Examination method based on fringe field and deep learning

Abstract

According to one embodiment of the present disclosure, a defect inspection apparatus is disclosed. The apparatus may comprise: an illumination system designed to transmit illumination patterns having different illuminance according to regions to the surface of an object to be inspected; an imaging unit for acquiring image data of the object to be inspected; one or more processors for processing the image data; and a memory storing the deep learning-based model. Furthermore, in one or more processors, by controlling the illumination system, the illumination system transmits illumination patterns having different illuminance according to regions to the surface of the object to be inspected, and image data acquired by the imaging unit is input to the model based on deep learning. The image data includes a dramatic change in illuminance generated in at least one region of the object surface due to the transmitted illumination pattern, and a defect on the surface of the object to be inspected can be determined using the model based on deep learning.

Description

Technical Field

[0001] The present invention relates to a defect inspection device, and more particularly, to a device for inspecting whether a surface of an inspection target object has defects. Background Art

[0002] During the process of manufacturing an object in a factory, the object may have defects due to machinery, process or other reasons. In the factory, before the object is completed and shipped out, a step of confirming whether the corresponding object has defects is usually performed.

[0003] Furthermore, recently, in order to shorten the process time and quickly inspect the defects of the object, a technology for judging the quality of the inspection object using computer vision has been introduced in this field.

[0004] However, the computer vision technology of the prior art only uses an illumination system with a diffuser attached, so the illumination is severely scattered. When inspecting an inspection object having a smooth surface, a sparkling surface, or a glossy surface, the recognition rate of surface defects such as fine scratches, height differences, and dents is reduced.

[0005] The Korean authorized patent No. 10-1517929 (invention name: high-speed robot viewing device capable of uninterrupted inspection of components) discloses the following device: a multi-joint robot arranged outside the inspection object is used to photograph the inspection object, and the quality of the inspection object is determined from the photographed image. Summary of the invention

[0006] The present disclosure has been made in response to the above-mentioned background art, and an object of the present disclosure is to provide an inspection device that can more effectively inspect defects on the surface of an inspection target object.

[0007] A defect inspection device of an embodiment of the present disclosure for achieving the above-mentioned purpose may include: an illumination system designed to transmit an illumination pattern having different illuminations according to regions to the surface of an object to be inspected; a photographing unit for acquiring image data of the object to be inspected; one or more processors for processing the image data; and a memory for storing a model based on deep learning. Furthermore, in one or more of the processors, the illumination system may be controlled to transmit an illumination pattern having different illuminations according to regions to the surface of the object to be inspected, and the image data acquired by the photographing unit may be input into the model based on deep learning, wherein the image data includes a drastic change in illumination in at least one region of the surface of the object due to the transmitted illumination pattern, and the model based on deep learning may be used to judge defects on the surface of the object to be inspected.

[0008] In an alternative embodiment, the illumination system of the defect inspection apparatus may include a plurality of light sources, and the plurality of light sources may be arranged at predetermined intervals.

[0009] In an alternative embodiment, the above-mentioned lighting system of the defect inspection device may include a plurality of light sources, and an illumination pattern having different illumination intensities according to regions may be transmitted to the surface of the inspection target object by controlling the operation of each of the plurality of light sources.

[0010] In an alternative embodiment, the above-mentioned illumination system of the defect inspection apparatus may include a plurality of light sources composed of at least two combinations.

[0011] In an alternative embodiment, the illumination system of the defect inspection apparatus includes a plurality of light sources, and when the light sources emit light, the number of fringe fields per unit area generated in a region of the surface of the inspection object is at least different from that in other regions.

[0012] In an alternative embodiment, the illumination system of the defect inspection apparatus transmits an illumination pattern having different illumination in each area so that an illumination variation area including a sharp variation in illumination covers at least a portion of the surface of the inspection object.

[0013] In an alternative embodiment, in the illumination pattern having different illumination intensities according to regions transmitted by the illumination system of the defect inspection apparatus, the plurality of light sources of the illumination system may be controlled to transmit in a linear, polygonal, or circular shape.

