KR101950681B1 - Defect inspection device, system, method and computer readable storage medium - Google Patents

Abstract

Provided is a method for inspecting a defect of an object. According to an embodiment of the present invention, the method comprises the steps of: determining an object type by using an object image; dividing the object image into one or more partial images based on the determined object type; converting the partial image into a rectangular patch image based on each shape of the one or more divided partial images; inspecting, by a machine learning module, whether an object is defective by using the converted rectangular patch image by using one or more rectangular patch images; and outputting, by the machine learning module, a result of an inspection.

Description

≪ Desc / Clms Page number 1 > DEFECT INSPECTION DEVICE, SYSTEM, METHOD AND COMPUTER READABLE STORAGE MEDIUM,

The present disclosure relates to an apparatus, a system, a method, and a computer-readable storage medium for inspecting a defect in an object having one or more pattern types.

Unless otherwise stated herein, the description in this section is not prior art to the claims in this application and should not be considered prior art for the reasons described in this section.

Recently, with the development of artificial intelligence technology, there are increasing attempts to combine machine learning techniques in various industrial fields. Defects can be formed on products manufactured through a series of processes. Normally, these defects are detected through a separate defect inspection process, and a machine learning technique can be used for inspection of such defects.

For example, Korean Patent Laid-Open Publication No. 10-2017-0074812 (Prior Art Document 1) discloses a system and method for inspecting a semiconductor sample, and performs learning about an inspection-related application using an image of a semiconductor sample . According to Korean Patent Registration No. 10-1711073 (Prior Art Document 2), an apparatus and a method for detecting micro cracks formed on a flexible touch screen panel to which a deep learning algorithm is applied are disclosed. In this prior art document 2, the image to be inspected is divided into blocks of predetermined sizes, and the divided blocks are separated into square patches to detect defects using the machine learning technique using the detailed patches.

In order to detect defects formed in a product by using a machine learning technique, it is necessary to perform a large number of learning operations on defect detection, and since such learning needs to be performed according to the type of defects, . As shown in the prior art documents 1 and 2, an object such as a semiconductor sample or a touch screen panel has a pattern of a single type, and for an object having a single type pattern, if an image of the object is cut into a patch of a predetermined size, Since the patches have patterns of the same type, defect inspection can be collectively performed for each patch using a machine learning technique. However, when the object to be inspected has multiple types of patterns, the conventional defect inspection method is not suitable, or a very large amount of information may be required for machine learning.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems and it is an object of the present invention to provide a defect detection apparatus, a system, a method, and a computer readable storage medium capable of improving the efficiency of learning about defect detection of a machine learning module, present.

In some embodiments of the present disclosure, a defect inspection method for inspecting an object having one or more pattern types is described. One exemplary method may be performed under the control of a computing device, and the computing device may include a machine learning module. In some examples, a defect detection method includes determining an object type using an object image for an object; Dividing an object image into one or more partial images, each based on a determined type of object, wherein each of the one or more partial images has one pattern, wherein one pattern is one of one or more pattern types; Converting the partial image into a rectangular patch image based on the shape of each of the one or more partial images divided; Checking whether an object is defective by using a patch image of the transformed rectangle by a machine learning module of a computing device that performs machine learning on defect detection using a patch image of one or more squares; And outputting a result of the inspection by the machine learning module.

In some embodiments, a defect inspection apparatus is described. One exemplary defect inspection apparatus may include an object type determination module, an image segmentation module, a patch image conversion module, a machine learning module, and an output module. The object type determination module may be configured to determine an object type using an object image for the object. The image segmentation module may be configured to segment the object image into one or more partial images based on the type of object determined by the object type determination module. The image segmentation module may divide each of the one or more images into one pattern, wherein one pattern is one of one or more pattern types. The patch image conversion module may be configured to convert the partial image into a rectangular patch image based on the shape of each of the one or more partial images divided by the image segmentation module. The machine learning module may be configured to use the patch image of the transformed rectangle to check whether the object is defective. The machine learning module may also be configured to perform learning for defect detection using one or more rectangular patch images containing patch images of the transformed rectangles. The output module may be configured to output the result of the inspection by the machine learning module.

