KR102486230B1 - Method and system for quality inspection of new product using deep learning - Google Patents

Abstract

A new product quality inspection method and system using deep learning are proposed, and the new product quality inspection method and system using deep learning perform quality inspection of the new product based on a defective image, which is an image of an existing product with defects. It may include a learning data generation system that generates learning data necessary for learning for learning, and a defect identification system that learns the learning data and performs a quality inspection of the new product based on the learning.

Description

Method and system for quality inspection of new products using deep learning {METHOD AND SYSTEM FOR QUALITY INSPECTION OF NEW PRODUCT USING DEEP LEARNING}

Embodiments disclosed herein relate to a method and system for inspecting the quality of a new product using deep learning, and more particularly, to a product production line using a pre-learned model to inspect defects of products produced on the production line. It relates to a method and system for inspecting defects of new products newly produced in

The application of machine learning is increasing with the development of computing technology. In particular, in recent years, deep learning technology represented by artificial neural networks among machine learning has been rapidly developed, and application cases are increasing in various industrial fields. Even in the manufacturing sector, artificial neural networks are innovatively changing manufacturing sites.

Existing rule-based test methods have not been able to detect atypical defects that occur on the production line, but artificial neural networks can extract atypical defects like humans. In the past, people directly judged defects on the product production line for atypical defects, but now, machine vision using artificial neural networks is used to determine whether or not a product is defective.

In this way, a learning process of the artificial neural network is required to identify whether or not a product is defective by using the artificial neural network. That is, it is common to acquire images of defective products that may occur in the production process, and to train an artificial neural network using a supervised learning method using the acquired images of defective products.

In order to detect product defects using artificial neural networks for newly produced products, artificial neural networks must be trained using defective images of new products. There is a problem that it takes a long time to learn an artificial neural network because it is difficult to acquire an image in a bad state.

In this regard, Korean Patent Publication No. 10-2000-0087346, which is a prior art document, relates to an Internet artificial intelligence learning and management method. Teachers and learners are registered through the Internet, and teachers can create and use problem questions and test questions that they want to use. Although it is doing, it is not able to perform the learning of artificial intelligence quickly.

Therefore, a technique for solving the above problems has been required.

On the other hand, the above-mentioned background art is technical information that the inventor possessed for derivation of the present invention or acquired in the process of derivation of the present invention, and cannot necessarily be said to be known art disclosed to the general public prior to filing the present invention. .

Embodiments disclosed herein are aimed at presenting a method and system for quality inspection of a new product using deep learning.

Embodiments disclosed in this specification are aimed at presenting a method and system for inspecting quality of a new product that filters defects by learning an image of a good product of the new product.

Embodiments disclosed herein are aimed at presenting a quality inspection method and system for a new product that generates a defect image of a new product based on a defect image generated based on a defect image of an existing product and a good image of a new product. there is

Embodiments disclosed in this specification are aimed at presenting a new product quality inspection method and system for performing patch learning based on a defect image of an existing product and a defect image of a new product.

Embodiments disclosed in this specification are aimed at presenting a new product quality inspection method and system for filtering defects among new products and classifying defect types of the filtered defective products.

Embodiments disclosed in this specification are aimed at presenting a method and system for inspecting the quality of a new product in which a classification model is trained using defective images.

As a technical means for achieving the above-described technical problem, according to an embodiment, in a system for inspecting the quality of a new product, quality inspection of the new product is performed based on a defective image, which is an image of an existing product with defects. It may include a learning data generation system that generates learning data necessary for learning for learning, and a defect identification system that learns the learning data and performs a quality inspection of the new product based on the learning.

According to another embodiment, in the method for inspecting the quality of a new product by a quality inspection system, learning data necessary for learning for quality inspection of the new product is generated based on a defective image, which is an image of an existing product with defects. It may include the step of performing a quality test of the new product based on the step of learning, the step of learning the learning data, and the learning.

