Body shape analysis apparatus and body shape analysis method

The body shape analysis device and method improve shoe fit compatibility by classifying foot shapes into detailed clusters using three-dimensional models, addressing the limitations of conventional sizing systems and enhancing fit accuracy and inventory management.

JP2024158135APending Publication Date: 2024-11-08ASICS CORP
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
JP2023073079
Authority / Receiving Office
JP · JP
Patent Type
Applications
Current Assignee / Owner
Filing Date
2023-04-27
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

Existing shoe sizing systems often fail to account for variations in foot width and shape beyond foot length, leading to mismatched fits, especially in online sales and international sales, and are limited by conventional grading methods that prioritize foot length over other anatomical features.

Method used

A body shape analysis device and method that utilizes three-dimensional homologous models to analyze foot shapes, employing cluster analysis based on cosine similarity and principal component analysis to classify foot models into detailed clusters, allowing for customized shoe design and inventory management.

Benefits of technology

Enhances shoe fit compatibility by considering foot shape variations, reduces inventory costs, and improves sales compatibility across different regions by providing tailored shoe sizes and designs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure 2024158135000001_ABST
    Figure 2024158135000001_ABST
Patent Text Reader

Abstract

To provide a technique that improves the adaptability between feet and shoes.SOLUTION: In a shape analysis apparatus 50, a model storage part 22 stores a set of three-dimensional coordinates indicating anatomical features in a foot shape as a plurality of three-dimensional homologous models for a plurality of foot shapes. The plurality of the three-dimensional homologous models are divided into multiple clusters in the upper hierarchy by cluster analysis based on the cosine similarity of the position vectors from a predetermined reference point to each point of the three-dimensional homologous model, and then the lower hierarchy is further divided into clusters by cluster analysis based on principal component analysis with dimensionality reduced to the second main component or less, excluding the first main component in the case where principal component analysis is performed for each cluster in the upper hierarchy.SELECTED DRAWING: Figure 2
Need to check novelty before this filing date? Find Prior Art

Description

[Technical field]

[0001] The present invention relates to a body shape analysis device and a body shape analysis method, and more particularly to a technique for analyzing shoe suitability based on foot shape measurements. [Background technology]

[0002] It is preferable to offer a wide variety of shoe sizes to accommodate a variety of foot shapes and sizes. However, expanding the size range entails increased manufacturing costs and the risk of having a large inventory, so in reality, most shoe manufacturers offer sizes centered on those that are in high demand.

[0003] Since the units and size guidelines for shoe size indications vary depending on the country or region, shoes of the same size must be adjusted for each country / region when sold in different countries / regions. Therefore, it is difficult to strictly match the size indications of all countries / regions for shoes sold internationally, and it may be difficult to know which size fits a foot without actually trying them on. Shoe stores may have specialized personnel called shoe fitters who can select shoes that fit the customer's feet based on their extensive knowledge and experience, but shoe fitters cannot be relied on when selling online. In recent years, technology has been developed to select shoes that fit the foot by measuring the foot shape in three dimensions.

[0004] A technology is known that selects a type of shoe that fits a foot shape based on three-dimensional data obtained by measuring the foot shape three-dimensionally (see, for example, Patent Document 1). In Patent Document 1, two-dimensional shape feature data is generated separately for the toe, middle, and heel parts from the three-dimensional data of the foot shape. Similarly, feature data for each part is generated for the shoe last, and the collected feature data for a large number of shoe lasts is clustered, and a code is assigned to each cluster and stored. The relationship between the feature data of the foot shape and the feature data of the shoe last that fits it is learned using a neural network, and a rule is set for determining the feature data of the shoe last from the feature data of the foot shape. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2000-90272 A Summary of the Invention [Problem to be solved by the invention]

[0006] Generally, shoes on the market are often made separately for men and women, taking into account the differences in physique and design preferences between men and women. In Japan in particular, sizes are often developed separately for men and women according to the JIS standard, so shoes are rarely designed as unisex models. Except for sports shoes, where fit is particularly important, sizes are rarely developed according to foot width in addition to foot length.

[0007] Here, if shoes are selected according to foot length, mismatches may occur, such as the width being too narrow or too wide. In addition, shoes that fit may not be found because the fit is not achieved in areas other than foot length and width. However, since sizes are basically developed according to foot length, which is the factor that has the greatest impact on the fit of shoes, it is cumbersome to expand the selection of shoes to include shoes of different foot length sizes while placing importance on the fit of foot width and other areas, and it is currently difficult to try multiple sizes, especially in online sales.

[0008] On the other hand, it is particularly difficult for shoe manufacturers to offer sizes that meet demand when selling shoes internationally, as the demand for sizes and shapes varies depending on the average body type from country / region to country / region.

[0009] The present invention has been made in consideration of the above circumstances, and has an object to provide a technique for improving the fit between feet and shoes. [Means for solving the problem]

[0010] In order to solve the above problems, a body shape analysis device of one embodiment of the present invention includes a model memory unit that pre-stores a group of three-dimensional coordinates indicating anatomical features of at least a portion of the body as a plurality of three-dimensional homology models for a plurality of body shapes, a measurement data acquisition unit that acquires a group of three-dimensional coordinates indicating anatomical features corresponding to the three-dimensional homology models in the body shape of the subject as measurement data, and an evaluation determination unit that determines a three-dimensional homology model that is similar to the body shape of the subject from the plurality of three-dimensional homology models by evaluating the similarity between coordinates between the group of three-dimensional coordinates indicating the anatomical features in the measurement data and the group of three-dimensional coordinates indicating the anatomical features in the plurality of three-dimensional homology models.

[0011] Another aspect of the present invention is a body shape analysis device, comprising: a model storage unit that stores in advance a group of three-dimensional coordinates indicating anatomical features of a foot as a group of three-dimensional homologous models for a group of body shapes, and divides the plurality of three-dimensional homologous models into a plurality of clusters by cluster analysis based on the cosine similarity between position vectors from a predetermined reference point to each point of the three-dimensional homologous models, and an output unit that outputs a group of three-dimensional coordinates of the three-dimensional homologous models for each cluster.

