KR102575508B1 - AI-based textile pattern inspection system for article of footwear - Google Patents

Abstract

The present invention relates to an AI-based fabric pattern inspection system for shoe parts, and more particularly, to an AI-based fabric pattern inspection system for shoe parts that inspects patterns of fabrics for shoe parts through computer vision and machine learning. In the AI-based textile pattern inspection system for shoe parts that collects and images images and determines defect patterns, a scanning unit that images a pattern of fabric through computer vision, an image collection unit that collects images generated by the scanning unit, and , It is configured to include an AI inspection unit that receives and analyzes the images collected by the image collection unit to determine a defective pattern, and a database unit that provides data based on the judgment to the AI inspection unit.

Description

AI-based textile pattern inspection system for article of footwear}

The present invention relates to an AI-based fabric pattern inspection system for shoe parts, and more particularly, to an AI-based fabric pattern inspection system for shoe parts that inspects a pattern of fabric for shoe parts through computer vision and machine learning.

In recent years, functional shoes have been spotlighted as high value-added products with improved physical properties by combining various materials and manufacturing technologies.

The existing footwear industry has grown based on price competitiveness centered on low-cost materials and low-cost processes, but recently, sales of high-priced functional products have been brisk due to the revitalization of the sports industry. In other words, lightweight yet high-quality, high-functional materials are being applied to shoes, and new production methods such as 3D printing are being applied to production methods. Accordingly, the existing woven and non-woven fabric processing technologies used in shoes are also required to be improved for more precise and better quality.

In addition, global brands leading the footwear industry are actively inducing a smart factory business that builds an intelligent factory using speed factories and sensors that drastically shorten delivery times, making process automation inevitable.

However, as the existing engineers in the footwear field are aging, the issue of technology transfer is emerging, and the level of product inspection varies depending on the proficiency of the operator, and as the shape of shoes diversifies, the design and color of shoe parts become complicated. As it deteriorates, it is difficult to perform quality inspection perfectly with the existing manual work.

Accordingly, there is an increasing demand for the introduction of a fabric inspection system suitable for an automated shoe manufacturing process.

Prior Art Korean Patent Registration No. 10-1873278 relates to an apparatus and method for inspecting uppers of shoes, wherein a plurality of support rollers disposed in parallel and rotated, an object to be inspected is moved in one direction by the support rollers, An inspection chamber installed on a moving path of the inspected object, a gantry robot installed inside the inspection chamber, an imaging member installed in the gantry robot, and a light installed inside the inspection chamber parallel to the ground and having light transmittance A transmissive member, a lower light emitting member spaced apart from the lower side of the light transmissive member, and an upper light emitting member spaced apart from the upper side of the light transmissive member in the test chamber and illuminating downward toward the light transmissive member. Although an apparatus for inspecting the upper of a shoe is disclosed, it is possible to inspect the size defect of the shoe upper, but there is a problem in that it is difficult to identify defects in the fabric pattern as described above.

Republic of Korea Patent No. 10-1873278

An object of the present invention to solve the above problems is to provide an AI-based fabric pattern inspection system for shoe parts that can image and collect fabric patterns through computer vision technology in inspecting shoe fabrics. there is.

Another object of the present invention is to provide an AI-based fabric pattern inspection system for shoe parts that can identify defective fabrics at high speed through AI learning in inspecting shoe fabrics.

An AI-based fabric pattern inspection system for shoe parts according to an embodiment of the present invention for solving the above problems is an AI-based fabric pattern inspection system for shoe parts that collects images of fabric patterns and determines defect patterns. , a scanning unit that images the pattern of fabric through computer vision, an image collection unit that collects images generated by the scanning unit, and an AI inspection unit that receives and analyzes the images collected by the image collection unit to determine defective patterns, and It is configured to include a database unit that provides data on the basis of judgment to the AI inspection unit.

Preferably, the AI check unit may be configured to include additional accumulation of data based on judgment by feeding back the judgment result data to the database unit.

