Method of reading point group pattern in cloth and apparatus performing thereof

Abstract

A clothing point cloud pattern reading method executed in a clothing point cloud pattern reading device according to the present invention comprises: a step of capturing a clothing surface to generate a two-dimensional original image; a step of applying a projection transformation matrix to the two-dimensional original image to correct perspective distortion; a step of detecting wrinkle distortion in the corrected image and applying non-linear filtering to correct wrinkle distortion; a step of analyzing the curvature information of the clothing to perform a curvature-based flattening operation and then detecting a point cloud pattern; and a step of providing a service connected to the point cloud pattern.

Description

Technology Field

[0001] The present invention relates to a method for reading a clothing point cloud pattern and an apparatus for executing the same. More specifically, it relates to a method for reading a clothing point cloud pattern and an apparatus for executing the same, which enables accurate reading of the point cloud pattern regardless of the wearer's movement or wrinkles by applying a technique that scans the clothing, analyzes depth information of the clothing surface, and flattens it to restore the original point cloud pattern. Background Technology

[0003] Point cloud pattern embedding technology is a widely used technique for printed materials, security documents, and authentication. It is designed to embed specific patterns that are not easily identifiable with the naked eye but can be detected using dedicated devices or software. This technology is primarily applied through methods such as offset printing, digital printing, and laser marking, and is used for purposes such as anti-counterfeiting, verifying the authenticity of products, and tracing origins.

[0004] However, existing point cloud pattern insertion and reading technologies primarily target hard surfaces such as paper, plastic, and metal, and are based on processing images captured in a relatively uniform environment.

[0005] Therefore, there are limitations to embedding point cloud patterns into materials, such as clothing, that are highly susceptible to wrinkling or deformation, and accurately reading them. In particular, for clothing used in diverse environments, such as military uniforms, there is a high likelihood that the point cloud patterns will be distorted or severely wrinkled due to the wearer's movements. Consequently, it is difficult to restore the embedded point cloud patterns to their original form using conventional image processing methods.

[0006] In particular, since existing point cloud pattern reading technologies primarily rely on two-dimensional (2D) image-based analysis techniques, complex distortions such as curvature or wrinkles on the subject's surface deform the point cloud pattern, leading to a decrease in reading rates. While general correction techniques can compensate for simple distortions using perspective correction or geometric transformations, accurate correction is difficult when complex curves or non-linear creases are present.

[0007] Furthermore, the issue of illegal replication and counterfeiting of military and security clothing has recently become serious. Military uniforms are not merely clothing but a crucial element representing a soldier's identity, and the circulation of counterfeit uniforms can directly lead to security threats. For instance, there may be instances of unauthorized entry or attempted illegal activities using counterfeit uniforms, which could pose a significant threat to national security and public safety.

[0008] Existing anti-counterfeiting methods for military uniforms include technologies such as the use of special fabrics, the insertion of holographic tags, and the embedding of RFID chips; however, these methods suffer from high manufacturing costs or the difficulty of fundamentally preventing replication itself. In particular, when counterfeiters imitate genuine products by using similar patterns and materials, it is difficult to determine authenticity using only existing visual or physical security elements. The problem to be solved

[0010] The present invention aims to provide a method for reading a clothing point cloud pattern and an apparatus for executing the same, which enables accurate reading of a point cloud pattern regardless of the wearer's movement or wrinkles by applying a technology that scans the clothing, analyzes depth information of the clothing surface, and flattens it to restore the original point cloud pattern.

[0011] In addition, the present invention aims to provide a clothing point cloud pattern reading method and an apparatus for executing the same, which can extract and read point cloud patterns with high accuracy by correcting distortion of the point cloud pattern caused by wrinkles in clothing or movement of the wearer, and can correct non-linear distortion by applying a deep learning-based depth estimation and flattening algorithm.

[0012] Furthermore, the present invention aims to provide a method for reading clothing point cloud patterns and an apparatus for executing the same, which enables reliable determination of authenticity through point cloud pattern detection since it is difficult for counterfeit military uniforms to perfectly reproduce the same point cloud pattern, thereby facilitating the identification of soldiers and security personnel and allowing for the expected effect of preventing security incidents involving counterfeit military uniforms.