[0012] Yet another aspect of the present invention is a body shape analysis method, which includes the steps of: reading out a group of three-dimensional coordinates indicating anatomical features of the foot from a plurality of three-dimensional homologous models for a plurality of body shapes; acquiring a group of three-dimensional coordinates indicating anatomical features corresponding to the three-dimensional homologous models in the body shape of a subject as measurement data; evaluating the similarity between the group of three-dimensional coordinates indicating the anatomical features in the measurement data and the group of three-dimensional coordinates indicating the anatomical features in the plurality of three-dimensional homologous models; and determining a three-dimensional homologous model similar to the body shape of the subject from the plurality of three-dimensional homologous models based on the evaluation of the similarity. Effect of the Invention

[0013] According to the present invention, a technique for improving the fit between feet and shoes can be provided. [Brief description of the drawings]

[0014] [Figure 1] FIG. 1 is a diagram showing a basic configuration of a shape analysis system. [Diagram 2] FIG. 2 is a functional block diagram showing each component of the shape analysis device. [Diagram 3] FIG. 1 is a diagram showing an example of a three-dimensional homology model. [Figure 4] This figure shows the relationship between the distribution of foot length and circumference by gender and the Japanese JIS standard. [Diagram 5] FIG. 1 is a diagram showing the difference between a general cluster analysis method and a new cluster analysis method. [Figure 6] FIG. 13 is a diagram illustrating a method of combining different cluster analyses in upper and lower hierarchical layers. [Figure 7] FIG. 1 is a diagram showing Euclidean distance and cosine similarity between points. [Figure 8] FIG. 13 is a diagram showing the relationship between corresponding points in two foot models. [Figure 9] FIG. 13 is a diagram showing a comparison between foot print models using Euclidean distance and cosine similarity. [Figure 10] FIG. 13 is a diagram showing the difference in classification results between cluster analysis based on cosine similarity or Euclidean distance. [Figure 11] FIG. 13 is a diagram showing classification results obtained by cluster analysis based on cosine similarity. [Figure 12] FIG. 13 shows regression lines of foot length and foot circumference for the top three clusters, and the results of principal component analysis of the top three clusters. [Figure 13] FIG. 13 is a diagram showing the results of principal component analysis in clusters. [Figure 14] FIG. 1 is a diagram showing a basic configuration of a shape analysis system. [Figure 15] FIG. 2 is a functional block diagram showing each component of the shape analysis device. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0015] Hereinafter, the present invention will be described based on preferred embodiments with reference to the drawings. In the embodiments and modifications, the same or equivalent components are given the same reference numerals, and redundant descriptions will be omitted as appropriate.

[0016] (First embodiment) In this embodiment, the three-dimensional shapes of the feet of many subjects are measured and stored as foot shape data, and then classified into multiple clusters using a cluster analysis method described later. Furthermore, shoe lasts and insoles are designed based on the classified multiple clusters, providing shoe product shape variations based on standards different from conventional standards, and providing shoes that fit the feet of many people. Furthermore, demand forecasts are made for each shape variation according to the distribution of foot shapes.

[0017] 1 shows the basic configuration of a shape analysis system 100. The shape analysis system 100 includes a three-dimensional foot shape measuring device 18 and a shape analysis device 50. The foot shape of the subject 10 is scanned by the three-dimensional foot shape measuring device 18. Foot shape data may be acquired, for example, in the shoe design stage, when a large number of subjects 10 are used as the subject 10 in order to acquire foot shape data of various sizes or shapes regardless of gender or age, or when a person considering purchasing shoes at a shoe store is used as the subject 10.

[0018] The three-dimensional foot shape measuring device 18 obtains three-dimensional data on the foot shape of the subject 10 by laser measurement. The "foot shape" here is a three-dimensional model that virtually reproduces the three-dimensional shape of the foot of the subject 10. The measurement value as the three-dimensional measurement result of the foot shape scanned by the three-dimensional foot shape measuring device 18 is transmitted from the three-dimensional foot shape measuring device 18 to the shape analyzing device 50. The shape analyzing device 50 may be realized as a server that is connected to multiple three-dimensional foot shape measuring devices 18 via a network and collects and analyzes measurement data from each of them. Note that the "body shape analyzing device" in the claims may refer to the entire shape analysis system 100 or the shape analyzing device 50. In addition, in this embodiment, since the shape analyzing device 50 is realized in a form that includes many of the characteristic functions included in the "body shape analyzing device" in the claims, the shape analyzing device 50 essentially corresponds to the "body shape analyzing device".

[0019] The shape analysis device 50 is a server computer that is connected to multiple 3D foot shape measuring devices 18 via a network line such as the Internet or a LAN (Local Area Network) and communication means such as wireless communication. The shape analysis device 50 may be configured as a single server computer or a combination of multiple server computers.

[0020] FIG. 2 is a functional block diagram showing each component of the shape analysis device 50. The shape analysis device 50 can be realized by various hardware and software configurations. FIG. 2 shows functional blocks of the shape analysis device 50 that are realized by cooperation of various hardware and software configurations. Therefore, it is understood by those skilled in the art that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof. The shape analysis device 50 is composed of a combination of hardware such as a microprocessor, a memory, a display, and a communication module. The shape analysis device 50 may be a server computer built and managed by a shoe manufacturer. However, the shape analysis device 50 runs a program having the following functions. The shape analysis device 50 includes, as its functions, a communication unit 19, a measurement data acquisition unit 20, a model generation unit 21, a model storage unit 22, an output unit 26, an analysis processing unit 30, and a statistical processing unit 35. The communication unit 19 of the shape analysis device 50 is connected to the three-dimensional foot shape measuring device 18 and the external terminal 17 via a network.