In some embodiments, a computer-readable storage medium storing one or more computer-executable instructions for inspecting a defect of an object having one or more pattern types is described. When the one or more computer executable instructions are executed, causing the computing device to determine an object type using an object image for the object; Dividing an object image into one or more partial images based on a determined type of object, each of the one or more partial images having one pattern, wherein one pattern is one of one or more pattern types; Converting the partial image into a rectangular patch image based on the shape of each of the one or more partial images segmented; Checking whether an object is defective by using a patch image of the transformed rectangle by a machine learning module of a computing device that performs machine learning on defect detection using one or more rectangular patch images; And outputting the result of the inspection by the machine learning module.

The foregoing brief summary and effect description are exemplary only and are not intended to limit the spirit of the invention disclosed in the present disclosure. Additional embodiments and technical features may be understood in addition to the above-described exemplary embodiments and technical features, with reference to the following detailed description and the accompanying drawings.

The features of the present disclosure described above and other additional features will be described in detail below with reference to the accompanying drawings. These drawings depict only a few embodiments in accordance with the present disclosure and are not to be construed as limiting the scope of the present teachings. The technical spirit of the present disclosure will be described in more detail and in detail using the accompanying drawings.
1 is a block diagram that schematically illustrates a defect inspection apparatus according to at least some embodiments of the present disclosure;
2 is an exemplary block diagram of a defect inspection system including a defect inspection apparatus according to Fig. 1;
3 is a flow chart illustrating an exemplary process for a defect inspection method in accordance with at least some embodiments of the present disclosure;
Figure 4 illustrates an exemplary computer program product that may be used to perform defect inspection in accordance with at least some embodiments of the present disclosure;
5 is an exemplary block diagram of a computing device arranged in accordance with at least some embodiments of the present disclosure;
6 is a diagram illustrating an example of a defect check according to at least some embodiments of the present disclosure;

Hereinafter, embodiments and embodiments of the present invention will be described in detail with reference to the accompanying drawings, so that those skilled in the art can easily carry out the present invention. It should be understood, however, that the present invention may be embodied in many different forms and is not limited to the embodiments and examples described herein.

The present disclosure generally relates to a defect inspection apparatus, system, method and computer-readable storage medium for inspecting an object having one or more pattern types, and is suitable for performing defect inspection for objects having patterns of one or more types Apparatus, system, method, and computer-readable storage medium.

1 is a block diagram that schematically illustrates a defect inspection apparatus 100 according to at least some embodiments of the present disclosure. The defect inspection apparatus 100 is a computing device that can perform processing for checking whether an object is defective by using an image of an object to be inspected. An object may be an object having multiple types of patterns, such as various types of mechanical parts, including, for example, gears, various types of sockets, etc., as well as objects having a single type of pattern such as semiconductor, . The defect inspection apparatus 100 may be, for example, a server-type computer such as a stand-alone server, a web server, etc., a desktop computer, a laptop computer, a smart phone, a tablet computer, Electronic devices such as, but not limited to, electronic devices.

1, the exemplary defect inspection apparatus 100 includes a memory 110, an object type determination module 120, an image segmentation module 130, a patch image conversion module 140, a machine learning module 150 And an output module 160. [0027] Each of the components included in the defect inspection apparatus 100, such as the object type determination module 120, the image segmentation module 130, the patch image conversion module 140, the machine learning module 150, and the output module 160, Or may be implemented in a manner such that two or more of the components combine to form one component. The components included in the defect inspection apparatus 100 may form a device, a server, a program unit, or a suitable combination thereof. In addition, some of the components of the defect inspection apparatus 100, such as the object type determination module 120, the image segmentation module 130, and the patch image conversion module 140, may perform functions, operations, May cooperate with the machine learning module in accordance with additional embodiments to be described below for implementation.

The memory 110 may store an image of an object (hereinafter referred to as an "object image"). The object images stored in the memory 110 may be stored as well-known various types of image files. In some examples, the memory 110 may store processed data according to the operation of the image segmentation module 130, the patch image transformation module 140, and the output module 160.

In some embodiments, the object type determination module 120 may read an object image stored in the memory 110. The object type determination module 120 can be configured to determine an object type using an object image. Various image recognition techniques can be used to determine the object type. In some instances, the object type determination module 120 may determine that the object being inspected from the object image is one of more than one type of predetermined object. The type of object may include, but is not limited to, the type, configuration or type of the object, the type of pattern the object has, the position of the pattern the object has, and the like. In some other instances, object type determination module 120 may cooperate with machine learning module 150 to determine the type of object from an object image.