Preferably, the AI inspection unit comprises a pattern division module for dividing pattern parts in an image, a coordinate division module for dividing the pattern divided by the pattern division module into a grid by dividing into a preset orthogonal coordinate system, and the coordinate division module dividing It may include a pattern inspection module that inspects the pattern of each grid and an AI learning module that learns and categorizes the coordinate position and defect type of the defective pattern inspected by the pattern inspection module.

Preferably, the AI inspection unit is further provided with a character recognition module for recognizing a character string appearing on the fabric, and the AI learning module automatically converts and learns the character string on the other side when recognizing a character string on either side of the left or right of the shoe. It can be configured including.

The AI-based fabric pattern inspection system for shoe parts according to the present invention is a computer for inspecting fabrics for shoes through a scanning unit that images fabric patterns through computer vision and an image collection unit that collects images generated by the scanning unit. Through vision technology, there is an effect of collecting images of textile patterns.

The AI-based fabric pattern inspection system for shoe parts according to the present invention is AI learning through an AI inspection unit that receives and analyzes images collected by an image collection unit to determine defective patterns and a database unit that provides data based on the judgment to the AI inspection unit. Based on this, there is an effect of providing a fabric pattern inspection system capable of identifying defective fabrics at high speed.

1 is a schematic diagram showing a configuration according to an embodiment of an AI-based textile pattern inspection system for shoe parts according to the present invention.
2 shows the configuration of an AI inspection unit according to an embodiment of the AI-based textile pattern inspection system for shoe parts according to the present invention.
3 is a flow chart showing a fabric pattern inspection process according to an embodiment of the AI-based fabric pattern inspection system for shoe parts according to the present invention.
4 is an embodiment showing a flowchart of an image output process after a scanner operation of the AI-based fabric pattern inspection system for shoe parts according to the present invention.
5 is a flowchart illustrating a process of determining defects in a fabric pattern of the AI-based fabric pattern inspection system for shoe parts according to another embodiment of the present invention.

The specific details, including the problems to be solved, the solutions to the problems, and the effect of the invention for the present invention as described above are included in the embodiments and drawings to be described below. Advantages and features of the present invention, and methods of achieving them, will become clear with reference to the detailed description of the following embodiments taken in conjunction with the accompanying drawings.

As shown in FIGS. 1 to 5, the AI-based fabric pattern inspection system for shoe parts according to the present invention collects images of fabric patterns and determines defect patterns. In the present invention, a scanning unit (1) for imaging a pattern of fabric through computer vision, an image collection unit (2) for collecting images generated by the scanning unit (1), and the image collection unit (2) collected It may be configured to include an AI inspection unit 3 that receives and analyzes images to determine defective patterns, and a database unit 4 that provides data based on judgment to the AI inspection unit 3.

1 is a schematic diagram showing a configuration according to an embodiment of an AI-based textile pattern inspection system for shoe parts according to the present invention.

First, the scan unit 1 may image a fabric pattern through computer vision.

Specifically, the computer vision of the scan unit 1 may be provided with a scanner or camera, and any equipment capable of generating a digital image by scanning or photographing may be provided, including single or heterogeneous types. A plurality may be provided.

More specifically, the scan unit 1 is preferably provided on a fabric movement path during a process of manufacturing a fabric for shoe parts.

Next, the image collection unit 2 may collect images generated by the scanning unit 1 .

Specifically, the image collection unit can serve to receive the image generated by the scanning unit 1 and deliver it to the AI inspection unit 3 so that the AI inspection unit 3 can determine the defect of the fabric. and can optionally store images in chronological order.

Next, the AI inspection unit 3 may determine a defective pattern by receiving and analyzing the images collected by the image collection unit 2 .

Specifically, as shown in FIG. 2, the AI inspection unit 3 includes a pattern classification module 31, a coordinate division module 32, a pattern inspection module 33, and an AI learning module 34. It can be.

The pattern division module 31 may classify and extract the pattern part in the image.