According to another embodiment, a computer readable recording medium on which a program for performing a quality inspection method is recorded, wherein the quality inspection method performs a quality inspection of the new product based on a defective image, which is an image of an existing product with defects. It may include generating learning data necessary for learning for learning, learning the learning data, and performing quality inspection of the new product based on the learning.

And according to another embodiment, the quality inspection method is performed by the quality inspection system and is a computer program stored in a medium to perform the quality inspection method, the new product based on a defective image, which is an image of an existing product with defects. It may include generating learning data necessary for learning for quality inspection, learning the learning data, and performing quality inspection of the new product based on the learning.

According to any one of the above-described problem solving means, it is possible to propose a quality inspection method and system for a new product using deep learning.

According to any one of the above-described problem solving means, it is possible to propose a new product quality inspection method and system capable of quickly filtering out defects of a new product by learning good quality images of the new product.

According to any one of the above-described problem solving means, a new product quality inspection method and system for generating a defective image of a new product based on a defective image generated based on a defective image of an existing product and a good image of a new product are proposed. can do.

According to any one of the above-described problem solving means, a new product quality inspection method and system for strengthening defect learning by performing patch learning based on a defect image of an existing product and a defect image of a new product are proposed. can do.

According to any one of the above-described problem solving means, it is possible to propose a new product quality inspection method and system for filtering defects among new products and classifying defect types of the filtered defective products.

Effects obtainable from the disclosed embodiments are not limited to those mentioned above, and other effects not mentioned are clear to those skilled in the art from the description below to which the disclosed embodiments belong. will be understandable.

1 is a configuration diagram showing a quality inspection system according to an embodiment.
2 is a block diagram showing the configuration of a quality inspection system according to an embodiment.
3 to 5 are flowcharts for explaining a quality inspection method according to an exemplary embodiment.

Hereinafter, various embodiments will be described in detail with reference to the accompanying drawings. Embodiments described below may be modified and implemented in various different forms. In order to more clearly describe the characteristics of the embodiments, detailed descriptions of matters widely known to those skilled in the art to which the following embodiments belong are omitted. And, in the drawings, parts irrelevant to the description of the embodiments are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout the specification, when a component is said to be "connected" to another component, this includes not only the case of being 'directly connected', but also the case of being 'connected with another component in between'. In addition, when a certain component "includes" a certain component, this means that other components may be further included without excluding other components unless otherwise specified.

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings.

However, prior to explaining this, the meaning of the terms used below is first defined.

'Artificial neural network' is an information processing technology that engineeringally imitates the advanced information processing mechanism of the biological nervous system and performs complex control by relating inputs and outputs to each other in detail. It is a network in which a plurality of three types of neurons (nerve cell) models are connected, consisting of a hidden layer that prioritizes inputs and outputs and adjusts mutual relationships, and an output layer that calculates and outputs the necessary control amount based on this.

The quality inspection system 10, which will be described below, is coupled or connected to an inspection device that inspects a product using machine vision, for example, and learns an artificial neural network using an acquired product image, or inspects or measures whether a product is defective or not. can do.

Here, machine vision means automating the industry through a camera (visual recognition), CPU, and SW instead of the existing method in which a person judges with the naked eye to inspect or measure an object.

Also, supervised learning is a method of machine learning performed using training data including values to be output as results.

Unsupervised learning is a learning method to find out how data is structured. Unlike supervised learning or reinforcement learning, the learning data is not given a target value to be output as a result. .

‘Unclassified product’ is a product that is classified as a non-normal product by an unsupervised artificial neural network based on the photographed image of a normal product, and ‘defective product’ is a product that has defects among the unclassified products.

'Defect image' is an image of an existing product identified as a defective product among existing products that were produced on the production line, and 'defect image' is a virtual defect in the good product image of a new product produced on the production line where the existing product was produced. It is a virtual bad image created by synthesizing .

In addition to the terms defined above, terms that require explanation are separately explained below.

1 is a configuration diagram for explaining a quality inspection system 10 according to an embodiment.