In some embodiments, the image segmentation module 130 may be configured to segment the object image into one or more partial images, based on the type of object determined by the object type determination module 120. The image segmentation module 130 may divide an object image according to a pattern type. That is, each of the one or more partial images divided by the image segmentation module 130 may have one pattern. In some instances, the image segmentation module 130 may utilize various shape recognition algorithms for segmenting the object image. In one example, an object may include one or more shape patterns, and an object may divide one or more shape patterns into partial images, respectively. For example, the image segmentation module 130 may use an algorithm such as the Hough circle transform to detect the circular pattern, or may use an algorithm such as the Ramer-Douglas-Peucker algorithm to detect the polygonal pattern. In some other instances, image segmentation module 130 may cooperate with machine learning module 150 to segment the object image into one or more partial images. Each of the one or more segment images may be in various forms such as a circle, a circle, an ellipse, a polygon such as a circle, a triangle, a rectangle, a pentagon, etc., a sector, a ring,

In some embodiments, the patch image conversion module 140 may be configured to convert the partial image into a patch image based on the shape of each of the one or more partial images that have been segmented. As such, the patch image transform module 140 may generate one or more patch images using one or more partial images segmented by the image segmentation module 130. The patch image is a rectangular image for facilitating processing by a computing device such as the defect inspection apparatus 100.

In some embodiments, the patch image transform module 140 may use various types of transforms to transform the partial image into a rectangular patch image. In some examples, the patch image transform module 140 may transform a partial image into a rectangular patch image using coordinate transformations such as polar coordinate to cartesian coordinate transforms. Additionally or alternatively, the patch image transform module 140 may use various types of image processing techniques, such as skew transform, shear transform, and / or image warping, Can be converted. In some examples, the patch image conversion module 140 may be configured to convert the partial image based on the pattern type corresponding to the pattern of the partial image. Partitioning of the object image by the image segmentation module 130 and an example of the transformation of the partial image by the patch image transformation module 140 are illustrated in FIG. 6 and will be described in detail below.

In some embodiments, the machine learning module 150 may perform pre-learning for defect detection using a rectangular patch image prior to examining the defect of the object. Thereafter, the machine learning module 150 may be configured to check whether there is a defect in the object using the patch image of the rectangle converted by the patch image conversion module 140. On the other hand, the machine learning module 150 can also use the rectangular patch image used for the inspection in the learning of the machine learning module 150. In some instances, a user of the fault detection apparatus 100 may input feedback on the test results determined by the machine learning module 150 via an input device (not shown) and, based on such feedback, Learning for detection can be performed. The machine learning module 150 may utilize various known machine learning techniques. For example, the machine learning module 150 may include a Deep Neural Network (DNN), a Convolutional Neural Network (CNN), a Recurrent Neural Network (RNN), a Support Vector Machine Vector Machine (SVM), and so on. In some instances, the machine learning module 150 may utilize various image processing techniques such as canny edge detection, SHIFT, SURF, etc. to detect defects in an object.

In some embodiments, the output module 160 may be configured to output the results of the inspection by the machine learning module 150. In some examples, the results of the inspection may include at least one of the type of the detected defect, the location of the detected defect, and the reliability for the detected defect. In some additional examples, the output module 160 may perform a final determination of the failure based on predetermined conditions, based on the inspection results of the machine learning module 150. For example, the predetermined criterion may include, but is not limited to, the sensitivity set by the user or the type of the detected object defect.

2 is an exemplary block diagram of a defect inspection system 200 including a defect inspection apparatus 100 according to FIG. The defect inspection system 200 according to FIG. 2 may include a defect inspection apparatus 100 and an image capture apparatus 210 configured to acquire an object image from an object. The image capturing apparatus 210 may be connected to the defect inspection apparatus 100 in a wired or wireless manner. The image capture device 210 may transmit the acquired object image to the defect inspection apparatus 100 via a wired or wireless network. The image capturing apparatus 210 may be formed integrally with the defect inspection apparatus 100, but may be configured as a separate apparatus from the defect inspection apparatus 100.

