KR20230067733A - Cloth size automatic measurement method, apparatus and system - Google Patents

Abstract

Disclosed is an automatic clothing size measurement method by a computing device, which is provided by one aspect of the present invention, which can assist the operation of a shopping mall more efficiently. The method comprises: a step of acquiring a clothing-filmed image obtained by filming a piece of clothing as a subject (the clothing-filmed image includes an image for a standard object, which is a standard for converting a pixel distance into an actual distance, and information on an actual size of the standard object is pre-stored); a step of calculating a relation between the pixel distance compared to the actual distance of the clothing-filmed image based on the standard object in the clothing-filmed image; a step of extracting landmarks of the clothing in the clothing-filmed image; a step of measuring a pixel distance between the extracted landmarks; and a step of converting the measured pixel distance between the landmarks into an actual distance by using the calculated relation between the pixel distance compared to the actual distance.

Description

Clothing size automatic measurement method, apparatus and system {CLOTH SIZE AUTOMATIC MEASUREMENT METHOD, APPARATUS AND SYSTEM}

The present invention relates to an apparatus and method for automatically measuring size, and more particularly, to an apparatus and method for automatically measuring clothing size.

In the past, since most people visit offline stores to try on clothes before purchasing them, it is rare for people to choose the wrong size of clothes.

However, recently, the number of cases of purchasing clothes through online shopping is rapidly increasing. Since it is not possible to try on and purchase clothes online, accurately entering the size of clothes for sale online is a very important factor not only in sales but also in terms of consumer satisfaction. has become established

However, as online shopping mall operators deal with a large number of clothes in the shopping mall and new products arrive every day, a lot of effort is put into measuring and disclosing the accurate size of many clothes to be sold. It is increasingly difficult for shopping mall operators to perform these tasks.

In addition, since the standard of the measured clothing size is different for each shopping mall, there is a problem in that there are many cases in which consumers are confused and purchase clothes of the wrong size.

An object according to an aspect of the present invention for solving the above problems is to provide an automatic clothing size measuring device and measuring method for automating clothing size measurement to reduce the effort of a shopping mall operator for actual measurement.

According to one aspect of the present invention for achieving the above object, a method for automatically measuring clothing size in a computing device includes acquiring a clothing photographed image in which clothing is photographed as a subject (the clothing photographed image is a pixel distance (pixel distance)). distance) to an actual distance, and information on the actual size of the reference object is pre-stored), in the clothing photographic image, based on the reference object calculating a relationship between pixel distance and actual distance of the clothing photographed image, extracting landmarks of the clothing from the clothing photographed image, measuring pixel distances between the extracted landmarks, and and converting a pixel distance between the measured landmarks into an actual distance using the relationship between the calculated pixel distance and the actual distance.

The method further includes generating detailed clothing information using the converted actual distance, wherein the generating detailed clothing information includes determining a size of the clothing using the converted actual distance. It may include generating detailed information.

The method may further include acquiring category information of the clothing.

One clothing size pattern is defined by matching a plurality of landmarks, and a plurality of clothing size patterns may be determined based on the clothing category information.

The extracting of the landmarks includes extracting matching landmarks according to the determined plurality of clothing size patterns, and the measuring of the pixel distance between the extracted landmarks comprises matching the landmarks to one clothing size. It can be determined by measuring a distance pixel value of a line created by drawing a line connecting a plurality of landmarks constituting the pattern.

Clothing of the upper category may have a clothing size pattern including a shoulder width clothing size pattern, a sleeve length clothing size pattern, a chest section clothing size pattern, and a total length clothing size pattern.

For clothing in the lower category, a clothing size pattern may be determined including a waist circumference clothing size pattern and a total length clothing size pattern.

For clothing in the skirt category among clothing in the bottom category, the clothing size pattern is determined by further including a hip circumference clothing size pattern and a hem circumference clothing size pattern, and for clothing in the pants category among clothing in the bottom category, A clothing size pattern may be determined further including a crotch length clothing size pattern, a hem length clothing size pattern, and a thigh circumference clothing size pattern.

The reference object includes at least one of a two-dimensional figure having a plurality of edges and a three-dimensional object having a plurality of edges and a plurality of faces, and at least one of the actual length of each edge and the actual size of each face of the reference object. is stored in advance, and calculates the ratio of the actual distance to the pixel distance based on at least one of the pre-stored actual length of each corner and the actual size of each side to measure the actual distance between landmarks in the clothing shooting image. can

At least one of the pre-stored actual length of each corner and the actual size of each surface is compared with at least one of the length of the edge and the size of the surface of the reference object in the clothing photographed image to determine a photographing angle and a photographing distance, and determine Actual distances between landmarks in the clothing photographed image may be converted using the result.

The landmark may be detected using an object detection algorithm in the clothing photographed image.

The landmark may be at least one of a convexhull point and an extreme point.

In applying the relationship between pixel distance and actual distance, different ratios are applied to individual clothing size patterns, and the different ratios may be determined by a machine learning algorithm.

In applying the relationship between pixel distance and actual distance, an estimation error corresponding to an individual clothing size pattern is applied, and the estimation error corresponding to the individual clothing size pattern may be calculated by machine learning.

According to another aspect of the present invention for achieving the above object, an automatic clothing size measurement device includes a clothing photographed image in which clothing is captured as a subject - the clothing photographed image converts a pixel distance into an actual distance An image of a reference object serving as a reference for image acquisition is acquired, and a relation between a pixel distance and an actual distance of the clothing captured image is calculated based on the reference object in the clothing captured image, and the clothing captured image extracting the landmarks of the clothing from within, measuring pixel distances between the extracted landmarks, and using the calculated pixel distance versus actual distance relationship to determine the actual pixel distances between the measured landmarks. A processor that converts the distance into a distance and a memory electrically connected to the processor and storing instructions related to the operation of the processor, the memory may store information about the actual size of the reference object.

The device may further include a photographing module for photographing the clothing image.

According to another aspect of the present invention for achieving the above object, an automatic clothing size measuring system provides a reference object that is a standard for converting a pixel distance into an actual distance (information on the actual size of the reference object is the clothing size pre-stored in the automatic measurement device), a fixing plate configured to fix clothing as a subject, a camera generating a clothing photographic image by photographing the clothing on the fixing plate, and the reference object, and the clothing photographed image from the camera. obtains, calculates a relation between a pixel distance and an actual distance of the clothing photographed image based on the reference object in the clothing photographed image, extracts a landmark of the clothing in the clothing photographed image, and an automatic clothing size measuring device for measuring a pixel distance between the extracted landmarks and converting the pixel distance between the measured landmarks into an actual distance using the calculated pixel distance versus actual distance relationship. .

The system may further include a camera fixing unit for fixing the camera, and the camera fixing unit may fix the camera so as to be spaced apart from the center of the clothing by a reference distance in a vertical direction.

According to another aspect of the present invention for achieving the above object, a computer readable storage medium comprising instructions executable by a processor, when the instructions are executed by the processor, clothing photographed as a subject Acquire a photographed image (the clothing photographed image includes an image of a reference object that is a reference for converting a pixel distance into an actual distance, and information on the actual size of the reference object is stored in advance present), calculates a relation between pixel distance and actual distance of the clothing photographed image based on the reference object in the clothing photographed image, extracts a landmark of the clothing in the clothing photographed image, It may be configured to measure pixel distances between the extracted landmarks and convert the pixel distances between the measured landmarks into actual distances using the calculated pixel distance versus real distance relationship.