[0014] The objects of the present invention are not limited to those mentioned above, and other unmentioned objects and advantages of the present invention may be understood from the following description and will be more clearly understood by the embodiments of the present invention. Furthermore, it will be readily apparent that the objects and advantages of the present invention can be realized by the means and combinations thereof set forth in the claims. means of solving the problem

[0016] A clothing point cloud pattern reading method executed in a clothing point cloud pattern reading device for achieving such purposes includes the steps of: capturing a clothing surface to generate a two-dimensional original image; applying a projection transformation matrix to the two-dimensional original image to correct perspective distortion; detecting wrinkle distortion in the corrected image and applying non-linear filtering to correct wrinkle distortion; analyzing the curvature information of the clothing to perform a curvature-based flattening operation and then detecting a point cloud pattern; and providing a service connected to the point cloud pattern.

[0017] In one embodiment, the step of correcting the perspective distortion can inversely transform the perspective deformation using the camera intrinsic parameter matrix and the extrinsic parameter matrix.

[0018] In one embodiment, the step of correcting the wrinkle distortion may include applying a Fourier Transform to detect the region containing the wrinkles and performing high-frequency filtering to remove the wrinkle components.

[0019] In one embodiment, the step of detecting a point cloud pattern after performing the curvature-based flattening operation may include: a step of applying a Gaussian curvature formula to calculate the surface curvature value; and a step of correcting the point cloud deformation value based on a curvature tensor to perform the curvature correction. Effects of the invention

[0021] According to the present invention as described above, by applying a technology that scans clothing, analyzes depth information of the clothing surface, and flattens it to restore the original point cloud pattern, there is an advantage in that the point cloud pattern can be accurately read regardless of the wearer's movement or wrinkles.

[0022] In addition, according to the present invention, the distortion of the point cloud pattern caused by wrinkles in clothing or movement of the wearer can be corrected to extract and read the point cloud pattern with high accuracy, and there is an advantage that non-linear distortion can be corrected by applying a deep learning-based depth estimation and flattening algorithm.

[0023] In addition, according to the present invention, since it is difficult for counterfeit military uniforms to perfectly reproduce the same point cloud pattern, the authenticity can be reliably determined through point cloud pattern detection, thereby facilitating the identification of soldiers and security personnel and having the effect of preventing security incidents involving counterfeit military uniforms. Brief explanation of the drawing

[0025] FIG. 1 is a network configuration diagram for explaining a clothing point cloud pattern reading system according to one embodiment of the present invention. FIG. 2 is a flowchart illustrating an embodiment of a clothing point cloud pattern reading method according to the present invention. Specific details for implementing the invention

[0026] The aforementioned objectives, signatures, and advantages are described in detail below with reference to the attached drawings, thereby enabling those skilled in the art to easily implement the technical concept of the present invention. In describing the present invention, detailed descriptions of known technologies related to the present invention are omitted if it is determined that such descriptions would unnecessarily obscure the essence of the invention. Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the attached drawings. In the drawings, the same reference numerals are used to indicate the same or similar components.

[0028] FIG. 1 is a network configuration diagram for explaining a clothing point cloud pattern reading system according to one embodiment of the present invention.

[0029] Referring to FIG. 1, the clothing point cloud pattern reading system includes a clothing point cloud pattern reading device (100) and a clothing manufacturing device (200).

[0030] The clothing point cloud pattern reading device (100) is a device that photographs a user wearing clothing with a point cloud pattern inserted therein and extracts the point cloud pattern from the clothing.

[0031] First, the clothing point cloud pattern reading device (100) is equipped with a smartphone camera, LiDAR sensor, or ToF (Time-of-Flight) sensor capable of high-resolution shooting. Accordingly, when the user is wearing the clothing, the clothing point cloud pattern reading device captures the surface of the clothing, and although wrinkles or creases may occur at the time of capturing the surface of the clothing, a flattening processing technique is applied.