Even when a new product is produced on a production line that used to produce existing products, the quality inspection system 10 photographs a new product being produced through a machine vision camera and inputs the acquired image to the artificial neural network to determine whether it is defective or not. can be inspected To this end, the quality inspection system 10 can generate learning data necessary for learning the artificial neural network used for quality inspection of a new product based on a defective image of an existing product with defects, and learn the generated learning data. to perform quality inspection of new products.

This quality inspection system 10 includes a learning data generation system 11 and learning data for generating learning data necessary for learning an artificial neural network for quality inspection of a new product based on a defective image, which is an image of an existing product with defects. It may include a defect identification system 12 that learns and performs a quality inspection of a new product.

In addition, each of the learning data generation system 11 and the defect identification system 12 constituting the quality inspection system 10 can be implemented as a computer capable of accessing a remote server through a network N or connecting to other terminals and servers. can Here, the computer may include, for example, a laptop computer, a desktop computer, and a laptop computer equipped with a web browser.

In addition, each of the learning data generation system 11 and the defect identification system 12 constituting the quality inspection system 10 may be physically separated and implemented in a separate server or in one server. In addition, the quality inspection system 10 may be implemented in the form of a cloud in which physically separated servers are connected through a network to logically form one system.

First of all, the learning data generation system 11 learns a defective image, which is an image of an existing product determined to have a defect among images of an existing product photographed through a machine vision camera, and learns a new product in the defect identification system 12 to be described later. It is possible to generate learning data used for quality inspection.

And the defect identification system 12 can learn by inputting the learning data generated in the above-described learning data generation system 11 into the artificial neural network, and using the learned artificial neural network, the new product photographed through the machine vision camera. Based on the image, it is possible to inspect whether the new product is defective, and for the new product determined to be defective, the type of defect may be identified.

At this time, each of the learning data generation system 11 and the defect identification system 12 may include an artificial neural network or be connected to an artificial neural network implemented in a physically separated third server, depending on the embodiment. Hereinafter, it is assumed that an artificial neural network is implemented in the learning data generation system 11 and the defect identification system 12.

2 is a block diagram showing the configuration of a quality inspection system 10 according to an embodiment. Referring to FIG. 2 , a quality inspection system 10 according to an embodiment may include a learning data generating system 11 and a defect identification system 12 .

First, the learning data generation system 11 may include a learning data control unit 111, a learning data communication unit 112, and a learning data memory 113.

The learning data control unit 111 controls the overall operation of the learning data generating system 11 and may include a processor such as a CPU. The learning data control unit 111 may control other components included in the learning data generation system 11 to perform an operation for generating learning data by processing data received through the learning data communication unit 112 .

For example, the learning data controller 111 may execute a program stored in the learning data memory 113, read a file stored in the learning data memory 113, or store a new file in the learning data memory 113. there is.

The learning data control unit 111 may obtain a defective image, which is an image of an existing product with defects, through the learning data communication unit 112 to be described later, and may learn the obtained defective image.

For example, the learning data control unit 111 may acquire a defective image, which is an image of an existing product having a defect among existing products for which a quality inspection was performed by the defect identification system 12 to be described later, and the defective image Through this, it is possible to learn the pattern, position, size or shape of defects included in the defective image.

Further, the learning data control unit 111 may learn a good product image, which is an image of a new product capable of passing the quality test.

For example, the learning data control unit 111 may acquire a non-defective product image for a new product passing the quality inspection through the learning data communication unit 112 and learn the obtained non-defective product image.

Thereafter, the learning data control unit 111 may generate a defect image for a new product based on the learned defective image of the existing product and good image of the new product.

For example, the learning data control unit 111 may generate various types of defects based on the pattern, position, size, or shape of defects included in the learned defective image, and convert the generated defects into a good image of a new product. can be synthesized to create a defect image.

In this way, the defect identification system 12, which will be described later, can be rapidly learned using a defect image generated by combining a defect occurring in an existing product with a non-defective image of a new product.