[0014] In an alternative embodiment, the above-mentioned processor of the defect inspection device can determine defects on the surface of the inspection target object by inputting multiple image data acquired under different lighting patterns into a deep learning-based model.

[0015] In an alternative embodiment, the above-mentioned lighting system of the defect inspection device can be composed of multiple lighting panels equipped with multiple light sources, and the above-mentioned lighting panels can be manufactured in the form of a punched area so that at least a portion of the light transmitted from the lighting panels located in the upper layer reaches the surface of the inspection object.

[0016] In an alternative embodiment, the processor of the defect inspection apparatus may further extract an illumination change region including a dramatic change in illumination from the image data.

[0017] In a computer-readable storage medium storing a computer program for determining whether there are defects on the surface of an object to be inspected, an embodiment of the present disclosure for achieving the purpose as described above, when the computer program is run on more than one processor, a neural network is used to perform the following actions for generating data, which may include: causing an illumination system designed to transmit different illumination patterns to transmit different illumination patterns according to different regions to the surface of the object to be inspected; acquiring image data, the image data including a dramatic change in illumination generated in at least one region of the surface of the object to be inspected by transmitting the illumination pattern through the illumination system; and determining whether the object has defects based at least in part on the illumination change portion of the image data.

[0018] A method for inspecting whether the surface of an inspection object has defects according to an embodiment of the present disclosure for achieving the purpose as described above may include the following steps: controlling a lighting system so that the lighting system designed to transmit different lighting patterns transmits different lighting patterns according to regions to the surface of the inspection object; inputting image data acquired by a shooting unit into a deep learning-based model, the image data including a drastic change in illumination caused in at least one region of the object surface by the transmitted lighting pattern; and determining defects on the surface of the inspection object using the deep learning-based model.

[0019] The present disclosure can provide an inspection device that can more effectively inspect defects on the surface of an inspection target object. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A diagram showing a block structure diagram of a defect inspection device according to an embodiment of the present disclosure.

[0021] Figure 2 FIG. 1 is a diagram showing an example for explaining the concept of an edge field defined in the present disclosure.

[0022] Figure 3a and Figure 3b An illustration of a variable diffuser and a lighting system according to an embodiment of the present disclosure is shown.

[0023] Figure 3c and Figure 3d This is a simplified diagram of light transmitted through the surface of an inspection target object according to the operating state of a variable diffuser according to one embodiment of the present disclosure.

[0024] Figure 4 A diagram showing the structure of an illumination system applicable to a defect inspection device according to an embodiment of the present disclosure.

[0025] Figure 5 FIG. 4 is a diagram illustrating the diffusion degree of light according to the transparency of a variable diffuser according to an embodiment of the present disclosure.

[0026] Figure 6 FIG. 4 is a diagram illustrating a method in which a processor adjusts the diffusion degree of light by controlling the transparency of a variable diffuser according to an embodiment of the present disclosure.

[0027] Figure 7 FIG. 4 is a diagram illustrating a method in which a processor adjusts the diffusion degree of light by controlling the configuration of a variable diffuser according to yet another embodiment of the present disclosure.

[0028] Figure 8 A diagram showing an example of controlling a variable diffuser for a defect inspection device according to regions according to an embodiment of the present disclosure.

[0029] Fig. 9 2 is a diagram illustrating the varying transparency when a processor controls the transparency of a variable diffuser according to an embodiment of the present disclosure.

[0030] Fig.10a , Fig.10b FIG. 1 is a diagram illustrating a change in visibility of a dark region on the surface of an inspection target object according to the transparency of a variable diffuser.

[0031] Fig.11 , Fig.12 , Fig.13 , Fig.14 , Fig.15 , Fig.16 This is a diagram illustrating image data acquired by controlling the transparency of a variable diffuser according to the surface characteristics of an inspection target object according to an embodiment of the present disclosure.

[0032] Fig.17 A diagram showing the structure of a plurality of light sources included in an illumination system applicable to a defect inspection device according to an embodiment of the present disclosure.