[0032] More specifically, the clothing point cloud pattern reading device (100) photographs the surface of the clothing while the user is wearing the clothing. At this time, the image is captured as a distorted 2D image. When the captured image is denoted as I(x, y) as in [Equation 1], x and y represent image coordinates, and distortion occurs depending on the shape of the crumpled military uniform.

[0034] [Mathematical Formula 1]

[0035]

[0037] I(x, y): Captured image,

[0038] f(x, y, z): A projection function that projects a point (x, y, z) in 3D space on the surface of the clothing onto a 2D image plane (x,y), as shown in [Equation 2]

[0039] (x, y, z): 3D point cloud coordinates

[0041] [Mathematical Formula 2]

[0042]

[0044] f(x, y, z): A projection function that projects a point (x, y, z) in 3D space on the surface of clothing onto a 2D image plane (x,y),

[0045] f perspective (x, y, z): Perspective distortion correction function,

[0046] f wrinkle (x, y, z): Wrinkle correction function,

[0047] f curvature (x, y, z): curvature correction function,

[0049] f perspective(x, y, z) is a perspective distortion correction function that corrects perspective distortion during camera capture to ensure the pattern is maintained at a constant ratio. This perspective distortion is a spatial deformation that occurs when capturing images with a camera. Generally, objects appear larger the closer they are to the camera and smaller the further they are. In particular, when projecting 3D surfaces, such as clothing, onto a 2D image, distortion due to depth occurs, which can cause the point cloud pattern to deform. The reason this perspective distortion is necessary is that for the point cloud pattern to be read in its original form, distortion caused by distance (depth) must be eliminated through perspective correction. In this process, an inverse transformation is performed using parameters such as the camera's focal length, sensor size, and shooting distance to convert the projected coordinates back into the original 3D space.

[0050] First, the clothing point cloud pattern reading device (100) converts 3D coordinates into 2D coordinates based on a projection model of a camera based on a perspective distortion correction function, then models radial and tangential distortions and performs distortion correction using a non-linear optimization technique. After that, the clothing point cloud pattern reading device (100) restores the point cloud pattern based on the corrected 2D coordinates.

[0051] First, the function for converting 3D coordinates to 2D coordinates is as shown in [Equation 3].

[0053] [Mathematical Formula 3]

[0054]

[0056] (x, y): 2D image coordinates,

[0057] c x , c y : Center coordinates of the image sensor,

[0058] f x , f y : Camera focal length,

[0059] (X w , Y w , Z w): Coordinates in 3D space

[0061] In the above [Equation 3], the smaller the Z-axis (depth), the stronger the perspective distortion occurs.

[0062] After going through this process, the clothing point cloud pattern reading device (100) models radial distortion and tangential distortion. The camera lens contains radial distortion and tangential distortion. Radial distortion exhibits the effect of straight lines bending as they move away from the center of the image, and the models for correcting this are [Equation 4] and [Equation 5].

[0064] [Mathematical Formula 4]

[0065]

[0066] k1, k2, k3: radiation distortion coefficients,

[0067] r 2 = x 2 + y 2 : Squared radius value at pixel location

[0068] (x', y'): Distortion-corrected coordinates

[0070] Tangential distortion is distortion that occurs when the lens is not perfectly aligned with the camera sensor. The formula for correcting this is [Equation 5].

[0072] [Mathematical Formula 5]

[0073]

[0075] P1, P2: Tangent distortion coefficients,

[0076] r 2 = x 2 + y 2 : Squared radius value at pixel location

[0077] (x', y'): Distortion-corrected coordinates

[0079] Then, the clothing point cloud pattern reading device (100) performs distortion correction using a non-linear optimization technique. To correct the above perspective distortion, the opposite operation to remove the above radial distortion and tangential distortion must be performed. This process is performed using non-linear optimization, which is as shown in [Equation 6] below.