Also, the learning data control unit 111 may generate learning data required for learning for quality inspection of a new product.

For example, the training data control unit 111 may generate training data by cropping an area in which defects exist in each of the defective image and the defective image, or may generate training data including a defective image and a defective image.

Thereafter, the learning data control unit 111 may provide the generated learning data to the defect identification system 12 to be described later.

Meanwhile, the learning data communication unit 112 may perform wired/wireless communication with other devices or networks. To this end, the learning data communication unit 112 may include a communication module supporting at least one of various wired/wireless communication methods. For example, the communication module may be implemented in the form of a chipset.

The wireless communication supported by the learning data communication unit 112 may be, for example, Wi-Fi (Wireless Fidelity), Wi-Fi Direct, Bluetooth, UWB (Ultra Wide Band), or NFC (Near Field Communication). . In addition, wired communication supported by the learning data communication unit 112 may be, for example, USB or High Definition Multimedia Interface (HDMI).

The learning data communication unit 112 may receive a good image of a new product or obtain a defective image of an existing product from a third server or a defect identification system 12 to be described later, and the learning data control unit 111 generates Learning data may be transmitted to the defect identification system 12 .

In the learning data memory 113, various types of data such as files, applications, and programs may be installed and stored. The learning data control unit 111 may access and use the data stored in the learning data memory 113, or may store new data in the learning data memory 113. In addition, the learning data control unit 111 may execute a program installed in the learning data memory 113 .

Meanwhile, the defect identification system 12 may include a defect identification input/output unit 121, a defect identification control unit 122, a defect identification communication unit 123, and a defect identification memory 124.

The defect identification input/output unit 121 of the defect identification system 12 displays information such as an input unit for photographing a new product being produced through a camera for machine vision and the result of work or the status of the defect identification system 12. It may include an output unit for For example, the defect identification input/output unit 121 may include a camera arranged on a production line to photograph a new product being produced and a display panel displaying a screen.

Specifically, the input unit may include devices capable of receiving various types of user inputs, such as a keyboard, a physical button, a touch screen, a camera, or a microphone. In particular, a camera for machine vision can be placed in a production line where products are produced and take pictures of products moving along the production line in real time.

Also, the output unit may include a display panel or a speaker. However, it is not limited thereto, and the defect identification input/output unit 121 may include a configuration supporting various inputs/outputs.

The defect identification control unit 122 of the defect identification system 12 controls the overall operation of the defect identification system 12 and may include a processor such as a CPU. And the defect identification control unit 122 is implemented as an artificial neural network or by using an artificial neural network implemented in a third server, based on the captured image of the new product acquired through the defect identification input/output unit 121, whether the new product is normal or normal. Defects can be inspected, and for this, other components included in the defect identification system 12 can be controlled.

The defect identification control unit 122 may learn good quality images of new products.

For example, the defect identification control unit 122 may acquire and learn a non-defective image of a new product included in the learning data or a non-defective image of a new product photographed through the defect identification input/output unit 121 .

And the defect identification control unit 122 can learn using the learning data obtained from the learning data generating system 11 .

For example, the defect identification control unit 122 divides only a defect region in a defect image, which is a virtual defect image for a new product included in the learning data generated by the learning data generation system 11, into a preset size, You can perform patch learning.

In this way, defects similar to the types of defects that occur in existing products occur because the production line (process) is the same through learning the area where defects exist through patch learning for learning data generated based on defective images of existing products. Quality inspections can also be performed on new products.

Also, the defect identification control unit 122 may perform a quality inspection of the new product based on the learning.

Depending on the embodiment, the defect identification control unit 122 may classify a new product that is different from the non-defective image of the new product among the new products subject to quality inspection as an unclassified product.

For example, the defect identification control unit 122 may compare an image of a new product produced on a production line with a machine vision camera of the defect identification input/output unit 121 and a previously learned non-defective product image, and may compare a new product image that is different from the non-defective product image. A product can be classified as an unclassified product.

Also, the defect identification control unit 122 may identify defective products among unclassified products.