[0033] Fig.18 This is an illustrative diagram of an illumination pattern having different illumination intensities according to regions transmitted to the surface of an inspection object by an illumination system of a defect inspection device according to an embodiment of the present disclosure.

[0034] Fig.19 This is an illustrative diagram of the structure of an illumination system applicable to a defect inspection device according to another embodiment of the present disclosure.

[0035] Fig. 20 A diagram showing a cross section of an illumination system applicable to a defect inspection device according to an embodiment of the present disclosure.

[0036] Fig.21 This is an illustrative diagram of a defect inspection device according to an embodiment of the present disclosure marking a defect location in an inspection object.

[0037] Fig. 22FIG. 4 is a block diagram of a computing device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0038] Hereinafter, various embodiments are described with reference to the accompanying drawings, and similar reference numerals are used to indicate similar structural elements throughout the specification. In this specification, various descriptions are used to provide an understanding of the present disclosure. However, it is obvious that such embodiments can be performed even without such specific descriptions.

[0039] The terms "component", "module", "system", etc. used in this specification refer to computer-related entities, hardware, firmware, software, a combination of software and hardware, or the operation of software. For example, a component can be a process, a processor, an object, an execution thread, a program and / or a computer running on a processor, but is not limited to this. For example, an application program and a computing device running in a computing device can both be components. One or more components can reside in a processor and / or an execution thread, a component can be located in a computer, or can be distributed between two or more computers. In addition, such components can be operated using a variety of computer-readable media with a variety of data structures stored internally. For example, multiple components can communicate through local and / or remote processing based on signals with more than one data packet (for example, data from a component that interacts with other components in a local system, a distributed system and / or through signals, data with other systems through a network such as the Internet).

[0040] Meanwhile, the term "or" refers to an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or the context is unclear, "X utilizes A or B" refers to one of the natural inclusive substitutions. That is, in the case where X utilizes A or X utilizes B or X utilizes A and B, "X utilizes A or B" can be used in any of the above cases. Furthermore, the term "and / or" used in this specification should be understood to refer to and include all possible combinations of more than one of the listed related items.

[0041] Furthermore, the terms "include" and / or "include..." should be understood as the existence of the corresponding features and / or structural elements. However, the terms "include" and / or "include..." should be understood as not excluding the existence or addition of one or more other features, structural elements and / or their combinations. Furthermore, in this specification and the scope of protection of the invention, if the context does not otherwise specify or clearly indicate the singular form, the singular should generally be interpreted as "one or more".

[0042] It is also necessary for those of ordinary skill to recognize that the various illustrative logic blocks, structures, modules, circuits, units, logic, and algorithm steps described in association with the embodiments disclosed herein are embodied by electronic hardware, computer software, or a combination of both. In order to clearly illustrate the interchangeability of hardware and software, various illustrative components, blocks, structures, units, logic, modules, circuits, and steps have been described above generally in terms of functionality. Whether the above functionality is embodied using hardware or software is related to the specific application (application) and design constraints assigned to the entire system. A person of ordinary skill can embody the functionality described in a variety of ways for each specific application. However, it should not be interpreted that the determination of the above embodiment exceeds the scope of the present disclosure.

[0043] The descriptions related to the disclosed embodiments are intended to enable a person of ordinary skill in the art to which the present disclosure belongs to utilize or implement the present disclosure. Various variations related to such embodiments are obvious to a person of ordinary skill in the art to which the present disclosure belongs, and the general principles defined herein may be applied to other embodiments without exceeding the scope of the present disclosure. Therefore, the present disclosure is not limited to the embodiments disclosed herein and must be interpreted in the widest scope consistent with the principles and novel features disclosed herein.

[0044] In this specification, the high diffusion of light transmitted to an object refers to a state in which the transparency of the variable diffuser is low. That is, in a state in which light is scattered, it means that the light reaches the object. Conversely, the low diffusion of light transmitted to an object refers to a state in which the transparency of the variable diffuser is high. That is, it means that the light is hardly scattered or the light emitted from the light source directly reaches the object. In other words, in this specification, the defect inspection device controls the transparency or configuration of the variable diffuser, and adjusts the diffusion of light transmitted to the object through the above notification.