FIG. 3 is a flow chart illustrating an exemplary process 300 for a defect inspection method in accordance with at least some embodiments of the present disclosure. For example, the process 300 may be performed under the control of a computing device, such as the fault detection device 100 of FIGS. Process 300 may be performed to examine objects having one or more pattern types. An object can include objects having multiple types of patterns, such as various types of mechanical parts, including, for example, semiconductor, textured objects, and objects having a single type of pattern, e.g., gears, various types of sockets, . The process 300 illustrated in FIG. 3 may include one or more actions, functions, or actions as illustrated by blocks 310, 320, 330, 340, and / or 350. On the other hand, the schematic operations illustrated in FIG. 3 are provided by way of example only, and some of the operations may be optional, combined with fewer operations, or extended with additional operations, without departing from the essence of the disclosed embodiments . Process 300 may begin at block 310, which determines the type of object.

At block 310, the computing device may determine the type of object using an object image for the object. In some instances, the computing device may determine from the object image that the object to be examined is one of more than one type of predetermined object. In some other instances, a computing device may use a machine learning module to determine the type of object from an object image. In determining the object, various image recognition techniques can be used. The type of object may include, but is not limited to, the type, configuration or kind of the object, the pattern type of the object, the position of the pattern of the object, and the like. The process 300 may proceed to block 310, which may proceed to block 320, where the object image is divided into one or more partial images.

At block 320, the computing device may partition the object image into one or more partial images, based on the type of object determined at block 310. [ The computing device may partition the object image by pattern type. In some instances, a computing device may partition an object image using various shape recognition algorithms such as, for example, a Hough circle transform, a Ramer-Douglas-Peucker algorithm, and the like. Process 300 may proceed to block 330, which converts the partial image to a rectangular patch image at block 320. [

At block 330, the computing device may convert each of the one or more sub-images segmented at block 320 into a rectangular patch image. The computing device may convert the partial image based on the shape of the partial image. In some instances, a computing device may transform a partial image into a rectangular patch image using coordinate transformations such as polar coordinate to cartesian coordinate transforms. Additionally or alternatively, the computing device may transform a partial image using various types of image processing techniques, such as skew transform and / or shear transform, image warping, and the like . In some instances, the computing device may convert the partial image based on the pattern type corresponding to the pattern of the partial image. The process 300 may lead to block 340, which uses the machine learning module at block 330 to check for an object defect using the translated patch image of the rectangle.

At block 340, the computing device can use the patch image of the transformed quadrangle at block 330 to inspect the object for defects by the machine learning module of the computing device. On the other hand, the machine learning module of the computing device can perform learning for defect detection using one or more square patch images such as a patch image of a transformed rectangle. Process 300 may proceed to block 350, which outputs the result of the check at block 340. [

At block 350, the computing device may output the result of the examination of the defect performed by the machine learning module at block 340. [ In some examples, the results of the inspection may include at least one of the type of the detected defect, the location of the detected defect, and the reliability for the detected defect. In some additional examples, the computing device may perform a final determination of a fault based on a pre-determined condition based on the inspection results of the machine learning module.

As described above, according to the present disclosure, it is possible to inspect defects for each pattern type by dividing an image for each pattern for an object having multiple types of patterns. By converting an image divided for each pattern into a rectangular patch image, Since the image inspected by the module can be normalized for each pattern type, it is possible to improve the efficiency of learning about the defect detection of the machine learning module and the defect detection performance.

FIG. 4 illustrates an exemplary computer program product 400 that may be used to perform defect inspection in accordance with at least some embodiments of the present disclosure. Exemplary embodiments of an exemplary computer program product are provided using signal embedding medium 402. [ In some embodiments, the signal-bearing medium 402 of one or more computer program products 400 may comprise a computer-readable medium 406, a recordable medium 408 and / or a communication medium 410.