For example, the defect identification control unit 122 may identify whether a defect learned through patch learning exists on an image of an unclassified product, and may specify an area where the defect is identified.

Thereafter, the defect identification control unit 122 may determine the defect type of the defective product based on the defect of the defective product among the unclassified products.

At this time, according to the embodiment, when learning defects using the learning data, the defect identification control unit 122 may classify and learn defects according to defect types.

For example, when learning the learning data, the defect identification control unit 122 may classify defects based on patterns, sizes, locations, or shapes of defects included in the learning data, and determine defect types according to the classified defects. there is.

Also, the defect identification control unit 122 may determine the defect type of the defective product based on the defects of the defective product.

For example, the defect identification control unit 122 determines that a defective product, which is a new organic light emitting diode panel, has a defect on the surface, such as a line defect, a spot defect, or an edge defect. Defect types of defective products can be determined.

In this way, even if a new product is produced on the same production line, the application time of machine vision for a new product is shortened by learning the defect area by patch learning the defective image of the existing product and applying it to the quality inspection of the new product. At the same time, accurate quality inspection can be performed.

3 to 5 are flowcharts for explaining a quality inspection method according to an exemplary embodiment.

The quality inspection method according to the embodiment shown in FIGS. 3 to 5 includes steps processed time-sequentially in the quality inspection system 10 shown in FIGS. 1 and 2 . Therefore, even if the contents are omitted below, the contents described above with respect to the quality inspection system 10 shown in FIGS. 1 and 2 can also be applied to the quality inspection method according to the embodiment shown in FIGS. 3 to 5. .

First of all, referring to FIG. 3, the learning data generation system 11 of the quality inspection system 10 generates learning data necessary for learning for quality inspection of a new product based on a defective image, which is an image of an existing product with defects. It can be done (S3001).

For example, the learning data generation system 11 uses a defect included in a defective image, which is a photographed image of an existing product with defects among existing products, through a machine vision camera installed on a production line to generate a good image of a new product. Defects can be combined to create a defect image.

4 is a flowchart specifically illustrating a method of generating learning data according to an embodiment.

Referring to FIG. 4 , the learning data generating system 11 may learn a defective image of an existing product (S4001).

For example, the learning data generation system 11 can acquire an image of an existing product photographed through a camera for machine vision installed on a production line, and learn the acquired image by an unsupervised learning method to determine the existing defective product. Defective images of products can be learned.

Alternatively, for example, the learning data generating system 11 may learn the defective image of the existing product acquired from the manager using a supervised learning method.

Further, the learning data generation system 11 may generate a defect image for the new product based on the good product image, which is an image for the new product capable of passing the quality inspection (S4002).

For example, the learning data generation system 11 may acquire a good product image of a new product, synthesize defects included in the existing product learned in step S4001 based on the acquired good product image, and create a defect image for the new product. can create At this time, the learning data generation system 11 may synthesize defects in a similar or identical pattern based on the defect image of the existing product for the position, size, and shape of the defect to be synthesized when generating the defect image.

Then, the learning data generating system 11 may generate learning data based on at least one of a defective image and a defective image (S4003).

For example, the learning data generation system 11 may generate learning data including defect images including various types of defects for new products through step S4002.

Thereafter, the defect identification system 12 of the quality inspection system 10 may learn the learning data generated in step S3001 (S3002).

For example, the defect identification system 12 may segment at least one of a defect image or a defect image, which is an image included in the training data generated in step S3001, to patch-learn a defect-containing region.

In addition, the defect identification system 12 may perform learning by obtaining a non-defective image of a new product.

For example, the defect identification system 12 may acquire an image of a new product and learn a non-defective image of the new product through an unsupervised learning method on the obtained image.

And the defect identification system 12 may perform a quality inspection of the new product based on the learning (S3003).

For example, the defect identification system 12 may identify whether a defect exists in an image of a new product photographed through a camera for machine vision based on the learning in step S3002.

5 is a flowchart illustrating a quality inspection process for a new product according to an embodiment.