[0021] The model storage unit 22 stores in advance a group of three-dimensional coordinates indicating at least a part of the anatomical features of the body as a plurality of three-dimensional models for a plurality of body shapes. The body shape here mainly refers to the foot shape, but may also refer to a body shape other than the foot shape, for example, body type data indicating the shape of the entire human body, or may refer to a partial body shape other than the foot shape. The three-dimensional model of the foot shape stored in the model storage unit 22 is, for example, a three-dimensional homologous model in which a group of coordinates of points indicating the anatomical features of an average foot in a three-dimensional space is defined. The three-dimensional homologous model of the foot shape is a model in which a plurality of foot shapes having various sizes and shapes are expressed as a polyhedron in which they are anatomically associated by the same topological geometric structure with the same number of points. The points indicating the anatomical features of the average foot are obtained in advance based on foot samples obtained from a large number of subjects. The three-dimensional model in the modified example may be, for example, a three-dimensional shape model in a spherical coordinate system, or various other models created in accordance with the same anatomical standard, such as a model in which a template foot shape generated by a three-dimensional CAD is deformed based on anatomical landmarks.

[0022] FIG. 3 shows an example of a three-dimensional homologous model. The three-dimensional homologous model 104 of the foot shape is composed of a group of points indicating the anatomical features of the foot. The three-dimensional homologous model 104 does not necessarily faithfully reproduce the foot shape of the subject 10, especially the shape around the toes, but is constructed in an intentionally deformed shape so as to be easily compared with other foot shapes or three-dimensional homologous models. In other words, the three-dimensional homologous model 104 does not have a perfect similarity to the actual foot shape or foot contour of the subject 10, and has a shape in which convex hull processing has been applied to the contour of the toes in particular. In addition, by detecting a plurality of characteristic points included in the foot contour of the three-dimensional homologous model 104 and using each point as a reference, the length of each part, such as the foot length, foot circumference, arch circumference, heel circumference, and heel width, can be measured. The method of measuring the length of each part is known, so the description will be omitted.

[0023] Returning to FIG. 2, the measurement data acquisition unit 20 acquires, as measurement data, a group of three-dimensional coordinates indicating anatomical features corresponding to a three-dimensional homologous model of the foot shape of the measurement subject 10 from the three-dimensional foot shape measuring device 18 via the communication unit 19. The measurement data acquisition unit 20 may acquire measurement data from an external terminal 17 other than the three-dimensional foot shape measuring device 18 via the communication unit 19. The model generation unit 21 generates a three-dimensional homologous model of the foot shape based on the measurement data acquired by the measurement data acquisition unit 20, and stores it in the model storage unit 22 as a new three-dimensional homologous model. In this case, even if measurement data is acquired using various types of foot shape measuring devices as the three-dimensional foot shape measuring device 18 or the external terminal 17, the model generation unit 21 converts it into a three-dimensional homologous model. This allows the measurement data to be generalized into a data format that is easy to use. On the other hand, in a modified example, the measurement data acquired by the measurement data acquisition unit 20 may be converted into a three-dimensional homologous model of the foot shape in advance by the three-dimensional foot shape measuring device 18 or the external terminal 17. In this case, the model generating unit 21 does not need to generate a three-dimensional homologous model of the foot shape based on the measurement data, and the measurement data may be directly stored in the model storage unit 22 as a new three-dimensional homologous model.

[0024] The analysis processing unit 30 has a cluster analysis processing unit 31 and a principal component analysis processing unit 32. The cluster analysis processing unit 31 executes cluster analysis of upper and lower hierarchical levels, which will be described later, on the multiple three-dimensional homologous models stored in the model storage unit 22. The cluster analysis processing unit 31 implements a method that combines different cluster analyses in the upper and lower hierarchical levels. The principal component analysis processing unit 32 executes principal component analysis, which will be described later, on the multiple three-dimensional homologous models of foot shapes stored in the model storage unit 22. After executing the cluster analysis of the upper hierarchical level, the cluster analysis processing unit 31 executes cluster analysis of the lower hierarchical level based on the results of the principal component analysis executed by the principal component analysis processing unit 32.

[0025] Information on each cluster as a classification result obtained by executing the cluster analysis by the cluster analysis processing unit 31 and the principal component analysis by the principal component analysis processing unit 32 is output to the outside by the output unit 26. Based on the classification result by the cluster analysis, the analysis processing unit 30 divides the three-dimensional homology model stored in the model storage unit 22 into a plurality of clusters and stores them.

[0026] The output unit 26 may display the classification results on a display screen of the shape analysis device 50 or the external terminal 17, or may transmit the classification results to the external terminal 17 via the communication unit 19. The output unit 26 outputs a group of three-dimensional coordinates of the three-dimensional homologous model for each cluster. Information on the classification results output by the output unit 26 may be used for designing a shoe last or an insole.

[0027] The statistical processing unit 35 executes demand prediction of shoe lasts corresponding to each foot shape model based on the distribution of foot shape models for each cluster. The result of the demand prediction is output to the external terminal 17 by the output unit 26. The statistical processing unit 35 manages a database of shoe lasts, footwear or insoles corresponding to the foot shape models for each cluster stored in the model storage unit 22 according to the result of the demand prediction.

[0028] Figure 4 shows the distribution of foot length and circumference by gender and its relationship to the Japanese JIS standard. Shoe sizes in Japan are classified and standardized by the Japanese Industrial Standards (JIS), which is a standard, according to foot length and circumference / width. This standard takes into account differences in physique and specifies different sizes for men, women, and children.

[0029] Figure 4(a) shows the JIS standard for women and the distribution of a sample of many adult women. Figure 4(b) shows the JIS standard for men and the distribution of a sample of many adult men. The JIS standard for women specifies a foot length range of 19.5 cm to 27.0 cm, while the JIS standard for men specifies a foot length range of 20.0 cm to 30.0 cm. Foot circumference / foot width is divided into shoe widths A to F for women and shoe widths A to G for men according to foot length. However, different ranges are specified, such as shoe width A for women being 183 mm for foot circumference and 76 mm for foot width, and shoe width A for men with a foot length of 20.0 cm being 189 mm for foot circumference and 79 mm for foot width.

[0030] The multiple diagonal lines in Figure 4(a) and Figure 4(b) indicate the relationship between foot length [mm] and foot circumference [mm] for each foot length size according to the JIS standard. The multiple diagonal lines indicate women's shoe widths A to F from the bottom in Figure 4(a), and men's shoe widths A to G from the bottom in Figure 4(b). The difference in foot circumference between shoe widths is approximately 6 mm.