According to the automatic clothing size measurement apparatus and method of the present invention, as the clothing size measurement is automatically performed by the photographing device and the computing device, the effect of reducing the shopping mall operator's effort to measure clothing and helping to operate the shopping mall more efficiently there is

1 is a block diagram schematically showing an apparatus for automatically measuring clothing size according to an embodiment of the present invention;
2 is a conceptual diagram showing a system configuration in more detail with the addition of a mechanical configuration interlocking with the automatic clothing size measuring device of FIG. 1;
3 is a conceptual diagram for explaining a method for automatically measuring clothing size according to an embodiment of the present invention;
4A to 4C are conceptual diagrams for explaining a process of measuring the size of clothing in the T-shirt category by the method of FIG. 3;
5A to 5C are conceptual diagrams for explaining the process of measuring the size of clothing in the lower skirt category by the method of FIG. 3;
6A to 6C are conceptual diagrams for explaining a process of measuring the size of clothing in a shirt category by the method of FIG. 3;
7A to 7B are conceptual diagrams for explaining a process of measuring the size of clothing in the lower pants category by the method of FIG. 3;
8 is a conceptual diagram for explaining a method for automatically measuring clothing size according to another embodiment of the present invention;
9 is a view showing a standardized table of sizes defined differently for each shopping mall in the clothing size standardization method according to an embodiment of the present invention;
10 is a detailed flowchart showing a method for standardizing a clothing size according to an embodiment of the present invention in detail;
11A and 11B are diagrams showing a standardized table of attributes defined differently for each individual shopping mall by the method of FIG. 10;
12 is a conceptual diagram showing a system for automatically measuring clothing size and operating related applications according to an embodiment of the present invention;
13 is a flowchart for explaining a method of operating an application operated by the system of FIG. 12;
14 is a view showing a screen in which a product is uploaded to an actual shopping mall through the application of FIG. 12;
15 is a view showing a detail cut for a specific product of FIG. 14;
16 is a diagram showing a user interface for modifying an automatic clothing size measurement result by way of example;
17 is a block diagram showing the configuration of a computing device according to an embodiment of the present invention.

Since the present invention can make various changes and have various embodiments, specific embodiments are illustrated in the drawings and described in detail.

However, this is not intended to limit the present invention to specific embodiments, and should be understood to include all modifications, equivalents, and substitutes included in the spirit and scope of the present invention.

Terms such as first and second may be used to describe various components, but the components should not be limited by the terms. These terms are only used for the purpose of distinguishing one component from another. For example, a first element may be termed a second element, and similarly, a second element may be termed a first element, without departing from the scope of the present invention. The terms and/or include any combination of a plurality of related recited items or any of a plurality of related recited items.

It is understood that when an element is referred to as being "connected" or "connected" to another element, it may be directly connected or connected to the other element, but other elements may exist in the middle. It should be. On the other hand, when an element is referred to as “directly connected” or “directly connected” to another element, it should be understood that no other element exists in the middle.

Terms used in this application are only used to describe specific embodiments, and are not intended to limit the present invention. Singular expressions include plural expressions unless the context clearly dictates otherwise. In this application, the terms "include" or "have" are intended to designate that there is a feature, number, step, operation, component, part, or combination thereof described in the specification, but one or more other features It should be understood that the presence or addition of numbers, steps, operations, components, parts, or combinations thereof is not precluded.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with the meaning in the context of the related art, and unless explicitly defined in the present application, they should not be interpreted in an ideal or excessively formal meaning. don't

Hereinafter, with reference to the accompanying drawings, preferred embodiments of the present invention will be described in more detail. In order to facilitate overall understanding in the description of the present invention, the same reference numerals are used for the same components in the drawings, and redundant descriptions of the same components are omitted.

1 is a schematic block diagram of an apparatus for automatically measuring clothing size according to an embodiment of the present invention. As shown in FIG. 1 , the apparatus for automatically measuring clothing size according to an embodiment of the present invention may include a photographing module 110 and an automatic clothing size measuring module 120 .

Referring to FIG. 1 , a photographing module 110 is a module that generates and outputs a photographic image of clothing by photographing a subject as clothing. This includes image processing devices such as cameras. The photographing module 110 provides the clothing photographed image to the clothing size automatic measurement module 120 . In one example, like a smart phone, the automatic clothing size measurement device may be implemented in a form including the photographing module 110 and the automatic clothing size measurement module 120 in one device. In another example, the photographing module 110 is separately provided as a camera device, and the automatic clothing size measurement device is implemented as a computing device including only the clothing size automatic measurement module 120 . At this time, the computing device communicates with the photographing module 110 to receive the clothing measurement image. Then, the device may be implemented as a camera and a dual device in a method of automatically measuring the size of clothing, which is a subject in the photographed image. The photographing module 110 may photograph clothes by placing them on a support plate (which may be referred to as a fixing plate). At this time, the photographing module 110 is preferably photographed in a direction and position perpendicular to the clothing as the subject. That is, a direction perpendicular to the center of the subject's clothing is appropriate for the photographing angle of the clothing photographed image. In a system capable of fixing the photographing module 110, the photographing module 110 may be fixedly mounted in a direction perpendicular to the fixing plate. It is preferable to arrange the clothing photographed images at an appropriate distance so that the entire shape of the clothing is included, but not smaller than the reference value or larger than the reference value. At this time, it is preferable to photograph while securing an appropriate zoom distance.

The automatic clothing size measurement module 120 acquires the clothing photographed image from the photographing module 110, extracts landmarks in the image, and performs an algorithm that converts pixel distances (distances in the image) into actual distances. This is a module that determines the size of clothes by calculating the actual distance. This may be implemented as a computing device including a processor (not shown) and a memory.

The clothing size automatic measurement module 120 may include a landmark extraction unit 122 , an inter-landmark distance measurement unit 124 , an actual distance conversion unit 126 and a size determination unit 128 .

The landmark extractor 122 extracts landmarks from the clothing photographed image using an object detection algorithm. The landmark is composed of a set of reference points related to the size of clothing. It includes two or more points. The landmark may be one or more points of the outline of clothing in the image. Alternatively, it may be one of feature points representing the shape of clothing, such as a center point of clothing and a point derived from a landmark point forming the contour (for example, a point on a line generated in a vertical/horizontal direction from a contour landmark). . For example, in the case of a T-shirt, a portion where the curvature of the contour line changes may be a landmark for measuring a shoulder line. Also, for the measurement of the total length, the lowest point and the highest point of the clothing may be extracted as landmarks. A landmark may be identified by coordinates in an image. They can be identified by giving a name or index.

The distance measurement unit 124 between landmarks is a component that measures a pixel distance between two landmarks by matching them to at least some of the extracted landmarks. The extracted landmarks are identified by coordinates, and one landmark functions as a meaningful clothing size pattern when associated with the corresponding other landmark. That is, as described above, the overall length becomes one clothing size pattern consisting of a pair of the uppermost point and the lowermost point of the clothing outline. In other words, the clothing size pattern is composed of two landmarks, and the distance measurement unit 124 between landmarks calculates the pixel distance between them based on the coordinates of the two landmarks constituting the clothing size pattern, Measure the distance (pixel distance) within the image of the size pattern. A plurality of clothing size patterns may exist in one clothing. In addition, clothing size patterns may be configured differently according to clothing categories and attributes. For example, if clothing size patterns are composed of a shoulder section, armhole, chest section, sleeve length, and total length for a T-shirt, a clothing size pattern is composed of a waist circumference, hip circumference, hem circumference, and total length for a skirt.

The actual distance conversion unit 126 converts the pixel distance of individual clothing size patterns calculated by the landmark distance measurement unit 124 into an actual distance. A reference object for determining the accumulation of the image may exist in the clothing photographed image. The actual distance calculator 126 may know the actual size (or length) of the reference object in advance. If the actual size of the reference object is known, the actual distance of one pixel or the actual distance between two points in the image may be calculated based on the size of the reference object in the clothing image. For example, if a 5 cm long reference object is displayed with 50 pixels in an image, one pixel may correspond to 0.1 cm. By calculating the actual distance of the pixels in the image in this way, the pixel distance of the two landmarks constituting the clothing size pattern is converted into the actual distance. For example, as in the above embodiment, if one pixel points to 0.1 cm and the distance between two landmarks is 210 pixels, the clothing size pattern composed of the distance between the two landmarks is 21 cm. If this indicates the shoulder section, the length of the shoulder section of the subject's clothing is actually determined to be 21 cm. In other words, the actual distance converter 126 calculates the actual distance between landmarks by substituting the proportional relationship between the actual length of the reference object and the length (pixel distance) in the image into the distance between landmarks.

The size determination unit 128 determines the overall size of the corresponding clothing by collecting the actual lengths of the clothing size patterns. The size determination unit 128 stores a comparison table of total size indices corresponding to the reference values of clothing size patterns in a memory (not shown), and retrieves the comparison table, and converts distance values of actually measured clothing size patterns to the stored table. By comparing with , the total size index is calculated. For example, a pants size of 28 may be set to a thigh circumference of 28 to 30 cm, a crotch circumference of 26 to 28 cm, a hem circumference of 21 to 22 cm, and a total length of 105 to 110 cm. For example, when actual clothing size patterns are calculated as thigh circumference 29cm, crotch circumference 27cm, hem circumference 21.5cm, and total length 109cm, the size of the corresponding pants may be determined as size 28 according to the stored table. A plurality of reference values of clothing size patterns for calculating the total size index may be generated in correspondence to categories and attributes of individual clothing and stored in memory in advance.