In addition, the coordinate division module 32 may divide the pattern extracted by the pattern division module 31 into a preset orthogonal coordinate system to form a grid.

In addition, the pattern inspection module 33 can inspect the pattern of each lattice divided by the coordinate division module 32 .

In addition, the AI learning module 34 may learn and categorize the coordinate location and defect type of the defective pattern inspected by the pattern inspection module 33 .

On the other hand, the AI inspection unit 3 may be configured to further include a character recognition module 35.

The character recognition module 35 may recognize a character string appearing on fabric.

In addition, the AI learning module 34 can automatically convert and learn the string on the opposite side when recognizing one of the left and right strings of the shoe.

Next, the database unit 4 may provide data based on judgment to the AI inspection unit 3 .

In addition, the AI inspection unit 3 feeds back the judgment result data to the database unit 4 so that data based on judgment can be additionally accumulated.

As shown in FIG. 3 and described above, an application example of the AI-based textile pattern inspection system for shoe parts according to the present invention is as follows.

It is preferable that the data of the database unit 4 include images and fabric patterns that are referenced in the determination of the AI inspection unit 3.

In addition, in the database unit 4, not only characters for recognizing a character string of a corresponding shoe product, but also information on the manufacturer, product name, target gender, shoe size, and left and right are stored in a database.

In addition, the AI learning unit may perform learning for determining defective products and good products based on data provided by the database unit 4 .

In addition, the data learned in this way can be transmitted to the pattern division module 31 and used to classify and discriminate an accurate pattern.

First, when fabric is transported, rolled into a roll, or calendered in a fabric process applied to shoe manufacturing, the scanning unit 1 scans or photographs the surface of the fabric in motion to form an image.

Scanned images are collected through the image collection unit 2 and stored in real time. Stored images may be provided by a manager by time or by lot.

Next, the AI inspection unit 3 analyzes and determines defects on the surface of the fabric by analyzing the image transmitted by the image collection unit 2 in real time.

Specifically, the pattern division module 31 may play a role in identifying and extracting objects to be analyzed, such as patterns, characters, and symbols, in an image. That is, it is possible to discriminate a defect by extracting a part necessary for discriminating a pattern with a focus.

Preferably, the pattern division module 31 may distinguish between a part actually used and a part not used in the image, and may extract a part to be inspected with a difference in a weight value.

The coordinate division module 32 divides the part to be inspected extracted by the pattern division module 31 into a Cartesian coordinate system according to a preset standard and performs latticeization.

Specifically, the minimum unit to be gridded may be a pixel unit, and may be selected as a size suggested by the AI learning module 34 to classify defects. That is, as the repetitive pattern analysis proceeds, the AI learning module 34 may suggest a grid size and division number optimized for finding fabric defects.

Next, the pattern inspection module 33 compares the non-defective product reference provided by the database unit 4 with the inspected unit to determine whether the product is defective or non-defective.

Specifically, the good product reference provided by the database unit 4 is divided into the same size as the lattice size divided by the coordinate division module 32 to make a one-to-one correspondence with each coordinate system of the inspected unit to discriminate defects such as color or pattern. it is desirable

As an example, the result of the pattern inspection module 33 may be expressed as a percentage (%) according to a preset standard. That is, the pattern inspection module 33 may provide the defect sector as an image and display the defect level of the corresponding portion as a %. This can be utilized to set weights for making a final defect decision.

In addition, the AI learning module 34 may learn and categorize the coordinate location and defect type of the defective pattern inspected by the pattern inspection module 33 . The learned information can be transmitted to the database unit 4 and used when inspecting defects of the next fabric pattern.

In addition, the character recognition module 35 recognizes the character string appearing on the fabric, and the separated character string information is transmitted to the pattern classification module 31 and can be inspected through the same process as when inspecting a general pattern. .