[0031] The distribution of female samples shown in Figure 4(a) is mostly within the range of foot length and foot circumference targeted by the JIS standard, but there are also samples that fall outside the range. The distribution of male samples shown in Figure 4(b) is mostly within the range of foot length and foot circumference targeted by the JIS standard, but like women, there are also samples that fall outside the range.

[0032] The bold diagonal lines in Figure 4(a) and Figure 4(b) indicate the sample averages for men and women. The foot circumference varies by about ±3 to 4 widths above and below the average line for each foot length size. In other words, even if the relationship between foot length and foot circumference as shown in the figure is determined and provided as a guideline, the actual variation or scatter of foot circumference is large. Moreover, while conventional shoe standards only determine the relationship between foot length and foot circumference, the size and shape of the foot are also determined by other parameters, so even if only the foot length and foot circumference are matched, the target shoe does not necessarily fit the target's foot. In order to solve this situation, we decided to classify the three-dimensionally measured foot shape models by cluster analysis using the method described below, and design size variations based on the classification.

[0033] FIG. 5 shows the difference between the general cluster analysis method and the new cluster analysis method. When using the general cluster analysis method to cluster the foot shape data group X, it is possible to perform cluster analysis based on Euclidean distance. However, in this method, the clusters are classified in a manner that is significantly influenced by foot length size. In addition, if it is desired to classify not only foot length size but also differences in foot shape, it is possible to divide by foot length and perform cluster analysis for each foot length size. However, even in the case of this method, it is significantly influenced by foot length size, and it is expected that it will be difficult to distinguish from the conventional grading method that classifies only by multiple foot circumferences for each foot length size. In addition, there is a risk that the number of clusters classified will differ for each size.

[0034] Therefore, the cluster analysis processing unit 31 in this embodiment performs cluster analysis across all foot length sizes as shown in the figure. The foot shape data group X is configured in such a way that a large cluster W separated by similar shapes (i.e., similar shapes) across foot length sizes is multiplied by a small cluster H separated by detailed features. The following figure illustrates the hierarchical state of such clusters.

[0035] Fig. 6 shows a method of combining different cluster analyses in upper and lower hierarchical layers. First, clusters 1 to n obtained by dividing a footprint data group X by cluster analysis based on cosine similarity constitute the first hierarchical layer 110, which is the upper hierarchical layer. Clusters (1-1, 1-2, . . . 1-m, 2-1, 2-2, . . . 2-m, . . . , n-1, n-2, . . . nm) divided finely by cluster analysis based on principal component analysis in which the dimensions of clusters 1 to n are reduced to the second principal component or less excluding the first principal component when principal component analysis is performed constitute the second hierarchical layer 112, which is the lower hierarchical layer. The cluster analysis methods for the upper and lower hierarchical layers will be described later.

[0036] FIG. 7 shows the Euclidean distance and cosine similarity between points. An example will be described in which four points, a first point 120, a second point 121, a third point 122, and a fourth point 123, exist in a vector space. A common method for measuring the closeness between two points is a method of judging the distance by looking at a distance called Euclidean distance. The Euclidean distance is the linear distance between two points, and the smaller the value, the more similar the two points are. On the other hand, in this embodiment, a method of judging the closeness by looking at a similarity called cosine similarity is mainly adopted. The cosine similarity is the cosine value (cosθ) of the angle θ formed by the position vectors from the origin of the two points, and the smaller the angle θ, the closer the cosine value is to 1, and the closer the cosine value is to 1, the more similar the two points are. In general, the cosine similarity is used in situations where the similarity between sentences is measured, and is rarely used in situations where the similarity between shapes is measured.

[0037] Since the Euclidean distance (distance 124) from the first point 120 to the second point 121 is shorter than the Euclidean distance (distance 125) from the first point 120 to the third point 122, when the Euclidean distance is used as the standard, the second point 121 is closer to the first point 120 than the third point 122.

[0038] On the other hand, the angle θ between the position vectors of the first point 120 and the second point 121 from the origin 1 The angle θ between the position vectors from the origin of the first point 120 and the third point 122 is 2is smaller and has a cosine value closer to 1, so that when the cosine similarity is used as a standard, the third point 122 is closer to the first point 120 than the second point 121.

[0039] Furthermore, the first point 120 and the fourth point 123 have almost the same position vector from the origin, the angle θ is almost zero, and the cosine similarity is almost 1. Therefore, the fourth point 123 is farther away than the second point 121 when the Euclidean distance is used as the standard, but is extremely close to the first point 120 when the cosine similarity is used as the standard.

[0040] FIG. 8 shows the relationship between corresponding points in two foot models. The foot model is composed of a cloud of points, and each point indicates an anatomical feature in the foot. Since each foot model includes points that indicate the same anatomical feature, the similarity between the entire foot models is determined based on the similarity of the positional relationship between the points. In comparison between a relatively small foot model 130 and a relatively large foot model 131, the cosine similarity is calculated based on the angle formed by the position vector from the origin between corresponding points among the many point clouds included in each foot model. In the side view of FIG. 8(a), points 132 and 133, points 134 and 135, and points 136 and 137 correspond to each other, and in the top view of FIG. 8(b), points 138 and 139, points 140 and 141, and points 142 and 143 correspond to each other, respectively. In Fig. 8(a), the angle between the position vectors of points 132 and 133, the angle between the position vectors of points 134 and 135, and the angle between the position vectors of points 136 and 137 are all almost zero. In Fig. 8(b), the angle between the position vectors of points 138 and 139, the angle between the position vectors of points 140 and 141, and the angle between the position vectors of points 142 and 143 are all almost zero. Therefore, it can be said that foot print model 130 and foot print model 131 are almost similar in shape although their sizes are different.

[0041] Figure 9 shows a comparison of Euclidean distance and cosine similarity when comparing foot models. The closer the foot length between the two foot models, the smaller the Euclidean distance value, and the more likely they are to be judged as "similar." Conversely, the larger the Euclidean distance between the two foot models, the more likely they are to be judged as "dissimilar." The closer the foot models are to similar shapes, the closer the cosine similarity is to 1, and the more likely they are to be judged as "similar." Conversely, the closer the cosine similarity is to 0, the more likely they are to be judged as "dissimilar."