On the other hand, although it is expressed as "clothing" in this specification, other fashion items (eg, shoes, shoes, etc.) that require size measurement while being handled in a shopping mall are also applied with the automatic size measurement method according to an embodiment of the present invention. It will be obvious to those skilled in the art that size measurement is possible.

FIG. 2 is a conceptual diagram showing a system configuration in more detail in which a mechanical configuration interworking with the automatic clothing size measuring device of FIG. 1 is added.

Referring to FIG. 2 , an automatic clothing size measuring device (not shown) interworks with a fixing plate 215 and a photographing device 220 (which may be a camera). The fixing plate 215 is a panel provided to properly position the clothing 205 to be measured for size. The fixing plate 215 may be provided on the table 210 . In another example, the fixing plate 215 may be provided on a wall surface. A white board or the like may also be the fixing plate 215 . The fixing plate 215 preferably has a larger size than the target clothing 205 .

A target garment 205 is placed on the fixing plate 215 . It is preferable that the target clothing 205 is arranged at or close to the center of the fixing plate 215 so that a part of the clothing 205 does not reach a range beyond the fixing plate 215 during shooting.

Meanwhile, the reference object 230 as well as the target clothing 205 are arranged on the fixing plate 215 . The reference object 230 is an object used to convert a pixel distance into an actual distance in a photographic clothing image of the target clothing 205 . The automatic clothing size measuring device knows in advance the length, shape, angle, etc. of each side of the reference object 230, analyzes the form in which the reference object 230 is captured in the clothing photographed image, and actually measures the pixel distance. convert to distance The reference object 230 may be a 2D figure and/or a 3D object.

In the system according to the embodiment of FIG. 2 , the photographing module 220 is preferably installed to be spaced apart from each other by a reference distance in a direction perpendicular to the fixing plate 215 . The photographing module 220 may be installed in a fixed form through a fixing unit. The photographing module 220 photographs an area of at least a portion of the fixing plate 215 with the clothing 205 as the center. At this time, it is preferable to take a picture so that the reference object 230 is included. The photographing module 220 transmits the photographed image to the clothing size automatic measurement module using wired or wireless communication.

According to another embodiment of the present invention, in a state in which the fixing plate 215 is arranged in a fixed form on the wall, an automatic clothing size measurement device including a photographing module, such as a smart phone, directly measures the clothing 205 and the reference object ( 230), it is possible to perform both image capture and automatic size measurement of the captured image in one device without going through a process of transmitting and receiving the captured image to another device.

3 is a conceptual diagram for explaining a method for automatically measuring clothing size according to an embodiment of the present invention.

Referring to FIG. 3 , a photographed image of clothing includes an area of clothing 305 as a subject and a reference object area 330 . In this embodiment, the reference object 330 is described as having a three-dimensional shape such as a rectangular parallelepiped, but it may be a two-dimensional figure. For example, it may be a three-dimensional solid shape such as a circle, a triangle, a quadrangle, other polygons, or a sphere, a cube, a cuboid, or a cylinder. The length and width of each corner and each side of the reference object 330 may be stored in advance in the automatic clothing size measuring device. For example, as in the embodiment of FIG. 3 , when the reference object 330 is a rectangular parallelepiped, the actual left-right length of the upper surface may be pre-stored as 10 cm. Through this, it is possible to convert the pixel distance in the image into the actual distance by analyzing how many pixels are displayed on the upper surface of the cuboid in the clothing photographic image.

Meanwhile, in the clothing photographed image, the reference object 330 may be arranged around the clothing 305, which is a target object. It is preferable to arrange at one corner of the upper left corner, the upper right corner, the lower left corner, and the lower right corner. The device extracts a landmark of clothing 305 as a subject in the clothing photographed image. A landmark may consist of a plurality of points A to J. These can be extracted by various object detection algorithms. The device selects at least some of the detected landmarks for use as a clothing size pattern. One clothing size pattern can be created by matching two landmarks among them. For example, a first clothing size pattern may be formed by matching landmarks A and J. A second clothing size pattern may be formed by matching landmarks A and B. As described above, these clothing size patterns may be defined corresponding to the category and/or attribute of the clothing 305 . Therefore, based on this, a plurality of clothing size patterns are extracted using at least some of the landmarks. Then, their pixel distances are measured in the apparel image.

Next, the device converts the pixel distance of the clothing photographed image into an actual distance using the reference object 330 . As described above, the left and right lengths of the upper surface of the reference object 330 are 10 cm. If this is displayed as 250 pixels in the image, one pixel corresponds to an actual length of 0.04 cm. The device converts the pixel distance of the clothing size pattern into an actual distance by applying this correspondence to the individual clothing size pattern. For example, if the first clothing size pattern (A-J) has a pixel distance of 300, this may result in an actual length of 300 X 0.04 cm = 12 cm. In this way, all actual distances of a plurality of clothing size patterns related to the corresponding clothing 305 are obtained. If the reference object 330 is a 3D object and is photographed at an oblique angle, the reference object 330 will be photographed in an image in a 3D shape as shown in the right side of FIG. 3 . Based on this, the device may analyze the degree of inclination in the vertical direction and the degree of inclination in the horizontal direction to calculate the vertical and horizontal imaging angles (θ, φ). Then, these photographing angles (θ, φ) may be reflected when converting the actual distance. In some cases, the device may calculate a photographing angle based on an asymmetry of distances between landmarks in a photographed image of clothing. For example, in the case of a bilaterally symmetrical shirt, if the left sleeve is photographed longer than the right sleeve, it may be determined that the photograph is biased to the left. In this case, a photographing angle may be inferred in consideration of asymmetry of landmarks to be present at corresponding left and right positions, and an actual distance may be calculated by reflecting this.

After the device calculates all actual distances for a plurality of clothing size patterns corresponding to the clothing 305, the total clothing size may be determined by collecting them and comparing them with a previously held clothing size reference table. Then, it can be created as detailed information of the clothing 305 and uploaded on a web or application related to a shopping mall.

According to an embodiment of the present invention, the device classifies clothing categories and/or attributes and stores corresponding clothing size patterns in advance. Usually, there are a plurality of clothing size patterns corresponding to one clothing category. Hereinafter, a corresponding clothing size pattern according to each clothing category will be described by way of example.

4A to 4C are conceptual diagrams for explaining a process of measuring the size of clothing in the T-shirt category by the method of FIG. 3 .

Referring to FIG. 4A , when clothing to be measured for size is a "top-t-shirt", clothing size patterns may include shoulder width, sleeve length, chest section, and total length (total length). The shoulder width clothing size pattern is created by matching landmarks A and B in the clothing photographed image. As shown in FIG. 4A , in the two-dimensional plan view consisting of the x-axis and the y-axis, landmarks A and B are the starting points of the sleeve part of the upper part of the clothing and become the point where the curvature is changed. To measure the pixel distance of shoulder width, the device draws a line from point A to point B, and measures the distance pixel value of the line. Second, the sleeve length can be made up of landmarks B and C. Point C may be the point where the sleeve ends. The device draws a line connecting points B and C, and measures the distance pixel value of the line. Next, the chest section consists of points D and E. Point D is a point below points A and D, and may be defined as a point at the lower end of the beginning of the sleeve. The device can measure the distance of the chest section by drawing a line connecting points B and C and measuring a distance pixel value of the line. Finally, the rector is made up of F and G points. Point F may be the topmost point of the T-shirt. Alternatively, the x value of point F may be the center point of A and B, but the y value of point F may be defined as the y value of a point constituting the uppermost part of the outline of the corresponding clothing image. The x value of point G is the same as that of point F, and the point where it meets the lowermost edge of the clothing can be defined as the y value. In this way, when the coordinates of points F and G are secured, a line connecting the two points is drawn to calculate a pixel distance value of the line. In another example, the total length may be calculated as a difference between the highest value and the lowest value based only on the y value among the points constituting the outline of the entire garment.