The instructions 404 included in the signal bearing medium 402 may be executed by a computing device, such as a fault detection device according to the present disclosure. Instruction 404, when executed, may cause the computing device to check for defects in objects having one or more pattern types. Instruction 404 may include one or more instructions for determining the type of object using an object image for the object; One or more instructions for partitioning an object image into one or more partial images based on a determined type of object, each of the one or more partial images having the one pattern, wherein one pattern is one of one or more pattern types; One or more instructions for transforming the partial image into a rectangular patch image based on the shape of each of the one or more partial images segmented; One or more instructions for inspecting an object for defects using a patch image of the transformed rectangle by a machine learning module of a computing device that performs machine learning on defect detection using one or more rectangular patch images; And one or more instructions for outputting the results of the inspection by the machine learning module.

5 is an exemplary block diagram of a computing device 500 that is arranged in accordance with at least some embodiments of the present disclosure. In one exemplary basic configuration 502, computing device 500 may include one or more processors 504 and system memory 506. [ The memory bus 508 may be used to communicate between the processor 504 and the system memory 506. Depending on the configuration desired, the processor 504 may be of any type, including, but not limited to, a microprocessor (uP), a microcontroller (uC), a digital signal processor (DSP), or any combination thereof. The processor 504 may include one or more levels of a cache, such as a level 1 cache 510, a level 2 cache 512, a processor core 514, and a register 516. The processor core 514 may comprise an arithmetic logic unit (ALU), a floating point unit (FPU), a digital signal processing core (DSP Core), or any combination thereof. The memory controller 518 may also be used with the processor 504 or, in some implementations, the memory controller 518 may be an internal portion of the processor 504. [

Depending on the configuration desired, the system memory 506 may be any type, including but not limited to volatile memory (such as RAM), non-volatile memory (such as ROM, flash memory, etc.), or any combination thereof, Do not. The system memory 506 may include an operating system 520, one or more applications 522, and program data 524. The application 522 may be implemented in any of a variety of ways, including the functional blocks and / or operations described with respect to the defect detection apparatus 100, the system 200, and / or the process 300 of FIG. 3, And a contribution computation algorithm 526 arranged to perform the function. In some embodiments, application 522 may be arranged to operate with program data 524 on operating system 520 such that an implementation for determining an optimal transmission environment may be provided as described herein. For example, the fault detection apparatus 100 may include all or a portion of the computing device 500 and may include all or some of the applications 522 such that an implementation that determines an optimal transmission environment may be provided as described herein. Lt; / RTI > This described basic configuration is illustrated by its components in dashed line 502 in FIG.

The computing device 500 may have additional features or functionality and additional interfaces to facilitate communication between the basic configuration 502 and any desired devices and interfaces. For example, the bus / interface controller 530 may be used to facilitate communication between the basic configuration 502 and the one or more data storage devices 532 via the storage interface bus 534. [ The data storage device 532 may be a removable storage device 536, a fixed storage device 538, or a combination thereof. Examples of removable storage devices and fixed storage devices include magnetic disk devices such as flexible disk drives and hard disk drives (HDD), compact disk (CD) drives or digital versatile disk (DVD) drives, Optical disk drives, solid state drives (SSDs), and tape drives, and the like. Exemplary computer storage media includes volatile and nonvolatile removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules, or other data can do.

The system memory 506, the removable storage device 536, and the fixed storage device 538 are all examples of computer storage media. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technology, CD-ROM, digital versatile disks (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, But is not limited to, any other medium which can be used to store the required information and which can be accessed by the computing device 500. Any such computer storage media may be part of the computing device 500.

The computing device 500 also includes an interface bus 542 for facilitating communication from various interface devices (e.g., an output interface, a peripheral interface, and a communication interface) to the basic configuration 502 via a bus / ). Exemplary output device 542 may include a graphics processing unit 648 and an audio processing unit 550 which may be coupled to one or more A / V ports 552 to communicate with various external devices, Lt; / RTI > The exemplary peripheral interface 544 may include a serial interface controller 554 or a parallel interface controller 556 which may be coupled to an input device (e.g., a keyboard, a mouse, a pen, A voice input device, a touch input device, etc.) or other peripheral device (e.g., a printer, a scanner, etc.). Exemplary communication device 546 includes a network controller 560 that may be arranged to facilitate communication with one or more other computing devices 562 on a network communication via one or more communication ports 564. [

6 is a diagram illustrating an example of a defect check according to at least some embodiments of the present disclosure; 6 (a) shows an exemplary object image 610, and 620-1, 620-2, 620-3, 620-4, 620-5, 620-6, 620-7, ..., 620-n represent partial images divided by the pattern from the object image 610, and 630-1, 630-2, 630-3, 630-4 , 630 - 6, 630 - 7, 630 - n are partial images 620 - 1, 620 - 2, 620 - 3, 620 - -7, ..., 620-n).