[0042] In the example of Figure 9(a), the two footprint models almost overlap. The Euclidean distance is relatively small, and the cosine similarity is almost 1. Therefore, it is easy to determine that they are "similar" using either the Euclidean distance or the cosine similarity as the criterion.

[0043] On the other hand, in the example of Figure 9(b), the two footprint models are clearly different in size. The Euclidean distance is relatively large, and it is easy to judge that they are "not similar" when based on Euclidean distance. On the other hand, the cosine similarity is almost 1, which is almost the same as the example of Figure 9(a), and it is easy to judge that they are "similar" when based on cosine similarity. In other words, when comparing two footprint models, even if there is a large difference when based on Euclidean distance, there may be almost no difference when based on cosine similarity, which shows that cosine similarity is not affected by size. In the example of Figure 9(b), the two footprint models are different in size, but are similar enough to be said to be almost similar in shape.

[0044] Figure 10 shows the difference in classification results between cluster analysis based on cosine similarity and Euclidean distance. As classification results of actual cluster analysis of foot shape data, the right shows a hierarchical cluster based on Euclidean distance, and the left shows a hierarchical cluster based on cosine similarity.

[0045] In the cluster analysis based on Euclidean distance, in the subdivided clusters at the bottom level, the foot length size gradually increases from left to right, such as decreasing toward the left and increasing toward the right. This also shows that the influence of foot length size is dominant in the cluster analysis based on Euclidean distance. In addition, regarding gender, there is a gradation in which the number of males gradually increases from left to right, such as more females toward the left and more males toward the right. This is because women tend to have smaller foot lengths than men, and it can be seen that the cluster analysis based on Euclidean distance is also influenced by gender.

[0046] On the other hand, the cluster analysis based on cosine similarity shows that the foot lengths are randomly mixed from left to right in the lowest cluster, and foot length size is hardly affected.Similarly, gender is randomly mixed and gender is hardly affected.

[0047] Figure 11 shows the classification results of cluster analysis based on cosine similarity. The footprint data samples used for the cluster analysis were from approximately 5,000 adult men and women. A hierarchical clustering method using Ward's method was used, based on the cosine similarity between each point in each footprint data.

[0048] The top three clusters (first cluster 150, second cluster 151, third cluster 152) are considered to be the upper hierarchy, and in each of first cluster 150, second cluster 151, and third cluster 152, the foot shape based on the coordinates of each point obtained by normalizing the foot length size is considered to be the average foot.

[0049] By superimposing the foot shape models of the average feet of each cluster, the characteristics of each cluster and the parts where such characteristics are likely to appear can be found from the variation in partial shapes. For example, in the height direction of the instep, the instep of the first cluster 150 is the lowest, the instep of the second cluster 151 is intermediate, and the instep of the third cluster 152 is the highest. For example, in the angle of the first toe, the angle of the first cluster 150 is the most protruding toward the inner foot, the angle of the second cluster 151 is intermediate, and the angle of the third cluster 152 is slightly inclined toward the second toe. For example, in the inclination around the ankle, the inclination of the first cluster 150 is the largest toward the inner foot, the inclination of the second cluster 151 is intermediate, and the inclination of the third cluster 152 is the smallest. For example, in the inclination around the ankle, the inclination of the first cluster 150 is the largest toward the inner foot, the inclination of the second cluster 151 is intermediate, and the inclination of the third cluster 152 is the smallest.

[0050] FIG. 12 shows the regression lines of foot length and foot circumference for the top three clusters, and the results of the principal component analysis for the top three clusters. First, FIG. 12 shows the regression lines showing the relationship between foot length and foot circumference for each of the first cluster 150, the second cluster 151, and the third cluster 152, superimposed on a graph of the JIS standard similar to FIG. 4. According to this, it can be seen that the regression lines are roughly parallel to each other, but their slopes are significantly different from the slope of the regression line showing the relationship between foot length and foot circumference in the JIS standard. Therefore, by designing shoe last variations according to the regression lines of each cluster, it becomes possible to design variations that are different from the conventional grading for men and women according to the JIS standard, that is, variations that do not distinguish between men and women and are divided by the shape characteristics of the foot shape other than the foot length size.

[0051] The principal component analysis processing unit 32 performs principal component analysis on the foot shape data for each cluster, calculates data by adding the first principal component PC1, which is the variance, to the average foot, which is the center of gravity for each cluster, and plots the relationship between foot length and foot circumference in the data on the graph in FIG. 12. Although each cluster is primarily clustered based on cosine similarity and classified in a manner that allows the influence of foot length size to be ignored, the data itself contains information on foot length. Therefore, when the principal component analysis processing unit 32 performs principal component analysis, the first principal component PC1 extracts a principal component that is largely influenced by foot length. However, since the average foot shape is divided into each cluster with different shapes based on cosine similarity as a large classification in the upper hierarchy, it is easy to classify the lower hierarchy by cluster based on principal components. In this regard, as plotted by the open circles in the graph, the plot showing the relationship between foot length and foot circumference in the average foot + first principal component PC1 for each cluster roughly follows the regression line for each cluster, so the results of the cluster analysis and the results of the principal component analysis are not contradictory.

[0052] FIG. 13 shows the results of principal component analysis on clusters. When the principal component analysis processing unit 32 executes principal component analysis on a cluster, it becomes clear which parameters have a large influence on the variance of the data held by that cluster. In FIG. 13, the first principal component PC1 is set on the vertical axis in the upper part, the second principal component PC2 is set on the vertical axis in the middle part, and the third principal component PC3 is set on the vertical axis in the lower part. On the left side, foot length is set on the horizontal axis, and on the right side, foot circumference is set on the horizontal axis. The group of plots surrounded on the left side of each graph represents females, and the group of plots surrounded on the right side represents males.

[0053] As shown in the two graphs on the left and right in the upper row, the first principal component PC1 is not only strongly correlated with foot length, but also strongly correlated with foot circumference. On the other hand, as shown in the graphs in the middle and lower rows, the regression line of the second principal component PC2 and below is almost horizontal, and it can be seen that they are almost not affected by foot length or foot circumference. In this embodiment, the characteristics of the results of such principal component analysis are utilized to exclude the first principal component PC1 that is affected by foot length and foot circumference, and only the second principal component PC2 and below that are not affected by foot length and foot circumference are used for sub-clustering. The utilization of such characteristics is considered to be advantageous in realizing the construction of clustered variations focusing on differences in detailed shapes, which is different from the conventional grading by size.