The right side view of FIG. 4A is a view for explaining the configuration of an automatic clothing size measurement system in which clothes 405 are hung on a fixing plate composed of a white board and photographed together with a circular reference object 430 . The apparatus unfolds the T-shirt 405 as much as possible and takes a picture on the fixing plate, and arranges the reference object 430 on the top right of the T-shirt 405 to serve as a standard for automatic size measurement.

Referring to FIG. 4B , the apparatus calculates clothing size patterns for the T-shirt in FIG. 4A through line drawing. In the left view of FIG. 4B , shoulder width is calculated by line 410 . The chest section can be calculated by line 412, and the total length can be calculated by line 414.

The diagram on the right side of FIG. 4B is a table showing the relationship between the measured distance and the estimated value obtained by using the ratio of the measured length of the T-shirt to the length in the image in the state where the circular magnet serves as the reference object. Looking at the actual length, the circular magnet has a diameter of 2.4 cm, the shoulder width of the T-shirt is 47 cm, the chest section is 54 cm, and the total length is 70 cm. However, in the image, the length of the reference object, the circular magnet, is measured as 0.42 cm, the shoulder width is 8.78 cm, the chest section is 9.92 cm, and the total length is 12.94 cm. Then, the actual and image ratios are calculated for each clothing size pattern, including the reference object, and an average thereof is calculated.

In this embodiment, the apparatus calculates an estimated (actually measured) distance by multiplying the averaged ratio by the pixel distance of the image, and calculates an estimation error by comparing it with the actual distance. It can be seen that the estimation error is calculated as a very small value of about 1 cm.

On the other hand, as described above, the device may calculate an estimate using a ratio of the pixel distance to the actual distance of the reference object, rather than the ratio calculated as the average. Alternatively, based on a large number of learning data, the pixel distance to actual distance ratio of the reference object and the pixel distance to actual distance ratio of each clothing size pattern are learned in advance to determine the actual distance based on the pixel distance to actual distance ratio of the reference object. A pixel distance versus actual distance of each clothing size pattern may be calculated, and lengths of actual clothing size patterns may be estimated based on the calculated pixel distance versus actual distance of each clothing size pattern. For example, basically, the measured length is calculated by applying the ratio of the measured distance to the pixel distance of the reference object to each individual clothing size pattern, but by the machine learning algorithm, the first clothing size pattern is the pixel distance to the measured distance of the reference object If the tendency of the length increased by 10% is inferred, the measured distance is calculated by applying the ratio of "pixel distance to measured distance ratio x 1.1" of the reference object to the first clothing size pattern. When it is inferred that the second clothing size pattern tends to have a length that is 5% less than the measured distance of the reference object, the device calculates the "pixel distance to measured distance ratio x 0.95" of the reference object for the second clothing size pattern. Calculate the measured distance by applying the ratio. In this case, a more accurate estimate may be calculated by adding or subtracting the estimation error learned by the machine learning algorithm for learning the estimation error.

Referring to the drawing on the left of FIG. 4C , the apparatus extracts landmarks whose contours and curvatures change by using a convexity defects function (image_CV) to detect the contours of an image. Through the convexity defect function, a landmark point value and its associated lines are extracted by comparing a contour line and a hull value. The device calculates the starting index constituting the convex hull, the index of the next connected part, the farthest index to the contour connected to the part connecting the two vertices, and the approximate distance to the farthest point. can do.

Meanwhile, after extracting landmark-related points and their coordinates, the device calculates the shortest distance between them by matching landmark pairs defined by a clothing size pattern. In this case, openCV functions may be used. In this embodiment, pixel distances of clothing size patterns such as shoulder width, sleeve length, and chest section can be calculated through this.

Referring to the right diagram of FIG. 4C, the device uses an extreme points function in parallel or independently of the contiguous defect function to land lands such as the uppermost point, the leftmost point, the rightmost point, and the lowest point. marks can be extracted. Through this, a gun length and the like can be calculated.

5A to 5C are conceptual diagrams for explaining a process of measuring the size of clothing in the lower skirt category by the method of FIG. 3 .

Referring to FIG. 5A , when clothing to be measured for size is “bottom-skirt”, clothing size patterns may include total length, waist circumference, hip circumference, and hem circumference. First, a waist circumference clothing size pattern is formed by matching landmarks A and B in a photographed clothing image. In the plan view of FIG. 5A , landmarks A and B are extracted as starting points of both ends of the upper edge of the skirt. The device draws a line from point A to point B, and measures the distance pixel value of the line. Second, the hip circumference can be made up of landmarks C and D. Points C and D may be portions where curvature changes among lines from points A and B to points E and F at both ends of the lower edge. Alternatively, it may be defined as a portion where a parallel line below a certain length from the upper edge meets the contour line of the skirt. The device draws a line connecting points C and D, and measures the distance pixel value of the line. Next, the hem circumference consists of points E and F. Points E and F can be defined as both ends of the lower edge of the skirt. The device can measure the distance around the hem by drawing a line connecting points E and F and measuring a distance pixel value of the line. Finally, the total length can be defined by the distance between the center of the AB line and the center point of the hem circumference, or the difference between the uppermost y value and the lowermost y value among the points forming the skirt outline.

The right side view of FIG. 5A is a view for explaining the configuration of an automatic clothing size measurement system in which a skirt 505 is hung on a fixing plate composed of a white board and photographed together with a circular reference object 530 . The apparatus unfolds the skirt 505 as much as possible to take a picture on the fixing plate, and arranges the reference object 530 at the top right of the skirt 505 to serve as a standard for automatic size measurement.

Referring to FIG. 5B , the apparatus calculates clothing size patterns for the T-shirt in FIG. 5A through line drawing. In the left diagram of FIG. 5B , the waist circumference is calculated by line 510 . Hip circumference can be calculated by line 512, hem circumference can be calculated by line 514, and total length can be calculated by line 516.

The drawing on the right side of FIG. 5B is a table showing the relationship between the measured distance and the estimated value estimated using the ratio of the measured length of the skirt to the length in the image in a state where the circular magnet serves as the reference object. Looking at the actual length, the circular magnet has a diameter of 2.4 cm, the waist circumference of the skirt is 33 cm, the hip circumference is 53 cm, the total length is 72 cm, and the hem circumference is 79 cm. However, in the image, the length of the reference object, the circular magnet, is measured as 0.6 cm, the waist circumference is 6.88 cm, the hip circumference is 11.01 cm, the total length is 15.98 cm, and the hem circumference is 17.15 cm. Then, the actual and image ratios are calculated for each clothing size pattern, including the reference object, and an average thereof is calculated.

In this embodiment, the apparatus calculates an estimated (actually measured) distance by multiplying the averaged ratio by the pixel distance of the image, and calculates an estimation error by comparing it with the actual distance. It can be seen that the estimation error is calculated as a very small value of about 1 to 2 cm.

In another example, the device may calculate the estimate from the ratio of pixel distance to actual distance of the reference object rather than an averaged ratio. Alternatively, based on a large number of learning data, the pixel distance to actual distance ratio of the reference object and the pixel distance to actual distance ratio of each clothing size pattern are learned in advance to determine the actual distance based on the pixel distance to actual distance ratio of the reference object. A pixel distance versus actual distance of each clothing size pattern may be calculated, and lengths of actual clothing size patterns may be estimated based on the calculated pixel distance versus actual distance of each clothing size pattern. In this case, a more accurate estimate may be calculated by adding or subtracting the learned estimation error.

Referring to the drawing on the left of FIG. 5C , the device extracts landmarks whose contours and curvatures change by using a convexity defect function (image_CV) to detect the contours of an image. After extracting these landmark-related points and their coordinates, the shortest distance between them is calculated by matching landmark pairs defined by the clothing size pattern. In this case, openCV functions may be used. In this embodiment, pixel distances of clothing size patterns such as waist circumference and hip circumference can be calculated through this.

Referring to the right diagram of FIG. 5C , the device uses an extreme points function together with or independently of the contiguity defect function to determine landmarks such as the highest point, the leftmost point, the rightmost point, and the lowest point. can be extracted. Through this, pixel distances of clothing size patterns such as total length and hem circumference may be calculated.

6A to 6C are conceptual diagrams for explaining a process of measuring the size of clothing in a shirt category by the method of FIG. 3 .