As shown in Figure 6, according to the present disclosure, an object image 610 is generated for each of the patterns 620-1, 620-2, 620-3, 620-4, 620-5, 620-6 , 620-7, ..., 620-n. Each of these partial images 620-1, 620-2, 620-3, 620-4, 620-5, 620-6, 620-7, ... 620- Direction, angle, and the like. For example, partial images 620-5, 620-6, and 620-7 have the same pattern type as images for bolts / nuts, but differ in position, direction, angle, etc. within object image 610 , And in the machine learning for defect detection, each of the partial images 620-5, 620-6, and 620-7 may be used as different types of learning data. However, according to the present disclosure, such partial image partial images 620-1, 620-2, 620-3, 620-4, 620-5, 620-6, 620-7, ..., 620- 630 - 6, 630 - 7,..., 630 - n) of the rectangular pattern image to perform normalization for each pattern type do. For example, according to the present disclosure, by converting partial images 620-5, 620-6, and 620-7 having the same pattern type into rectangular patch images 630-5, 630-6, and 630-7, A patch image for a pattern having a pattern type may have a similar shape. Thus, according to the present disclosure, the machine learning module can efficiently perform learning for defect detection on a pattern-by-pattern basis.

The claimed subject matter is not to be limited in scope to the specific embodiments described herein. For example, some implementations may be in hardware, such as may be used to operate on a device or combination of devices, while other implementations may be in software and / or firmware, for example. Likewise, although the claimed subject matter is not limited in scope in this respect, some embodiments may include one or more articles, such as signal bearing media, storage media. Such a storage medium, such as a CD-ROM, a computer disk, a flash memory, etc., may, for example, be executed by a computing device, such as a computing system, computing platform or other system, Thereby storing instructions that can cause the execution of the processor. As one possibility, a computing device may include one or more processing units or processors, one or more input / output devices such as a display, a keyboard and / or a mouse, and a static random access memory, a dynamic random access memory, a flash memory and / Lt; / RTI > memory.

There is little distinction between hardware and software implementations of aspects of the system; The use of hardware or software is a design choice that generally represents a trade-off between cost-effectiveness (although not always in that the choice between hardware and software in some contexts may be important) . There are a variety of vehicles (e.g., hardware, software and / or firmware) in which the processes and / or systems and / or other technologies described herein may be affected and preferred means include processes and / And / or the context in which other technologies are used. For example, if the implementer decides that speed and accuracy are the most important, the implementer can chose mainly hardware and / or firmware means, and if flexibility is of paramount importance, the implementer can chose mainly the software implementation; Alternatively, as an alternative, the implementer may select any combination of hardware, software, and / or firmware.

The foregoing detailed description has described various embodiments of devices and / or processes by way of block diagrams, flowcharts, and / or illustrations. As long as such block diagrams, flowcharts, and / or illustrations contain one or more functions and / or operations, those skilled in the art will recognize that each function and / or operation in such block diagrams, flowcharts or illustrations may be implemented in hardware, software, firmware, It is to be understood that they may be implemented individually and / or collectively by virtually any of a wide range of combinations. In one embodiment, some portions of the subject matter described in this disclosure may be implemented in the form of an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), or other integrated form. However, those skilled in the art will appreciate that some aspects of embodiments of the present disclosure may be implemented as a combination of one or more computer programs (e.g., one or more programs running on one or more computer systems) running on one or more computers, (I. E., One or more programs running on one or more microprocessors), firmware, or substantially any combination thereof, may be implemented in whole or in part in an integrated circuit, software and / And / or the design of the circuitry is within the skill of one of ordinary skill in the art in light of the present disclosure. It will also be understood by those skilled in the art that the mechanisms of the subject matter of the present disclosure may be distributed in various types of program products and examples of objects of the present disclosure may include certain types of signal bearing media ≪ / RTI >

Although specific example techniques have been described and illustrated herein using various methods and systems, it should be understood by those skilled in the art that various other modifications can be made, and equivalents may be substituted, without departing from the claimed subject matter. In addition, many modifications may be made to adapt a particular situation to the teachings of the claimed subject matter without departing from the central concept described herein. Accordingly, it is intended that the claimed subject matter is not limited to the specific examples disclosed, but that the claimed subject matter may also include all embodiments falling within the scope of the appended claims and their equivalents.