[0054] As shown in Fig. 6, in the first hierarchical layer 110, which is a higher hierarchical layer, the cluster analysis processing unit 31 performs cluster analysis based on cosine similarity to separate the clusters. In the second hierarchical layer 112, which is a lower hierarchical layer, the cluster analysis processing unit 31 performs cluster analysis based on principal component analysis in which the dimensions of each of the clusters separated in the first hierarchical layer 110 are reduced to the second principal component or less, so that the clusters are separated into many smaller clusters. The principal component analysis in this case is a principal component analysis in which the dimensions are reduced to the second principal component or less, excluding the first principal component, in the principal component analysis result by the principal component analysis processing unit 32. In the cluster analysis in the second hierarchical layer 112, hierarchical clustering based on the Ward method using Euclidean distance as a criterion is used, based on the second principal component PC2 and the principal components with a cumulative contribution rate of up to 99%.

[0055] The group of foot models divided into many clusters according to the detailed features can be used not only for designing variations of shoe lasts and footwear, but also for designing insoles. For example, the number of variations to be absorbed in the design of shoe lasts and footwear may be reduced by absorbing variations in the design of insoles regarding the detailed shapes that are not dependent on the foot length or foot circumference. In other words, the variations of shoe lasts and footwear may be combined with the variations of insoles to design the overall variations and cover the entire group of foot models.

[0056] In addition, each foot model may be assigned to a size range according to the JIS standard based on which of the shoe lasts in the size range designed according to the JIS standard it fits among the numerous foot models included in the foot model group.

[0057] Since the number of people of the footprint data included in each cluster is different, the ratio of the number of people of each cluster to the entire footprint data group, which is the population, is different. On the other hand, when a database holds a footprint data group that has been collected and accumulated for various purposes other than the footprint data group used in the cluster analysis in the first embodiment, the distribution of each foot length size in the population can be calculated based on the data. Therefore, the statistical processing unit 35 can quantify the ratio of each combination of cluster and foot length by multiplying the ratio of the number of people of each cluster by the distribution of each foot length size. The ratio of each combination of cluster and foot length when the foot length size is divided into a range of 22 cm to 31 cm in 1 cm increments is quantified.

[0058] When the statistical processing unit 35 designs variations of combination patterns of sizes and shapes of footwear products and insole products based on the ratio of each combination of clusters and foot lengths, it can predict demand as initial stock according to the ratio of each combination of clusters and foot lengths. In this case, it is possible to exclude combination patterns of clusters and foot lengths for which demand is predicted to be extremely low from the variations of footwear products from the beginning. The statistical processing unit 35 can reflect the demand prediction not only in the demand prediction as initial stock, but also in the production plan when the stock is reduced due to actual sales. The statistical processing unit 35 may output information regarding the ratio of each combination of clusters and foot lengths to the outside via the output unit 26.

[0059] Second embodiment This embodiment differs from the first embodiment in that the three-dimensional shape of the subject's foot is measured to obtain foot shape data, and similar foot shapes are extracted from three-dimensional homologous models of foot shapes stored in the shape analysis device 50, and can be used as reference information for purchasing shoes. The following description will focus on the differences from the first embodiment, and a description of the commonalities will be omitted.

[0060] FIG. 14 shows a basic configuration of a shape analysis system 100. The shape analysis system 100 mainly includes a shape data acquisition device and a shape analysis device 50. The shape data acquisition device is, for example, a three-dimensional foot shape measuring device 18 or an information terminal 16 such as a mobile phone equipped with a camera function. The subject 10 scans the foot shape of the subject 10 using the three-dimensional foot shape measuring device 18. Alternatively, the subject 10 places his / her foot on a dedicated measurement mat 12 and scans his / her own foot shape by photographing the foot using the camera function of the information terminal 16 and measuring it using a foot shape measurement app. An opportunity for acquiring foot shape data is, for example, when the subject 10 is a person considering purchasing shoes at a shoe store or at home.

[0061] The information terminal 16 may generate the foot shape of the subject 10 by a three-dimensional scanner function using technology such as LiDAR (Light Detection And Ranging) or image synthesis processing such as photogrammetry. The shape analysis system 100 may further include at least one of the measurement mat 12, the three-dimensional foot shape measuring device 18, and the information terminal 16. The subject 10 may operate the information terminal 16 as a user, or a person other than the subject 10, such as a clerk at a shoe store, may scan the foot shape using the three-dimensional foot shape measuring device 18 or the information terminal 16. Therefore, the information terminal 16 may be a terminal owned by a shoe store where the three-dimensional foot shape measuring device 18 is installed, or may be a terminal owned by the subject 10 himself.

[0062] The measurement values ​​as the three-dimensional measurement results of the foot shape scanned by the three-dimensional foot shape measuring device 18 or the information terminal 16 are transmitted from the three-dimensional foot shape measuring device 18 or the information terminal 16 to the shape analysis device 50. The shape analysis device 50 may be realized as a server connected to a plurality of three-dimensional foot shape measuring devices 18 or information terminals 16 via a network, and collects and analyzes measurement data from each of them. The "body shape analysis device" in the claims may refer to the entire shape analysis system 100 or the shape analysis device 50. In this embodiment, the shape analysis device 50 is realized by having many of the characteristic functions included in the "body shape analysis device" in the claims, so that the shape analysis device 50 essentially corresponds to the "body shape analysis device". However, the characteristic functions of the "body shape analysis device" may be distributed between the information terminal 16 and the shape analysis device 50, or may be realized by having most of them in the information terminal 16.

[0063] The shape analysis device 50 is a server computer connected to multiple 3D foot shape measuring devices 18 and multiple information terminals 16 via a network line such as the Internet or a LAN, or a communication means such as wireless communication. The information terminals 16 may be portable information terminals such as smartphones or tablet terminals, or may be personal computers.