Referring to FIG. 6A , when clothing to be measured for size is a “top-shirt,” clothing size patterns may include a shoulder section, a sleeve section, a chest section, and a total length. The shoulder width clothing size pattern extracts the top points of the part where the curvature changes among the corners of the upper area, that is, the part where the sleeve length starts, as landmarks in the clothing shooting image, draws a line between the two landmarks, and measures the distance pixel value. do. The distance pixel values of the sleeve length, chest section, and total length are measured in the same or similar way as the T-shirt of FIG. 4a. The total length can be calculated by the difference between the y value of the landmark constituting the shoulder width and the lowest y value of the lower edge of the shirt.

6A shows an automatic clothing size measuring system in which a shirt 605 is hung on a fixing plate composed of a white board, and a plurality of fixing pins (or fixing magnets) are used to fix the shirt 605 to the white board. It is a drawing for explaining the configuration of. Here, a circular reference object 605 is included. The apparatus unfolds the shirt 605 as much as possible to take a picture on the fixing plate, and arranges the reference object 630 at the upper left of the shirt 605 to serve as a standard for automatic size measurement.

Referring to FIG. 6B , the apparatus calculates clothing size patterns for the shirt in FIG. 6A through line drawing. In the left view of FIG. 6B , shoulder width is calculated by line 610 . The chest section can be calculated by line 612, the sleeve length can be calculated by line 614, and the total length can be calculated by line 616.

The diagram on the right side of FIG. 6B is a table showing the relationship between the measured distance and the estimated value obtained by using the ratio of the measured length of the shirt to the length in the image in a state where the circular magnet serves as the reference object. Looking at the actual length, the circular magnet has a diameter of 2.4 cm, the shoulder width of the shirt is 54 cm, the chest section is 60 cm, the sleeve length is 51 cm, and the total length is 69 cm. However, in the image, the length of the reference object, the circular magnet, is measured as 0.64 cm, the shoulder width is 11.82 cm, the chest section is 14.61 cm, the sleeve length is 11.85727 cm, and the total length is 15.56 cm. Then, the actual and image ratios are calculated for each clothing size pattern, including the reference object, and an average thereof is calculated.

In this embodiment, the apparatus calculates the estimated distance by multiplying the averaged ratio by the pixel distance of the image, and calculates the estimation error by comparing it with the actual value. It can be confirmed that the estimation error is calculated as a very small value of around 1 to 3 cm.

In another example, the device may calculate the estimate from the ratio of pixel distance to actual distance of the reference object rather than an averaged ratio. Alternatively, based on a large number of learning data, the pixel distance to actual distance ratio of the reference object and the pixel distance to actual distance ratio of each clothing size pattern are learned in advance to determine the actual distance based on the pixel distance to actual distance ratio of the reference object. A pixel distance versus actual distance of each clothing size pattern may be calculated, and lengths of actual clothing size patterns may be estimated based on the calculated pixel distance versus actual distance of each clothing size pattern. In this case, a more accurate estimate may be calculated by adding or subtracting the learned estimation error.

Referring to the drawing on the left of FIG. 6C , the apparatus extracts landmarks whose contours and curvatures change by using a convexity defects function (image_CV) to detect the contours of an image. After extracting these landmark-related points and their coordinates, the shortest distance between them is calculated by matching landmark pairs defined by the clothing size pattern. In this case, openCV functions may be used. In this embodiment, pixel distances of clothing size patterns such as shoulder width and chest cross section can be calculated through this.

Referring to the diagram on the right of FIG. 6C, the device uses an extreme points function together with or independently of the contiguity defect function to determine landmarks such as the highest point, the leftmost point, the rightmost point, and the lowest point. can be extracted. Through this, sleeve length, left and right length, and total length can be calculated.

7A to 7B are conceptual diagrams for explaining a process of measuring the size of clothing in the lower pants category by the method of FIG. 3 .

Referring to FIG. 7A , when clothing to be measured for size is “bottom pants,” clothing size patterns may include waist circumference, crotch length, thigh circumference, hem length, and total length. The waist circumference is calculated by extracting both ends of the upper edge of the pants as a pair of landmarks and measuring the distance between the two. Rise is calculated as the distance from the waistline to the groin line. That is, it can be calculated by extracting the center point of the waistline and the point where the two legs of the pants are split (groin point) as landmarks and measuring the distance therebetween. The thigh can be calculated by extracting the point where the horizontal direction from the groin point and the outline of the pants meet and the groin point as landmarks and measuring the distance between them. The hem can be calculated by extracting both end points of the lower edge where the leg part of the maji ends as landmarks and measuring the distance between them. The total length can be calculated by extracting one end of the waistline and one point of the hem as landmarks and measuring the distance between them.

Referring to FIG. 7B , the device may determine the overall size of the pants by comprehensively comparing the lengths of the individual clothing size patterns of the pants with reference to the pants size reference table. Basically, pants are based on waist size, so for a 28-inch waist, the thighs are 29cm, the rise is 27cm, the hem is 21.5cm, and the total length is 109cm. The standard range can be determined by ±1cm in the individual clothing size pattern. The reference range does not necessarily have to be determined only by ±1 cm from the reference value, and may be defined using other offset values such as ±0.5 cm and ±2 cm. The device determines the final size by comparing this reference range with the estimated length of the individual garment size pattern.

Referring to the left drawing of FIG. 7B , the apparatus extracts landmarks whose contours and curvatures change by using a convexity defects function (image_CV) to detect the contours of an image. After extracting these landmark-related points and their coordinates, the shortest distance between them is calculated by matching landmark pairs defined by the clothing size pattern. In this case, openCV functions may be used. In this embodiment, pixel distances of clothing size patterns such as waist circumference, rise, and thigh circumference can be calculated through this.

Referring to the diagram on the right of FIG. 7B , the device uses an extreme points function together with or independently of the contiguity defect function to determine landmarks such as the highest point, the leftmost point, the rightmost point, and the lowest point. can be extracted. Through this, a gun length and the like can be calculated.

According to an embodiment of the present invention, clothing size patterns extracted according to clothing categories and locations and characteristics of landmarks constituting the patterns may be learned by a machine learning algorithm. That is, the landmark extraction method may be learned by bundling a clothing shooting image of a specific clothing category and information on landmark coordinates thereof into one learning data and inputting a plurality of them. At this time, the correct answer data that is output is a shopping mall operator that the user wants to use as size information (or detailed information) for the clothing without modifying or deleting the actual length information of the clothing size pattern estimated through the above method. It may be actual measurement result data when there is a user input from

8 is a conceptual diagram for explaining a method for automatically measuring clothing size according to another embodiment of the present invention. As shown in FIG. 8 , a system to which the method of automatically measuring clothing size is applied may include a distance measurement module 810 and an automatic clothing size measuring device 820 .

Referring to FIG. 8 , similar to the previous embodiment, the system arranges clothing 805 on a fixing plate 830 and measures distances of landmarks B ₁ and B ₂ in the clothing 805. Determine your clothing size by At this time, the distance measurement module 810 may be utilized.

The distance measurement module 810 is a device (or module) that measures an actual distance to a specific point. The distance measurement module 810 includes a light-based distance measurer, preferably, may include a laser distance measurer, but is not necessarily limited thereto. The laser distance measurer measures an accurate distance by emitting a laser toward specific points (A ₁ , A ₂ ) in the clothing 805 and then detecting the reflected laser. At this time, the distances d ₁ and d ₂ to the plurality of reference points A ₁ and A ₂ in the clothing 805 are measured. The reference point in the clothing 805 may be a point for which the automatic clothing size measuring device 820 knows the location in advance. For example, the automatic clothing size measurement device 820 can know the center point (A ₁ ) of the clothing and the point (A ₂ ) where the center point (A ₁ ) meets the left or right outer line of the clothing through image analysis. A branch is enough. That is, it may be a point that can be identified as a singular point on the shape of clothing. The reference points A ₁ and A ₂ may be determined corresponding to the type of clothing 805 . The reference points A ₁ and A ₂ may be defined as one of landmarks constituting a clothing size pattern.

The automatic clothing size measuring device 820 obtains information about the distances _{d 1 and} d ₂ from the distance measurement module 810 to the reference points A 1 and A ₂ ( S810 ). This can be received through wired or wireless communication. Wireless communication may include broadband communication such as LTE and 5G as well as short-distance communication such as Bluetooth, zigbee, and Wi-Fi. In this case, the automatic clothing size measuring device 820 may calculate the distance between the distance measuring module 810 and the automatic clothing size measuring device 820 based on the strength of the wireless communication signal, and reflect the distance when measuring the actual distance. can do. In another example, the distance measuring module 810 may be integrally included in the automatic clothing size measuring device 820 and may acquire information about the distances d ₁ and d ₂ by exchanging information internally.