Referring to FIG. 5, the defect identification system 12 can photograph new products produced on the production line through a camera for machine vision (S5001), and based on the good product image of the new product learned in step S3002, the new product. can be classified into normal products and unclassified products.

For example, the defective identification system 12 may classify a new product different from the non-defective product image as an unclassified product based on the non-defective image of the new product learned by the unsupervised learning method in step S3002. In addition, the defect identification system 12 may classify new products that are not classified as unclassified products as normal products.

And the defect identification system 12 can identify the existence of a defect in the image of the new product classified as an unclassified product (S5003).

For example, the defect identification system 12 may identify whether a defect of the same or similar type as the defect learned by the patch method in step S3002 exists based on the image of the new product classified as an unclassified product. And if there is no defect in the image of the new product classified as unclassified product, the defect identification system 12 can classify the new product as the unclassified product as a normal product. can be classified as

Thereafter, the defect identification system 12 may identify the defect type of the defect product (S5004).

To this end, the defect identification system 12 may learn defects included in the learning data for each defect type using an unsupervised learning method when learning the learning data in step S3002.

For example, the defect identification system 12 may learn the pattern, size, position, shape, etc. of each defect generated according to the type of defect, and based on the learning, the image of the new product classified as a defective product in step S5003. Based on the defects included in the line defect, spot defect, edge defect, light leak defect, butterfly defect, stain defect, and un-uniformity The defect type may be identified as one of uniform defects, cloud defects, and terrain defects.

In this way, by using the defective image of the existing product to create and learn the defect image for the new product, the time required for the collection and learning of learning data required for quality inspection using machine vision for the new product is minimized and at the same time fast It can perform quality inspections on new products and identify types of defective products.

The term '~unit' used in the above embodiments means software or a hardware component such as a field programmable gate array (FPGA) or ASIC, and '~unit' performs certain roles. However, '~ part' is not limited to software or hardware. '~bu' may be configured to be in an addressable storage medium and may be configured to reproduce one or more processors. Therefore, as an example, '~unit' refers to components such as software components, object-oriented software components, class components, and task components, processes, functions, properties, and procedures. , subroutines, segments of program patent code, drivers, firmware, microcode, circuitry, data, databases, data structures, tables, arrays, and variables.

Functions provided within components and '~units' may be combined into smaller numbers of components and '~units' or separated from additional components and '~units'.

In addition, components and '~units' may be implemented to play one or more CPUs in a device or a secure multimedia card.

The quality inspection method according to the embodiment described with reference to FIGS. 3 to 5 may also be implemented in the form of a computer-readable medium storing instructions and data executable by a computer. In this case, instructions and data may be stored in the form of program codes, and when executed by a processor, a predetermined program module may be generated to perform a predetermined operation. Also, computer-readable media can be any available media that can be accessed by a computer and includes both volatile and nonvolatile media, removable and non-removable media. Also, a computer-readable medium may be a computer recording medium, which is a volatile and non-volatile memory implemented in any method or technology for storage of information such as computer-readable instructions, data structures, program modules, or other data. It may include both volatile, removable and non-removable media. For example, computer storage media include magnetic storage media such as HDDs and SSDs, optical media such as CDs, DVDs and Blu-ray discs, or media accessible through a network. It may be memory included in the server.

In addition, the quality inspection method according to the embodiment described with reference to FIGS. 3 to 5 may be implemented as a computer program (or computer program product) including instructions executable by a computer. A computer program includes programmable machine instructions processed by a processor and may be implemented in a high-level programming language, object-oriented programming language, assembly language, or machine language. . Also, the computer program may be recorded on a tangible computer-readable recording medium (eg, a memory, a hard disk, a magnetic/optical medium, or a solid-state drive (SSD)).