[0081] [Mathematical Formula 6]

[0082]

[0084] (x', y'): Distorted included pixel coordinates

[0085] (x corrected , y corrected ): Corrected image coordinates,

[0086] (c x , c y ): Normalize coordinates based on the center of the camera sensor

[0087] f x , f y : Corrected by camera focal length (scaled in pixels)

[0088] Z w : Reconstruct the position in 3D space using depth information

[0090] As mentioned above, since severe perspective distortion occurs when Zw is small, the original coordinates can be restored by performing an inverse transformation based on depth. That is, after inputting the 2D image coordinates (x', y') with perspective distortion applied, camera intrinsic parameters (focal length, sensor center), and depth information (Zw) into the non-linear optimization technique, the center point (c) from the pixel coordinates containing distortion x , c y Perform normalization based on ), and coordinate distance (f x , f y Scale using ), and depth (z w The final corrected coordinates are restored by multiplying by ) to restore the position in 3D space, and are as shown in [Equation 6].

[0091] The clothing point cloud pattern reading device (100) can restore the point cloud pattern to its original form using corrected coordinates. Through this process, the pattern can be read regardless of the wearer's movement or deformation of the clothing.

[0092] In addition, wrinkle distortion occurs when the surface of a garment deforms non-linearly. Garments printed with point cloud patterns can be deformed due to wearer movement or environmental factors (folding, creases, etc.), causing the point cloud pattern to appear different from its original state.

[0093] Accordingly, the clothing point cloud pattern reader (100) detects a wrinkled area through curvature calculation, analyzes high-frequency components using a Fourier transform, and then performs filtering to flatten the surface using curvature information and remove high-frequency components so that the point cloud pattern can be restored to its original form. Then, the clothing point cloud pattern reader (100) can reconstruct the corrected point cloud pattern by mapping it to its original position. Through the above process, distortion of the point cloud pattern caused by wrinkles and creases in the clothing can be corrected. This allows the point cloud pattern to be maintained in its original state and enables accurate reading.

[0094] In the above, a periodic nonlinear function is used to mathematically model wrinkles. Wrinkles in clothing can often be modeled as a sine function and are expressed as [Equation 7].

[0096] [Mathematical Formula 7]

[0097]

[0099] a: Amplitude of the wrinkle (intensity, height of the wrinkle)

[0100] w: period of the fold (interval of the fold),

[0101] Θ: Phase of the fold,

[0103] The above [Equation 7] indicates that the surface of the clothing is deformed non-linearly in a specific direction, and the value of a increases as the number of wrinkles increases, and the value of w increases as the number of wrinkles is narrow.

[0104] The clothing point cloud pattern reading device (100) detects wrinkles based on surface curvature. At this time, the surface curvature (K) changes rapidly in the area where wrinkles are present. Therefore, the clothing point cloud pattern reading device (100) can detect wrinkled areas by performing curvature calculations using second derivatives as in [Equation 8].

[0106] [Mathematical Formula 8]

[0107]

[0108] K: curvature,

[0109] : Second derivative in the X and Y directions (degree of curvature),

[0110] Cross partial derivatives

[0112] The clothing point cloud pattern reading device (100) can detect areas with wrinkles based on the fact that the curvature value increases rapidly in the parts where wrinkles occur.

[0113] Additionally, the clothing point cloud pattern reading device (100) can detect high-frequency components by applying a Fourier transform, since wrinkles contain high-frequency components. A Fourier transform is a process of converting a signal into the sum of multiple frequency components.

[0114] To this end, the clothing point cloud pattern reading device (100) analyzes wrinkle information by performing a two-dimensional Fourier transform on the input image I(x, y) as in [Equation 9]. The input image I(x, y) is expressed in the spatial domain (pixel coordinates), but when this transformation is performed, it is converted into a representation in the frequency domain called F(u, v).