The automatic clothing size measurement device 820 calculates the actual distance d ₃ from point A ₁ to point A ₂ by triangulation using the distances d ₁ and d ₂ (S820). Then, a photographing module (not shown) of the automatic clothing size measuring device 820 determines the distance (pixel distance) within the image from point _A1 to point _A2 in the clothing photographed image of the corresponding clothing 805. Compare. Based on this, the relationship (ratio) between the pixel distance and the actual distance is calculated.

The automatic clothing size measuring device 820 extracts landmarks B ₁ and B ₂ related to the clothing 805 and measures a pixel distance of a clothing size pattern composed of the extracted landmarks. Then, the pixel distance of the clothing size pattern is converted into an actual distance using the relationship between the pixel distance calculated using d ₃ and the actual distance (S830). The method of extracting the corresponding clothing size pattern according to the clothing category 805 and calculating the actual distance of the extracted clothing size pattern based on the ratio of the actual distance to the pixel distance is the method of FIGS. 4A to 7B described above. can be run similarly to

9 is a diagram showing a standardized table of sizes defined differently for each shopping mall in the method for standardizing clothing sizes according to an embodiment of the present invention.

Referring to FIG. 9 , even if the category of clothes is the same for each shopping mall, the actual size-measured size list is configured differently. For example, in shopping mall 1, for the category of "clothing-top-knit top", the size list disclosing the actual size is composed of total length, shoulder, chest, hem, sleeve length, armhole, arm section, and sleeve hem. However, for products in the same category of “Clothing-Top-Knit Top”, shopping mall 2 has a size list consisting only of the shoulder section, chest section, armhole, sleeve length, and total length.

As described above, products of the same category are managed as different size lists for each shopping mall. Furthermore, even for the same product, different size lists are managed for each shopping mall. In this way, since products are managed in different sizes for the same clothing product, consumers are often confused because the clothing size is not unified.

According to the clothing size standardization method according to an embodiment of the present invention, the apparatus acquires a clothing shooting video file and size information of the corresponding clothing from a shopping mall operator terminal or an online shopping mall service platform operating server, and includes them in a standardized size list. It can be returned after being converted into the used clothing size patterns. In this case, the standardized size list may be determined corresponding to the type (category) of clothing.

10 is a detailed flowchart showing a method for standardizing a clothing size according to an embodiment of the present invention in detail.

Referring to FIG. 10 , in order to standardize clothing size information measured differently for each individual shopping mall, the apparatus first acquires clothing-related image information from an individual shopping mall (S1010). In this case, the image information may include information on the size of corresponding clothing measured differently for each individual shopping mall.

The device extracts a raw landmark dimension value based on the obtained image information related to clothing. Raw landmarks refer to landmarks constituting a clothing size pattern managed by measurement in a size list possessed by each individual shopping mall. That is, in the actual example of FIG. 9 , in the case of shopping mall 1, the size list including the total length, shoulder, chest, hem, sleeve length, hole in the arm, arm section, and sleeve hem for the P000FTQF product is image information and When received together, the landmarks constituting the above clothing size patterns become row landmarks. The device extracts landmarks from image information.

The device identifies the category and attribute of the target clothing (S1030). Categories are divided into tops and bottoms in the first class, and in the second class, tops can be divided into knit tops, shirts, T-shirts, hoodies, blouses, etc., and bottoms, skirts, pants, shorts, etc. can be divided into The attribute is information indicating the peculiarity of the clothing, and the device is a long sleeve attribute indicating that the arm sleeve or pants sleeve is a long form of clothing, a high neck or half high neck attribute indicating that the neck part is raised high, and a vest. Singularities in the shape of various garments, such as sleeveless properties indicating that the garment is in the form of sleeveless clothing, may be defined as attributes. Then, target clothes are classified according to the categories and attributes defined as above. This classification may be performed by an image analysis algorithm. In this case, the device may analyze the shape of the clothing in the image using machine learning to infer which category the clothing belongs to and what properties the corresponding clothing has.

After identifying the category and attribute of the target clothing, the device loads a standardized size list corresponding to the category and attribute (S1040). It may be stored in advance. This standardized size list is the same as or similar to the apparel size patterns included therein versus apparel categories defined in FIGS. 4A to 7B . Referring to the embodiment of FIG. 4A , the device loads shoulder width, chest section, sleeve length, and total length as a standardized size list of the T-shirt category. If it corresponds to another clothing category, a standardized size list corresponding thereto is loaded, and a plurality of standardized clothing size patterns are defined in the loaded standard size list.

Based on the row landmark related information extracted in step S1020, the device converts the measured distance value of the raw landmark to the measured distance value of the clothing size pattern included in the loaded standardized size list (S1050). The device first calculates a ratio of pixel distances between row landmarks in an image to actual distances of row landmarks (a ratio value of pixel distances to actual distances). Then, by substituting this into a pixel distance value between landmarks constituting the clothing size pattern included in the standardized size list in the image, the measured distance value of the clothing size pattern is converted. The conversion method may utilize the method of FIGS. 4A to 7B identically or similarly.

11A and 11B are diagrams illustrating tables standardizing attributes defined differently for each shopping mall by the method of FIG. 10 .

Referring to FIG. 11A , a device (which may be referred to as a "clothing size standardization device") receives an image and related size information for blouse product #1 from shopping mall 1, and sets its category to "clothing-over-blouse". classified as In addition, a “long sleeve” attribute is given as a property of the corresponding product. The raw landmark-based size list received from shopping mall 1 is shoulder section, chest section, sleeve length, hem section, and total length. The device uses the method of FIG. It can be converted to section, sleeve length, hem and total length. Images and size information of blouse products #2 and #3 may also be received from shopping malls 2 and 3. Shoulder section, chest section, armhole, sleeve length, and gun length information are received from shopping mall 2, and shoulder, chest, sleeve, armhole, sleeve sleeve, hem, and total length information are received from shopping mall 3. The device converts the above non-uniform size list into a standardized size list.

Referring to the drawing on the left of FIG. 11B , the device may obtain an image and size information of blouse product #4 of shopping mall 4. By analyzing the image, the device can classify the category as “over-clothing-blouse” and define the attribute of the corresponding product as “half-heinneck”. The size list of the corresponding shopping mall can be managed by shoulder width, chest section, armhole circumference, sleeve length, neck length/total length. Reflecting the attribute part, unlike the previous embodiment of FIG. 11A, the device adds a neck length to a standardized size list including shoulder section, chest section, armhole, sleeve section, sleeve length, neck length/gun length. manage

Referring to the right side view of FIG. 11B , the device may obtain an image and size information of blouse product #5 of shopping mall 5. The device may analyze the image to classify the category as “over-clothing-blouse” and define the attribute of the corresponding product as “sleeveless”. The size list of the corresponding shopping mall may be managed in terms of length, total length, sleeve A length, shoulder section, chest section, waist section, and hem section. The device reflects the attribute part and manages it as a standardized size list including shoulder section, chest section, armhole, bottom section, and total length.

In this way, the device may store different standardized size lists by reflecting categories and attributes of clothing, and classify categories and attributes of corresponding clothing through image analysis when clothing images are received to correspond to standardized sizes. A unified size system can be established by loading a list and converting the measured distances of clothing size patterns included therein.

12 is a conceptual diagram illustrating a system for automatically measuring clothing size and operating related applications according to an embodiment of the present invention. As shown in FIG. 12, the automatic clothing size measurement system according to an embodiment of the present invention includes shopping malls 1210-1 to 1210-N, an automatic clothing size measurement device 1220, and a shopping mall (service) platform operating server ( 1230) may be included.

Referring to FIG. 12 , shopping malls 1210 - 1 to 1210 - N may refer to shopping malls that sell clothing products through the Internet, and include a server handling information related to the shopping mall or a terminal of an operator of the shopping mall. The terminal may include a computing device such as a smart phone or a PC.

According to an embodiment of the present invention, the plurality of shopping malls 1210-1 to 1210-N may interwork with the automatic clothing size measuring device 1220 through a communication network (not shown). The shopping malls 1210-1 to 1210-N may upload original image information of clothing products and request the automatic clothing size measuring device 1220 to generate size information accordingly.