Therefore, the quality inspection method according to the embodiment described with reference to FIGS. 3 to 5 can be implemented by executing the above-described computer program by a computing device. A computing device may include at least some of a processor, a memory, a storage device, a high-speed interface connected to the memory and a high-speed expansion port, and a low-speed interface connected to a low-speed bus and a storage device. Each of these components are connected to each other using various buses and may be mounted on a common motherboard or mounted in any other suitable manner.

Here, the processor may process commands within the computing device, for example, to display graphic information for providing a GUI (Graphic User Interface) on an external input/output device, such as a display connected to a high-speed interface. Examples include instructions stored in memory or storage devices. As another example, multiple processors and/or multiple buses may be used along with multiple memories and memory types as appropriate. Also, the processor may be implemented as a chipset comprising chips including a plurality of independent analog and/or digital processors.

Memory also stores information within the computing device. In one example, the memory may consist of a volatile memory unit or a collection thereof. As another example, the memory may be composed of a non-volatile memory unit or a collection thereof. Memory may also be another form of computer readable medium, such as, for example, a magnetic or optical disk.

Also, the storage device may provide a large amount of storage space to the computing device. A storage device may be a computer-readable medium or a component that includes such a medium, and may include, for example, devices in a storage area network (SAN) or other components, such as a floppy disk device, a hard disk device, an optical disk device, or a tape device, flash memory, or other semiconductor memory device or device array of the like.

The above-described embodiments are for illustrative purposes, and those skilled in the art to which the above-described embodiments belong can easily transform into other specific forms without changing the technical spirit or essential features of the above-described embodiments. You will understand. Therefore, it should be understood that the above-described embodiments are illustrative in all respects and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope to be protected through this specification is indicated by the following claims rather than the detailed description above, and should be construed to include all changes or modifications derived from the meaning and scope of the claims and equivalent concepts thereof. .

10: Quality inspection system
11: Learning data generation system
111: learning data control unit
112: learning data communication unit
113: learning data memory
12: defect identification system
121: defect identification input and output unit
122: defect identification control unit
123: defect identification communication unit
124: defect identification memory

Claims (18)

In the quality inspection system for inspecting the quality of new products,
Learning a defective image, which is an image of an existing product with defects, and generating a defect image by creating a defect in a good image, which is an image of a new product that can pass the quality inspection based on the learning of the defective image, a learning data generating system for generating learning data by cropping a defective image and a region in which defects exist in each of the defective images; and
A defect identification system that learns the learning data and performs quality inspection of the new product based on the learning;
The defect identification system,
The good product image is learned in an unsupervised manner, and new products that are different from the good product image among the new products subject to the quality inspection are classified as unclassified products based on the learning of the good product image, and new products not classified as the unclassified product A quality inspection system that classifies products as normal products and classifies and identifies defective products with defects and normal products without defects among the unclassified products based on the learning of the learning data. delete delete delete delete delete delete According to claim 1,
The defect identification system,
A quality inspection system for determining a defect type of the defective product based on defects on the defective product. In the method for the quality inspection system to inspect the quality of a new product,
generating learning data necessary for learning for quality inspection of the new product based on a defective image, which is an image of an existing product with defects; and
Performing a quality inspection of the new product based on learning of the learning data;
The step of generating the learning data,
learning the defective image, and generating a defect image for the new product by generating a defect in a good product image, which is an image for a new product capable of passing a quality test, based on the learning of the defective image; and
generating the learning data by cropping an area in which a defect exists in each of the defective image and the defective image;
The step of performing the quality inspection is,
The good product image is learned in an unsupervised manner, and new products that are different from the good product image among the new products subject to the quality inspection are classified as unclassified products based on the learning of the good product image, and new products not classified as the unclassified product Classifying as a normal product; and
And learning the learning data, and classifying and identifying defective products with defects and normal products without defects among the unclassified products based on the learning of the learning data. delete delete delete delete delete delete According to claim 9,
The quality inspection method,
and determining a defect type of the defective product based on defects on the defective product. A computer-readable recording medium in which a program for performing the method according to claim 9 is recorded. A computer program performed by the quality inspection system and stored in a medium to perform the method according to claim 9.

Images