[0116] [Mathematical Formula 9]

[0117]

[0119] I(x, y): Original image,

[0120] F(u, v): transformed frequency component,

[0121] M, N: Width and height of the image,

[0122] u, v: frequency coordinates,

[0123] : Complex exponential function,

[0124] j: imaginary unit,

[0126] In the above [Equation 9], the original image is converted into frequency components. To this end, the clothing point cloud pattern reader (100) decomposes all pixel values ​​of the image into frequency components, where low frequencies represent large structures (background, lighting, etc.) and high frequencies represent details (texture, edges, wrinkles, etc.). Therefore,

[0127] After that, the clothing point cloud pattern reader (100) detects a repetitive pattern by applying a Fourier transform. At this time, that is, by applying a Fourier transform, high-frequency components such as wrinkles can be detected by showing a specific pattern in the frequency domain. Therefore, the clothing point cloud pattern reader (100) can obtain a flattened image by controlling the wrinkles.

[0128] At this time, the clothing point cloud pattern reading device (100) can obtain a flattened image by controlling wrinkles through low-frequency filtering that removes high-frequency components.

[0129] Specifically, the clothing point cloud pattern reading device (100) can convert to the frequency domain by performing a Fourier transform, remove high-frequency components above a specific threshold, and convert to the original image by performing an inverse Fourier transform. Therefore, since wrinkles can be removed and a flattened image restored through the above process, the point cloud pattern of the clothing can be corrected so that it can be read without distortion.

[0130] In addition, clothing is not a completely flat surface and has the characteristic of bending or curving along the curves of the body. In this case, if curvature is present, the point cloud pattern exists in a deformed state in 3D space, and since the original shape may be distorted when converted into a 2D image, curvature correction must be performed.

[0131] A representative method for modeling such curvature is the curvature function used in differential geometry, as shown in [Equation 10].

[0133] [Mathematical Formula 10]

[0134]

[0136] K: curvature,

[0137] f(X, Y): A surface height function in 2D coordinates, which is the height function of the surface at a given point,

[0138] : Second partial derivative in the x, y directions (indicates how curved the surface is),

[0139] : Cross partial derivative,

[0141] Curvature is a value that mathematically expresses the bending (degree of deformation) of a surface, where the curvature value is large in areas with wrinkles and small in flat areas. If the curvature (K) is large, the surface is severely bent, and if the curvature (K) is 0, it means a flat surface. Therefore, the clothing point cloud pattern reading device (100) can obtain a new flattened surface by applying a correction value in the direction of flattening the curvature.

[0142] That is, the clothing point cloud pattern reading device (100) can calculate the curvature, then apply a correction formula as in [Equation 11] to create a new flattened surface, then calculate a new curvature, and if the new curvature is below a threshold value, repeat the above process.

[0144] [Mathematical Formula 11]

[0145]

[0147] f(x, y): original surface

[0148] K(x, y): curvature value (degree of bending),

[0149] λ: Adjustment factor (flattening strength),

[0151] As described above, the clothing point cloud pattern reading device (100) performs curvature correction and then reapplies the corrected coordinates to the original point cloud pattern to restore the flattened point cloud. As described above, distortion caused by wrinkles and curvature on the surface of the clothing can be removed, thereby correcting the pattern to enable accurate pattern reading.

[0152] After that, the clothing point cloud pattern reading device (100) performs a transformation to restore the transformed point cloud data to its original state in order to detect the point cloud pattern after applying a correction process.

[0153] The projection function of [Equation 2] above is a mathematical expression of how the point cloud is distorted and is not a function that performs correction. Therefore, to inversely transform the actual distortion, the clothing point cloud pattern reading device (100) performs perspective distortion correction, wrinkle distortion correction, and curvature distortion correction by applying the inverse matrix T-1 of the transformation matrix as in [Equation 12]. At this time, the correction process is as in [Equation 4], [Equation 5], [Equation 6], [Equation 9], and [Equation 11].

[0155] [Mathematical Formula 12]

[0156]

[0158] T -1 : The inverse matrix that restores the point group to its original form,

[0159] P': Original point cloud pattern

[0160] P: Modified point cloud pattern

[0162] The clothing manufacturing device (200) is a device that manufactures clothing (300) by generating a final pattern image using a point cloud pattern image that connects a basic pattern image for clothing (300) and authenticity information of clothing (300), and then printing the final pattern image onto a fiber.