[0064] FIG. 15 is a functional block diagram showing each configuration of the shape analysis device 50. The shape analysis device 50 can be realized by various hardware configurations and software configurations. FIG. 15 illustrates functional blocks of the shape analysis device 50 and the information terminal 16 that are realized by cooperation of various hardware configurations and software configurations. Therefore, it is understood by those skilled in the art that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof. The information terminal 16 is composed of a combination of hardware such as a camera module, a distance measurement sensor, a microprocessor, a touch panel, a memory, a communication module, etc. The information terminal 16 may be a general-purpose information terminal such as a smartphone or tablet terminal prepared by the measurement subject 10 or a shoe store. The information terminal 16 accesses a website provided by the shape analysis device 50 via a web browser, or displays information provided by the shape analysis device 50 by application software running on the information terminal 16. The shape analysis device 50 in this embodiment further includes an evaluation determination unit 24 as a function in addition to the shape analysis device 50 in the first embodiment. The communication unit 19 of the shape analyzing device 50, the three-dimensional foot shape measuring device 18 and the information terminal 16 are connected via a network.

[0065] The three-dimensional homologous models of a plurality of foot shapes stored in the model storage unit 22 may be classified into a plurality of clusters by the cluster analysis of the first embodiment. Alternatively, they may be three-dimensional homologous models corresponding to each of the variations of shoe lasts, footwear, or insoles designed based on the classification results of the cluster analysis of the first embodiment. Alternatively, they may be three-dimensional homologous models corresponding to combinations of foot length and foot circumference according to the conventional JIS standard.

[0066] The measurement data acquisition unit 20 acquires, as measurement data, a group of three-dimensional coordinates indicating anatomical features corresponding to a three-dimensional homologous model of the foot shape of the subject 10 from the three-dimensional foot shape measuring device 18 or the information terminal 16 via the communication unit 19. The model generation unit 21 generates a three-dimensional homologous model of the foot shape based on the measurement data acquired by the measurement data acquisition unit 20, and outputs it to the evaluation determination unit 24 as the three-dimensional homologous model of the foot shape of the subject 10.

[0067] The evaluation determination unit 24 evaluates the similarity between coordinates between a group of three-dimensional coordinates indicating anatomical features in the measurement data of the subject 10 and a group of three-dimensional coordinates indicating anatomical features in the multiple three-dimensional homologous models stored in the model storage unit 22. The evaluation determination unit 24 determines a three-dimensional homologous model that is similar to the foot shape of the subject 10 from the multiple three-dimensional homologous models based on the similarity between the coordinates between the corresponding three-dimensional homologous models.

[0068] More specifically, the evaluation determination unit 24 obtains position vectors from a predetermined reference point to each point in the multiple three-dimensional homology models, and obtains position vectors from a predetermined reference point to each point of the three-dimensional homology models in the measurement data. The evaluation determination unit 24 selects a three-dimensional homology model similar to the footprint of the measurement subject 10 from the multiple three-dimensional homology models based on the distance between the points of the corresponding three-dimensional homology models or the similarity of the position vectors. The distance between the points is, for example, Euclidean distance, and the similarity of the position vectors is, for example, cosine similarity. The evaluation determination unit 24 adds up the selection result of the three-dimensional homology model similar based on the Euclidean distance and the selection result of the three-dimensional homology model similar based on the cosine similarity, and determines the three-dimensional homology model most similar to the footprint of the measurement subject 10 from among them. It is assumed that about 6,000 three-dimensional footprint homology models are stored in the model storage unit 22. In this case, even if the evaluation determination unit 24 calculates the similarity between all three-dimensional homologous models and determines the most similar three-dimensional homologous model, the calculation can be completed in about 10 seconds, making it possible to extract the foot print in a very short time.

[0069] The evaluation determination unit 24 scales the determined three-dimensional homologous model to various sizes according to the foot length of the subject 10, and calculates the similarity between each of the three-dimensional homologous models and the three-dimensional homologous model of the foot shape of the subject 10. Based on each similarity, the evaluation determination unit 24 determines the scale of the three-dimensional homologous model with the highest similarity. The evaluation determination unit 24 extracts the determined three-dimensional homologous model of the foot shape and the type of shoe or insole corresponding to that scale from the database of the model storage unit 22, and generates information to be presented to the subject 10.

[0070] The output unit 26 transmits information about the foot shape model similar to that of the subject 10, as determined by the evaluation determination unit 24, and recommendation information indicating the type of shoes or insoles corresponding to that foot shape, to the information terminal 16 via the communication unit 19.

[0071] In a modified example, the evaluation determination unit 24 may evaluate which of the three-dimensional homologous foot print models of multiple celebrities stored in the model storage unit 22 has the highest similarity, and output the evaluation result to the information terminal 16.

[0072] The present invention has been described above based on the embodiment. This embodiment is merely an example, and it will be understood by those skilled in the art that various modifications are possible in the combination of each component and each processing step, and that such modifications are also within the scope of the present invention. Furthermore, the above-described embodiment can be generalized to obtain the following aspects.

