and discarding and regenerating yields the final solution. The specific steps are as follows: Local shape correction modifies the partition boundaries and protective component set boundaries within the effective area while maintaining the outer contour of the sole and satisfying the internal containment relationship. After local shape correction, the measurement and verification are repeated. If it fails, local redrawing is performed within the effective area. After local redrawing, the measurement and verification are repeated. If it still fails, the footwear design is discarded and the footwear design is regenerated until it passes. When all designs pass, the final design is determined and written into the design requirements record.

9. The AI ​​generation method for protective footwear based on physical process constraints and safety standard verification as described in claim 1, characterized in that: The process of reviewing the final design and outputting multi-view design drawings, zoning annotation diagrams, vectorized outlines, key dimension tables, material lists, and process specifications involves the following steps: If the list is empty, a consistency review is performed. After the consistency review is passed, the verification records are summarized in the order of the rows in the decisionable rule table. The multi-view design drawings, partition annotation drawings, vectorized outlines, key dimension tables, material lists and process specifications are output and written into the design requirement records.

10. The AI ​​generation method for protective footwear based on physical process constraints and safety standard verification as described in claim 9, characterized in that: The encryption / decryption, plaintext clearing, and access log recording removal include: The multi-view design drawings, partition annotation drawings, vectorized outlines, key dimension tables, bill of materials, and process specifications are compiled into an output package and bound to the design requirements record as a plaintext output package. The plaintext output package is encrypted to generate an ciphertext output package and written into the design requirements record. The ciphertext output package is decrypted and verified. After the decryption verification is successful, the plaintext output package and temporary files are cleared, leaving only the ciphertext output package. The access log of the output package generation, encryption, decryption verification, and plaintext clearing process is written into the design requirements record.

Images