[0163] First, the clothing manufacturing device (200) generates a data code by representing data connecting the authenticity information of the clothing (300) as binary bits, generates a verification code for the data code, and generates a data code with the verification code inserted using the verification code for the data code.

[0164] After that, the clothing manufacturing device (200) generates a point cloud pattern by displaying a data code with a verification code inserted on a point cloud representation area. At this time, the point cloud representation area may be formed with a width and height between, for example, 32 bits and 35 bits. However, this can be modified and applied by a person skilled in the art to suit the purpose or environment of implementation.

[0165] In one embodiment, the clothing manufacturing device (200) generates a point cloud pattern by representing 0 or 1 on a point cloud representation area according to the binary bits of the data code into which a verification code is inserted.

[0166] The clothing manufacturing device (200) generates a fast Fourier transform point cloud pattern image by performing a fast Fourier transform on a point cloud representation area and a position determination pattern. At this time, the position determination pattern is used to determine the starting position of the point cloud on the point cloud pattern image when the clothing (300) is scanned and the point cloud pattern is decoded after the point cloud pattern image is printed on the clothing (300).

[0167] After that, the clothing manufacturing device (200) generates a final point cloud pattern image using a plurality of point cloud pattern images. Each point cloud pattern image constituting the point cloud pattern is fast Fourier transformed together with a position determination pattern.

[0168] The clothing manufacturing device (200) generates a final pattern image using a basic pattern image and a point cloud pattern image.

[0169] In one embodiment, the clothing manufacturing device (200) includes the step of placing the point cloud pattern image on top of a basic pattern image to form a layer structure, and adjusting the color step difference between the basic pattern image and the point cloud pattern image to generate a final pattern image.

[0170] In the above embodiment, the clothing manufacturing device (200) can generate a final pattern image by increasing or decreasing the pixels of a point cloud pattern image by a certain amount based on the pixel color value of a basic pattern image.

[0171] For example, the clothing manufacturing device (200) can make the point group pattern slightly darker than the existing pattern by applying -10 to each value when the basic pattern pixel value is (R, G, B) = (120, 130, 140) and the point group pattern is negative (i.e., 0).

[0172] As another example, a clothing manufacturing device (200) can make the point group pattern appear slightly brighter than the original pattern by applying +10 to each value when the basic pattern pixel value is (R, G, B) = (120, 130, 140) and the point group pattern is positive (i.e., 1), thereby adjusting it to (130, 140, 150).

[0173] To this end, the clothing manufacturing device (200) can convert the RGB values ​​of the basic pattern image and the point cloud pattern image into the Lab color space and then calculate the color difference to check whether it is located within a specific range.

[0174] In one embodiment, the clothing manufacturing device (200) can calculate the color difference after converting the RGB values ​​of the basic pattern image and the point cloud pattern image into the Lab color space as in [Equation 1].

[0175] In this context, the Lab color space is a color representation method based on how the human eye perceives color. The Lab color space consists of three axes: L, a, and b, where L represents brightness, a represents the green-red axis, and b represents the blue-yellow axis.

[0177] [Mathematical Formula 13]

[0178]

[0180] △E: Color difference,

[0181] L: Brightness,

[0182] a: Red-green value,

[0184] As described above, the reason for converting the RGB values ​​of the basic pattern image and the point cloud pattern image to the Lab color space is that in RGB, (255, 0, 0) and (254, 0, 0) appear to be almost the same red to the naked eye, but in Lab, the color difference (ΔE) is slightly different and detectable.

[0185] The larger the color difference value mentioned above, the more the color is perceived differently. Color difference (ΔE) can be set to 76, 94, or 2000, but general smartphone cameras have high recognition accuracy based on a ΔE of 2000. Therefore, when ΔE is 5 or less, the difference is barely detectable by both the naked eye and the camera; when ΔE is 10 to 15, it is invisible to the naked eye but detectable by the camera; and when ΔE is 20, it can be distinguished even with the naked eye.