Claims (13)

A defect inspection method for inspecting an object having at least one pattern type, the inspection being performed under the control of a computing device, the computing device comprising a machine learning module for performing machine learning on defect detection,
Determining a type of the object using an object image for the object;
Dividing the object image into one or more portions based on the determined type of the object, wherein the object image has a plurality of portions, each of the one or more portions having one pattern and the one pattern being one of the one or more pattern types Divided to have a pattern type -;
Wherein the machine learning module is configured to perform a machine learning of the defect detection for each pattern type based on the type of each of the divided one or more portions and the pattern type corresponding to the pattern of the corresponding portion, Converting each of the one or more portions into a patch image, wherein the patch image of the rectangle to be transformed is transformed to be normalized for each pattern type;
Checking, by the machine learning module of the computing device, whether the object has a defect using the transformed rectangular patch image; And
Outputting a result of the inspection by the machine learning module
And a defect inspection method. The method according to claim 1,
Wherein determining the type of the object comprises determining one of one or more types of predetermined objects. The method according to claim 1,
Wherein determining the type of object comprises determining the type of the object using the machine learning module. The method according to claim 1,
Wherein dividing the object image into one or more parts comprises using the figure recognition algorithm or the machine learning module to divide the object image such that each of the one or more parts has one pattern. method of inspection. The method according to claim 1,
The step of converting the quadrangle into a patch image may include transforming the corresponding portion using at least one of a coordinate transform, a skew transform, a shear transform, and image warping And a defect inspection method. delete The method according to claim 1,
Wherein the result of the checking includes at least one of a type of the detected defect, a position of the detected defect, and a predicted reliability for the detected defect. A defect inspection apparatus for inspecting an object having at least one pattern type,
An object type determination module configured to determine a type of the object using an object image for the object;
An image segmentation module configured to segment the object image into one or more portions based on the type of the object determined by the object type determination module, the image segmentation module comprising: The pattern having a pattern type of one of the at least one pattern type;
A patch image conversion module configured to convert the portion into a rectangular patch image based on the type of each of the one or more portions divided by the image segmentation module and the pattern type corresponding to the pattern of the portion, The patch image of the rectangle to be transformed is transformed to be normalized for each pattern type;
Checking whether there is a defect in the object using the converted patch image of the quadrangle and performing defect detection for each pattern type using one or more rectangular patch images including the converted patch image of the quadrangle A machine learning module configured to perform machine learning for the machine; And
An output module configured to output a result of the inspection by the machine learning module
And a defect inspection apparatus. 9. The method of claim 8,
Wherein the image segmenting module is configured to divide the object image so that each of the one or more portions has one pattern using a graphics recognition algorithm. 9. The method of claim 8,
Wherein the patch image conversion module is configured to convert the corresponding portion using at least one of a coordinate transform, a skew transform, and a shear transform. delete A defect inspection apparatus according to claim 8; And
An image capture device configured to acquire an image of an object from an object
And a defect inspection system. A computer-readable medium having stored thereon one or more computer-executable instructions for examining a defect in an object having one or more pattern types, the computer-readable medium having computer-executable instructions that, when executed, Allowing a computing device, including a machine learning module,
Determining an object type of the object using an object image of the object;
Dividing the object image into one or more portions based on the determined type of the object, wherein the object image has a plurality of portions, each of the one or more portions having one pattern and the one pattern having one of the one or more pattern types Divided to have a pattern type -;
Wherein the machine learning module is configured to perform a machine learning of the defect detection for each pattern type based on the type of each of the divided one or more portions and the pattern type corresponding to the pattern of the corresponding portion, Converting each of the one or more parts into a patch image of a rectangle to be transformed so as to be normalized for each pattern type corresponding thereto;
Checking, by the machine learning module of the computing device, whether the object has a defect using the transformed rectangular patch image; And
An operation for outputting a result of the inspection by the machine learning module
≪ / RTI >

Images