In addition, the AI learning module 34 learns information about character strings stored in the database unit 4, manufacturer, product name, target gender, shoe size, shoe left and right classification, logo, and the like. In addition, the pattern inspection module 33 does not simply process defects when performing inspection, but checks the corresponding data such as character string, manufacturer, product name, target gender, shoe size, shoe left and right classification, logo, etc. in the database to determine whether the product is good or defective. to determine whether

As an example, if a fabric pattern applied to the right shoe is input after repeatedly inspecting the fabric pattern applied to the left shoe, the administrator is guided, and according to the response, it is recognized as a normal pattern and the inspection continues. can

In one embodiment, if a change is found in the size of the fabric pattern according to the size, a step of informing the manager, and according to the response, it may be recognized as a normal pattern and the inspection may be continued.

As an example, if the direction of the character string or logo is in the opposite direction, a step of providing guidance to the manager may be performed, and it may be recognized as a normal pattern according to a response and the inspection may be continued.

That is, by repeatedly inspecting the fabric pattern for one shoe, the fabric pattern for the opposite shoe can be learned and determined in advance. In addition, according to the change in pattern size, it is possible to confirm that it is the shoe size of another number and continue the inspection work.

In addition, based on the manufacturing result, the selection weight of the analysis target part classified by the pattern classification module 31 and the defect determination weight determined by the pattern judgment module 33 may be different, and each through the AI learning unit 34 Discrimination weights can be actively updated.

As one embodiment, as shown in FIG. 4 , a process from scanning a fabric pattern to image collection may be performed in the following flow.

Start → Scanner ON → Test → Continuous scan → Scanned image output → Stop scanning → End

As an embodiment, as shown in FIG. 5 , fabric pattern analysis may be performed in the following flow.

Start → Select 3 columns from the image analysis menu → Output scanned images in 3 columns (image merging using the Python opencv module) → Select 5 columns from the image analysis menu → Output 5 fabric patterns of the scanned image (using the Yolov5 model) Image cropping through R, L classification and fabric boundary detection) → Start of analysis → Determination of defects (Inspection using Tensorflow Keres CNN. Inspection is performed by cropping with 64X64 images, and output as one image upon completion of inspection) → Actual use Determining whether it is a part that is a problem (distinguish between actual use part and external part by making a difference with a weight value through product masking) → Notifying the location of a defect and emphasizing the location through RGB value change → Based on the manufacturing result, the defect detection weight is actively determined update to → quit

As such, it will be understood that the technical configuration of the present invention described above can be implemented in other specific forms without changing the technical spirit or essential features of the present invention by those skilled in the art to which the present invention pertains.

Therefore, the embodiments described above should be understood as illustrative and not restrictive in all respects, and the scope of the present invention is indicated by the claims to be described later rather than the detailed description, and the meaning and scope of the claims and their All changes or modified forms derived from equivalent concepts should be construed as being included in the scope of the present invention.

1: scan unit
2: Image collection unit
3: AI inspection department
31: pattern division module
32: coordinate division module
33: pattern inspection module
34: AI learning module
35: character recognition module
4: database part

Claims (4)

In an AI-based fabric pattern inspection system for shoe parts that collects and images patterns of fabrics and determines defect patterns,
A scan unit that images the fabric pattern through computer vision;
an image collection unit for collecting images generated by the scanning unit;
an AI inspection unit that receives and analyzes the images collected by the image collection unit to determine a defective pattern, and feeds the judgment result data back to the database unit so that data based on the judgment is additionally accumulated; and
Including; database unit for providing data on the basis of judgment to the AI inspection unit,
The AI inspection unit, a pattern division module for dividing and extracting the pattern portion in the image; a coordinate division module that divides the pattern extracted by the pattern division module into a preset orthogonal coordinate system and converts the pattern into a lattice; a pattern inspection module for inspecting the pattern of each lattice divided by the coordinate division module; an AI learning module that learns and categorizes the coordinate location and defect type of the defective pattern inspected by the pattern inspection module; And a character recognition module for recognizing a character string appearing on the fabric; wherein the AI learning module automatically converts and learns the character string on the opposite side when recognizing a character string on either side of the left and right shoes of the shoe. For fabric pattern inspection system. delete delete delete