The automatic clothing size measuring device 1220 (which may be referred to as a “detailed clothing automatically generating device”) receives a request for generating the original image information and size information from at least one of the shopping malls 1210-1 to 1210-N, and , It is possible to automatically recognize clothing categories and attributes by performing image analysis on uploaded clothing shooting images, and automatically create a size list corresponding thereto. To this end, the method of automatically measuring the clothing size described above may be used. The automatic clothing size measuring device 1220 may be referred to as a “device” or a “server”. The automatic clothing size measuring device 1220 may be implemented as a computing device. The computer device referred to herein may be implemented as a server-class computer terminal. The computer device may include an input device of a typical computer terminal, a display device, a networking device, a hard disk, a memory for storing a program, and a processor for executing the program stored in the memory. However, it is not necessarily implemented as a server-class computer terminal.

The shopping mall platform operating server 1230 may be a server of an operator operating an Internet shopping mall platform. The shopping mall platform operation server 1230 may be implemented as a separate device from the automatic clothing size measuring device 1220, or may be implemented as the same device as needed. When implemented as a separate device, the shopping mall platform operating server 1230 may provide the clothing size measurement device 1220 with clothing photographed images and information related thereto. This may include information about raw landmark dimension values used for size standardization.

According to another embodiment of the present invention, the automatic clothing size measurement device 1220 may be a device that develops a program related to automatic size measurement as an application and controls and manages it through a web or app. The terminal associated with the shopping malls 1210-1 to 1210-N downloads an application provided by the automatic clothing size measurement device 1220, and performs the automatic size measurement operation performed by the automatic clothing size measurement device 1220 within the terminal. You can control it to do it yourself. That is, without a separate server device, a processor (not shown) of a terminal executes a downloaded application to automatically measure the size of clothes based on a photographed image, visualizes the size, and provides it to the user through a display unit.

13 is a flowchart illustrating a method of operating an application operated by the system of FIG. 12 .

Referring to FIG. 13 , a user terminal uploads an image of a clothing product, and the device receives it (S1310). The user terminal proceeds with an option selection step of determining which one of the product cut version and the image cut version to output as a result of size measurement, and the device controls to output the result with the selected option (S1320). Here, the product cut version is a version in which, after landmarks related to clothing size patterns are extracted, the extracted landmarks and corresponding size information are overlaid and displayed on the original image. The image cut version is an image of a size display-related summary version based on a line connecting at least some of the plurality of landmarks of the original clothing image. That is, the image cut can be regarded as a version in which the attributes and characteristics of the corresponding clothing are revealed based on the original clothing image, and the overall outline and shape are simplified and displayed.

When the option selection is completed, the device determines the category and attribute of the clothing, automatically measures the size of the clothing accordingly, and generates and visualizes the result (S1330). First, the device analyzes an image of an original clothing image to classify a corresponding clothing category. This uses a machine learning image analysis algorithm. That is, in a state in which feature points of the image are extracted and learned based on a plurality of (about 10,000) image learning data classified as T-shirts, when a T-shirt image to be judged is input, the image is selected as a T-shirt based on the feature points. It can be judged that In a similar way to category classification, clothing property classification can also be trained through other machine learning algorithms. As described above, the attributes include a V-neck, long sleeves, button-down collars that indicate a V-shaped neck shape, and a plain pattern that refers to clothing that is made entirely of one color without a special pattern. distinguished by various properties. One garment may have a plurality of attributes.

The device may automatically generate a product name for the clothing based on the category and attribute. Product names are generated by appropriately combining attributes and categories. For example, a blouse with V-neck, button-down, and long-sleeve attributes may be branded as “V-neck long-sleeve blouse,” “V-neck button-down long-sleeve blouse,” “button-down blouse,” or “long-sleeve button-down blouse.” can This can be generated based on an algorithm that collects product name tags and machine learns them.

Next, landmarks constituting a clothing size pattern corresponding to a category of an original clothing image are extracted by calling a landmark detection API (including convexity defect and extreme point functions, or various other object detection algorithms). Then, the pixel distance of the clothing size pattern is calculated based on the extracted landmarks, and the pixel distance of the clothing size pattern is calculated as the actual distance using the ratio of the actual distance to the pixel distance. Then, it is displayed as the result of the size table. At this time, corresponding to the option selected by the user, in the case of a product cut, a landmark may be displayed on the original image, and the size result may be overlaid and visualized. In the case of an image cut, a size result may be overlaid and displayed on a video converted into an image cut. In addition, automatically measured sizes may be provided as a table using a standardized size table. Information related to the output result is provided to the user terminal from the automatic clothing size measurement device so that the user can check it through a display unit of the user terminal.

FIG. 14 is a view showing a screen in which a product is uploaded to an actual shopping mall through the application of FIG. 12 .

Referring to FIG. 14 , the device may provide the results of FIG. 13 to a user terminal, and the user terminal may upload them to a shopping mall managed by the user according to a user's input command. That is, the apparatus for automatically measuring clothing size may provide the result as an e-commerce application (EC API) in conjunction with a shopping mall platform operating server, reflect it on an actual shopping mall site, and control uploading as detailed product information. At this time, product cuts (or image cuts) are arranged at the top, and automatically generated product names, categories, and attribute information may be displayed at the bottom. And, it can be controlled so that a detail cut is displayed at the bottom and size information is displayed in a table form at the bottom.

FIG. 15 is a view showing a detail cut of a specific product of FIG. 14 .

Referring to FIG. 15 , the device may derive a detail cut from an original clothing image. This is an image in which a part of high importance is enlarged and displayed in detail in a clothing shooting image. These detail cuts may be generated corresponding to categories and/or attributes to which the target clothing belongs. As in the embodiment of FIG. 15 , in the case of a blouse, the neck, chest, and hem are formed as detail cuts. That is, a part of clothing that shows the corresponding attribute well is selected as a target for detail cut and constitutes a part of detailed product information in an enlarged form.

16 is a diagram showing a user interface for modifying an automatic clothing size measurement result by way of example.

Referring to FIG. 16 , the apparatus for automatically measuring clothing size displays landmarks and lines connecting the landmarks on an image of clothing to be measured through a product cut or image cut, and displays the length of the corresponding line. At this time, a name or index is given to the line, and its length is displayed in a table form corresponding to the name or index.

Meanwhile, according to an embodiment of the present invention, a user may directly input a line name. For example, in line 1, the name "gijang" can be entered. In line 2, you can enter the name "shoulder". In addition, the automatically measured length can also be modified or deleted by the user. For example, if you want to modify the length of 28cm, you can click the corresponding length part and enter another value such as 29cm to modify it. Information related to the line can be deleted. The device provides a user interface for adding, deleting and modifying automatically measured information.

The line name input using the above user interface is visualized in the same form as the clothing size detail table at the bottom of FIG. 14 . Deleted parts may not appear here.

In another example, a user may directly draw and add a line to be measured. At this time, the coordinates of the start and end points of the line are identified and added as size data together with the name of the line given by the user. Meanwhile, the added information may be generated as learning data. By collecting such learning data, the device can be trained to be the subject of clothing size measurement through a machine learning tool. Accordingly, the device may generate a new clothing size pattern of a category related to target clothing. In addition, as a number of users give names to newly added lines, the names of the lines can also be updated by the machine learning tool.

17 is a block diagram showing the configuration of a computing device according to an embodiment of the present invention. As shown in FIG. 17 , the device may include a communication unit 1710, a processor 1720, a memory 1730, and an input/output module 1740. The computing device becomes an automatic clothing size measurement device (which may be a clothing size standardization device and/or an automatic clothing detailed information generating device). Alternatively, it may be a user terminal for downloading and executing an automatic clothing size measurement application.

Referring to FIG. 17 , a communication unit 1710 is a component for exchanging information with other devices such as a shopping mall operator terminal or a shopping mall platform server through a wireless and/or wired network. The communication unit 1710 includes a communication device associated with a wired or wireless network. The communication unit 1710 may include an antenna.

The processor 1720 receives the clothing photographed image received through the communication unit 1710 and automatically measures and outputs the size of the target clothing by executing the method of automatically measuring the clothing size. The measurement result may be provided to a shopping mall operator or a shopping mall platform operator (manager) handling the clothing. The processor 1720 may standardize clothing sizes, and may perform a function of automatically generating size information, detail cuts, product names, etc. of corresponding clothing from a photographed clothing image and uploading them to a shopping mall. The processor 1720 may execute an object detection algorithm, learn and execute various machine learning algorithms used for automatic size measurement, automatic product name generation, and the like. In some cases, learning of the machine learning algorithm may be executed through a separate device, and the learned algorithm may be executed in the processor 1720. The processor 1720 may self-learn the machine learning algorithm.