[0186] [Table 1] below explains the process for setting the optimal color step difference.

[0188] [Table 1]

[0189]

[0191] For example, if a specific pixel of the basic pattern image is Lab(70, 0, 0), the corresponding pixel of the point cloud pattern image is adjusted to Lab(75, 5, -5), there is no significant difference to the naked eye, but the clothing point cloud pattern reader (100) can distinguish it by setting the color difference (ΔE) value to 12.

[0192] As described above, by utilizing the Lab color space in this way, color differences that are invisible to the naked eye but detectable by a camera can be accurately controlled.

[0193] The clothing manufacturing device (200) manufactures the clothing (300) by printing a final pattern image onto a fiber.

[0194] A point cloud pattern may be printed on the garment (300) produced as described above, but it may be visually indistinguishable. In this case, each of the intaglio and relief may indicate 0 or 1. That is, it may be predefined that the relief indicates 0 and the intaglio indicates 1, or it may be predefined that the relief indicates 1 and the intaglio indicates 0.

[0195] The above point cloud pattern recognizes a code value that is reflected and displayed through a flash or camera light provided to a clothing point cloud pattern reader (100), and then links the authenticity information of the clothing (300) corresponding to the code value. The code value can be various information utilizing the characteristics of light.

[0197] FIG. 2 is a flowchart illustrating an embodiment of a clothing point cloud pattern reading method according to the present invention.

[0198] Referring to FIG. 2, a clothing point cloud pattern reading device (100) captures the surface of the clothing to generate a two-dimensional original image (step S210).

[0199] The clothing point cloud pattern reading device (100) corrects perspective distortion by applying a projection transformation matrix to a 2D original image (step S220).

[0200] The clothing point cloud pattern reading device (100) detects wrinkle distortion in the corrected image and corrects the wrinkle distortion by applying non-linear filtering (step S230).

[0201] The clothing point cloud pattern reading device (100) analyzes the curvature information of the clothing, performs a curvature-based flattening operation, and then detects a point cloud pattern (step S240).

[0202] The clothing point cloud pattern reading device (100) provides a service connected to the point cloud pattern (step S250). At this time, the types of services may include, but are not limited to, a service for determining the authenticity of clothing, a service for providing information about a user wearing clothing, etc.

[0204] Although the present invention has been described by the embodiments and drawings described above, the present invention is not limited to the above embodiments, and various modifications and variations are possible from this description by those skilled in the art to which the present invention pertains. Accordingly, the concept of the present invention should be understood only by the claims set forth below, and all equivalent or analogous variations thereof shall be considered to fall within the scope of the concept of the present invention.

Claims

Claim 1 A method for reading a point cloud pattern of clothing executed in a clothing point cloud pattern reading device, comprising: a step of capturing a surface of clothing to generate a two-dimensional original image; a step of applying a projection transformation matrix to the two-dimensional original image to correct perspective distortion; a step of detecting wrinkle distortion in the corrected image and applying non-linear filtering to correct wrinkle distortion; a step of analyzing curvature information of the clothing to perform a curvature-based flattening operation and then detecting a point cloud pattern; and a step of providing a service connected to the point cloud pattern. Claim 2 A method for reading a clothing point cloud pattern, wherein, in claim 1, the step of correcting the perspective distortion is characterized by inversely transforming the perspective deformation using a camera intrinsic parameter matrix and an extrinsic parameter matrix. Claim 3 A method for reading a point cloud pattern of clothing, characterized in that, in claim 1, the step of correcting the wrinkle distortion comprises: a step of applying a Fourier Transform to detect an area containing the wrinkles; and a step of performing high-frequency filtering to remove wrinkle components. Claim 4 A method for reading a point cloud pattern of clothing according to claim 1, wherein the step of detecting a point cloud pattern after performing the curvature-based flattening operation comprises: a step of applying a Gaussian curvature formula to calculate the surface curvature value; and a step of correcting the point cloud deformation value based on a curvature tensor to perform the curvature correction.

Images