[0073] [Aspect 1] a model storage unit that stores in advance a group of three-dimensional coordinates indicating anatomical features of at least a part of a body as a plurality of three-dimensional homologous models for a plurality of body shapes; a measurement data acquisition unit that acquires, as measurement data, a group of three-dimensional coordinates that indicate anatomical features corresponding to the three-dimensional homologous model in a body shape of a subject; an evaluation and determination unit that evaluates a similarity between a group of three-dimensional coordinates indicating anatomical features in the measurement data and a group of three-dimensional coordinates indicating anatomical features in the plurality of three-dimensional homologous models to determine a three-dimensional homologous model that is similar to a body shape of the measurement subject from the plurality of three-dimensional homologous models; A body shape analysis device comprising: [Aspect 2] The body shape analysis device of aspect 1, characterized in that the evaluation determination unit acquires position vectors from a predetermined reference point to each point in the multiple three-dimensional homology models, acquires position vectors from the predetermined reference point to each point of the three-dimensional homology models in the measurement data, and determines a three-dimensional homology model that is similar to the body shape of the subject from the multiple three-dimensional homology models based on the distance between points of corresponding three-dimensional homology models or the similarity of the position vectors. [Aspect 3] The body shape analysis device of aspect 1 or 2, characterized in that the multiple three-dimensional homology models are divided into multiple clusters by cluster analysis based on the cosine similarity between position vectors from a predetermined reference point to each point of the three-dimensional homology models and stored in the model memory unit. Aspect 4 a model storage unit that stores in advance a group of three-dimensional coordinates indicating anatomical features of the foot as a plurality of three-dimensional homologous models corresponding to a plurality of body shapes, and the plurality of three-dimensional homologous models are divided into a plurality of clusters by cluster analysis based on the cosine similarity between position vectors from a predetermined reference point to each point of the three-dimensional homologous models; an output unit that outputs a group of three-dimensional coordinates of the three-dimensional homology model for each cluster; A body shape analysis device comprising: Aspect 5 The body shape analysis device of any one of aspects 1 to 4, wherein the multiple three-dimensional homology models are divided into multiple clusters at an upper level by cluster analysis based on cosine similarity, and then divided into multiple clusters at a lower level by cluster analysis based on principal component analysis that reduces the dimensionality to the second principal component or lower, excluding the first principal component, when principal component analysis is performed on each cluster at the upper level. Aspect 6 A body shape analysis device described in any one of aspects 1 to 5, characterized in that the model memory unit further stores a value indicating each demand for multiple three-dimensional homologous models classified into multiple clusters based on the number of the three-dimensional homologous models for each cluster corresponding to the multiple clusters classified by cluster analysis. Aspect 7 retrieving a plurality of three-dimensional coordinates representing anatomical features of the foot into a plurality of three-dimensional homologous models for a plurality of body shapes; A step of acquiring a group of three-dimensional coordinates indicating anatomical features corresponding to the three-dimensional homologous model in a body shape of a subject as measurement data; evaluating a degree of similarity between a set of three-dimensional coordinates representing anatomical features in the measurement data and a set of three-dimensional coordinates representing anatomical features in the plurality of three-dimensional homologous models; determining a three-dimensional homologous model similar to a body shape of the measurement subject from among the plurality of three-dimensional homologous models based on the evaluation of the similarity; A body shape analysis method comprising: [Explanation of symbols]

[0074] 10 measurement subject, 20 measurement data acquisition section, 21 model generation section, 22 model storage section, 24 evaluation determination section, 26 output section, 30 analysis processing section, 31 cluster analysis processing section, 32 principal component analysis processing section, 35 statistical processing section, 50 shape analysis device, 104 three-dimensional homology model.

Claims

1. a model storage unit that stores in advance a group of three-dimensional coordinates indicating anatomical features of at least a part of a body as a plurality of three-dimensional homologous models for a plurality of body shapes; a measurement data acquisition unit that acquires, as measurement data, a group of three-dimensional coordinates that indicate anatomical features corresponding to the three-dimensional homologous model in a body shape of a subject; an evaluation and determination unit that evaluates a similarity between a group of three-dimensional coordinates indicating anatomical features in the measurement data and a group of three-dimensional coordinates indicating anatomical features in the plurality of three-dimensional homologous models to determine a three-dimensional homologous model that is similar to a body shape of the measurement subject from the plurality of three-dimensional homologous models; A body shape analysis device comprising:

2. The body shape analysis device of claim 1, characterized in that the evaluation determination unit obtains position vectors from a predetermined reference point to each point in the multiple three-dimensional homology models, obtains position vectors from the predetermined reference point to each point of the three-dimensional homology models in the measurement data, and determines a three-dimensional homology model that is similar to the body shape of the subject from the multiple three-dimensional homology models based on the distance between points of corresponding three-dimensional homology models or the similarity of the position vectors.

3. The body shape analysis device according to claim 1, characterized in that the multiple three-dimensional homology models are divided into multiple clusters by cluster analysis based on the cosine similarity between position vectors from a predetermined reference point to each point of the multiple three-dimensional homology models and stored in the model storage unit.

4. a model storage unit that stores in advance a group of three-dimensional coordinates indicating anatomical features of the foot as a plurality of three-dimensional homologous models corresponding to a plurality of body shapes, and the plurality of three-dimensional homologous models are divided into a plurality of clusters by cluster analysis based on the cosine similarity between position vectors from a predetermined reference point to each point of the three-dimensional homologous models; an output unit that outputs a group of three-dimensional coordinates of the three-dimensional homology model for each cluster; A body shape analysis device comprising:

5. The body shape analysis device according to claim 3 or 4, characterized in that the plurality of three-dimensional homologous models are divided into a plurality of clusters in an upper hierarchy by cluster analysis based on cosine similarity, and then the lower hierarchy is further divided into a plurality of clusters by cluster analysis based on principal component analysis in which dimensionality is reduced to the second principal component or less excluding the first principal component when principal component analysis is performed for each cluster in the upper hierarchy.

6. The body shape analysis device according to claim 3 or 4, characterized in that the model storage unit further stores a value indicating each demand for a plurality of three-dimensional homologous models classified into a plurality of clusters, based on the number of the three-dimensional homologous models for each cluster corresponding to the plurality of clusters classified by cluster analysis.

7. retrieving a plurality of three-dimensional coordinates representing anatomical features of the foot into a plurality of three-dimensional homologous models for a plurality of body shapes; A step of acquiring a group of three-dimensional coordinates indicating anatomical features corresponding to the three-dimensional homologous model in a body shape of a subject as measurement data; evaluating a degree of similarity between a set of three-dimensional coordinates representing anatomical features in the measurement data and a set of three-dimensional coordinates representing anatomical features in the plurality of three-dimensional homologous models; determining a three-dimensional homologous model similar to a body shape of the measurement subject from among the plurality of three-dimensional homologous models based on the evaluation of the similarity; A body shape analysis method comprising:

Citation Information

Patent Citations

  • Selecting method for shoes

    JP2000090272A

  • Three-dimensional shape model generation device, three-dimensional shape model generation method, computer program, and three-dimensional model generation system

    JP2008171074A

  • Method and system for foot shape generation

    US20090051683A1

  • Footprint information distributing system

    WO2002040941A1