The memory 1730 is a storage device that stores instructions related to programs to be executed by the processor 1720 and various data requested by the processor 1720 . The memory 1730 may store clothing size pattern information and attribute information corresponding to clothing categories. The memory 1730 may be implemented as a local memory inside the device or as an external large-capacity database.

The input/output module 1740 includes information input means such as a keyboard and mouse, and information output means such as a monitor, TV, and touch screen. The input/output module 1740 is used to input and change various setting values such as line drawing, name of a new line, and the like, and to display related contents.

The system or apparatus described above may be implemented as hardware components, software components, and/or a combination of hardware components and software components. For example, systems, devices and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA) ), programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. A processing device may run an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of software. For convenience of understanding, there are cases in which one processing device is used, but those skilled in the art will understand that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that it can include. For example, a processing device may include a plurality of processors or a processor and a controller. Other processing configurations are also possible, such as parallel processors.

Software may include a computer program, code, instructions, or a combination of one or more of the foregoing, which configures a processing device to operate as desired or processes independently or collectively. The device can be commanded. Software and/or data may be any tangible machine, component, physical device, virtual equipment, computer storage medium or device, intended to be interpreted by or provide instructions or data to a processing device. , or may be permanently or temporarily embodied in a transmitted signal wave. Software may be distributed on networked computer systems and stored or executed in a distributed manner. Software and data may be stored on one or more computer readable media.

The method according to the embodiments may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program commands recorded on the medium may be specially designed and configured for the embodiment or may be known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with limited examples and drawings, those skilled in the art can make various modifications and variations from the above description. For example, the described techniques may be performed in an order different from the method described, and/or components of the described system, structure, device, circuit, etc. may be combined or combined in a different form than the method described, or other components may be used. Or even if it is replaced or substituted by equivalents, appropriate results can be achieved.

Therefore, other implementations, other embodiments, and equivalents of the claims are within the scope of the following claims.

Claims (19)

A method for automatically measuring clothing size in a computing device,
Obtaining a clothing photographic image in which clothing is photographed as a subject, wherein the clothing photographed image includes an image of a reference object serving as a reference for converting a pixel distance into an actual distance, and the actual image of the reference object Size information is pre-stored;
calculating a relation between a pixel distance and an actual distance of the clothing photographed image based on the reference object in the clothing photographed image;
extracting landmarks of the clothing from the photographic image of the clothing;
measuring a pixel distance between the extracted landmarks; and
and converting a pixel distance between the measured landmarks into an actual distance using the relationship between the calculated pixel distance and the actual distance. According to claim 1,
Further comprising generating detailed clothing information using the converted actual distance,
Wherein the generating detailed clothing information comprises generating the detailed clothing information by determining a size of the clothing using the converted actual distance. According to claim 1,
The method of automatically measuring the size of clothing, further comprising acquiring category information of the clothing. According to claim 3,
Define one clothing size pattern by matching a plurality of landmarks,
A method of automatically measuring clothing size, wherein a plurality of clothing size patterns are determined based on the clothing category information. According to claim 4,
The step of extracting the landmarks includes extracting matching landmarks according to the determined plurality of clothing size patterns,
The step of measuring the pixel distance between the extracted landmarks is determined by measuring a distance pixel value of a line generated by drawing a line connecting a plurality of landmarks that are matched and form one clothing size pattern. A method for automatically measuring clothing size. According to claim 4,
The method of automatically measuring clothing size, wherein clothing in the upper category includes a shoulder width clothing size pattern, a sleeve length clothing size pattern, a chest section clothing size pattern, and a total length clothing size pattern, wherein clothing size patterns are determined. According to claim 4,
A method for automatically measuring clothing size, wherein a clothing size pattern is determined by including a waist circumference clothing size pattern and a total length clothing size pattern for clothing in the bottom category. According to claim 7,
Clothing in the skirt category among clothing in the bottom category further includes a hip circumference clothing size pattern and a hem circumference clothing size pattern to determine a clothing size pattern,
The method of automatically measuring clothing size, wherein clothing in the pants category among clothing in the bottom category further includes a crotch length clothing size pattern, a hem length clothing size pattern, and a thigh circumference clothing size pattern, so that a clothing size pattern is determined. According to claim 1,
The reference object includes at least one of a two-dimensional figure having a plurality of edges and a three-dimensional object having a plurality of edges and a plurality of faces,
At least one of the actual length of each edge and the actual size of each side of the reference object is stored in advance;
Method for automatically measuring clothing size, which calculates a ratio of pixel distance to actual distance based on at least one of the pre-stored actual length of each corner and actual size of each side, and measures the actual distance between landmarks in the clothing image. . According to claim 9,
Comparing at least one of the pre-stored actual length of each corner and the actual size of each surface with at least one of the length of the edge and the size of the surface of the reference object in the clothing photographed image to determine a photographing angle and a photographing distance,
A method of automatically measuring a clothing size by using a determination result to convert an actual distance between landmarks in the clothing photographed image. According to claim 1,
The method of automatically measuring clothing size, wherein the landmark is detected using an object detection algorithm in the clothing photographed image. According to claim 11,
The method of automatically measuring clothing size, wherein the landmark is at least one of a convexhull point and an extreme point. According to claim 1,
In applying the pixel distance versus actual distance relationship, different ratios are applied for each individual clothing size pattern,
Wherein the different ratios are determined by a machine learning algorithm. According to claim 1,
In applying the pixel distance versus actual distance relationship, an estimation error corresponding to an individual clothing size pattern is applied,
The method of automatically measuring clothing size, wherein the estimation error corresponding to the individual clothing size pattern is calculated by machine learning. In the automatic clothing size measuring device,
A photographic clothing image in which clothing is photographed as a subject is obtained, wherein the photographic clothing image includes an image of a reference object serving as a standard for converting a pixel distance into an actual distance. In , based on the reference object, a relation between pixel distance and actual distance of the clothing photographed image is calculated, landmarks of the clothing are extracted from the clothing photographed image, and pixels between the extracted landmarks are calculated. a processor that measures distances and converts pixel distances between the measured landmarks into actual distances using a relationship between the calculated pixel distances and actual distances; and
A memory electrically connected to the processor and storing instructions related to the operation of the processor,
wherein the memory stores information about the actual size of the reference object. According to claim 15,
The automatic clothing size measurement device further comprising a photographing module for photographing the clothing image. In the automatic clothing size measuring system,
A reference object serving as a standard for converting a pixel distance into an actual distance, and information on the actual size of the reference object is pre-stored in the automatic clothing size measuring device;
a fixing plate configured to fix clothing as a subject;
a camera for generating a photographic image of clothing by photographing the clothing on the fixing plate and the reference object; and
The clothing captured image is acquired from the camera, a pixel distance versus actual distance relationship of the clothing captured image is calculated based on the reference object in the clothing captured image, and the land of the clothing in the clothing captured image Clothing size that extracts landmarks, measures pixel distances between the extracted landmarks, and converts the pixel distances between the measured landmarks into actual distances using the calculated pixel distance versus actual distance relationship An automatic garment size measurement system comprising an automatic measurement device. 18. The method of claim 17,
Further comprising a camera fixing unit for fixing the camera,
The automatic clothing size measurement system of claim 1 , wherein the camera fixing unit fixes the camera to be spaced apart from the center of the clothing by a reference distance in a direction perpendicular to the clothing size. A computer-readable storage medium containing instructions executable by a processor, which when executed by the processor:
A photographic image of clothing in which clothing is photographed as a subject is acquired, the photographic clothing image includes an image of a reference object serving as a standard for converting a pixel distance into an actual distance, and the actual size of the reference object Information about is stored in advance;
calculating a relation between a pixel distance and an actual distance of the clothing photographing image based on the reference object in the clothing photographing image;
extracting landmarks of the clothing from the photographic image of the clothing;
measure a pixel distance between the extracted landmarks; and
A computer readable storage medium configured to convert a pixel distance between the measured landmarks into an actual distance using the calculated pixel distance versus real distance relationship.

Images