[0045] Below, refer to Figure 1 The structure of the defect inspection device according to the embodiment of the present disclosure will be described.

[0046] Figure 1 The defect inspection device 100 is a block diagram showing a defect inspection device for determining whether an inspection object has an abnormality according to an embodiment of the present disclosure. The defect inspection device 100 may include a processor 110 , a memory 120 , a network 130 , a camera 140 , and a lighting system 200 .

[0047] Figure 1 The structure of the defect inspection device 100 shown is only a simplified example. In one embodiment of the present disclosure, the defect inspection device 100 may include other structures for inspecting whether there are defects on the surface of the inspection object, and only a part of the disclosed structure may constitute the defect inspection device 100.

[0048] The processor 110 may be composed of more than one core, and may include a data analysis processor such as a central processing unit (CPU) of a computing device, a general purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), etc., and a processor for deep learning. The processor 110 may run data processing for machine learning in one embodiment of the present disclosure by reading a computer program stored in the memory 120. According to one embodiment of the present disclosure, the processor 110 may run operations for learning a neural network. The processor 110 may run operations for learning a neural network, such as processing input data for deep learning (DL), feature extraction in input data, error calculation, weight value update of a neural network using backpropagation, etc. At least one of the central processing unit, the general purpose graphics processing unit, and the tensor processing unit of the processor 110 may process the learning of a network function. For example, the central processing unit and the general purpose graphics processing unit may process the learning of a network function and data classification using a network function together. Moreover, in one embodiment of the present disclosure, processors of multiple computing devices may be used together to process the learning of a network function and data classification using a network function. Furthermore, the computer program running in the computing device of an embodiment of the present disclosure may be a program executable by a central processing unit, a general-purpose graphics processing unit, or a tensor processing unit.

[0049] According to an embodiment of the present disclosure, the memory 120 may store any form of information generated or determined by the processor 110 and any form of information received by the network 130 .

[0050] According to an embodiment of the present disclosure, the memory 120 may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory (e.g., SD or XD memory, etc.), a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, and an optical disk. The computing device may operate in association with a web storage that performs the storage function of the above-mentioned memory 120 on the Internet. The records related to the above-mentioned memory are only for example, and the present disclosure is not limited thereto.

[0051] The network 130 of one embodiment of the present disclosure may use a variety of wired communication systems such as a public switched telephone network (PSTN), a digital subscriber line (xDSL), a rate adaptive digital subscriber line (RADSL), a multi-rate digital subscriber line (MDSL), a very high speed digital subscriber line (VDSL), a universal asymmetric DSL (UADSL), a high bit rate digital subscriber line (HDSL) and a local area network (LAN).

[0052] Furthermore, the network 130 disclosed in this specification may use a variety of wireless communication systems such as code division multiple access (CDMA), time division multiple access (TDMA), frequency division multiple access (FDMA), orthogonal frequency division multiple access (OFDMA), single carrier frequency division multiple access (SC-FDMA), and other systems.

[0053] In the present disclosure, the network 130 may be constructed regardless of the communication method such as wired or wireless, and may be composed of various communication networks such as a personal area network (PAN), a wide area network (WAN), etc. Furthermore, the above network may be the well-known World Wide Web (WWW), or may utilize wireless transmission technology for short-range communication such as infrared (IrDA) or Bluetooth.

[0054] The technology described in this specification is applicable not only to the above-mentioned network but also to other networks.

[0055] The defect inspection apparatus 100 according to an embodiment of the present disclosure may be embodied by the components described below to inspect defects of the inspection object 10. The inspection object 10 is an additional structure for explaining the structure or operation of the defect inspection apparatus, and is not a component of the defect inspection apparatus 100.

[0056] The defect inspection device 100 may include an illumination system 200 and a processor 110. The illumination system 200 may include a light source for projecting light toward an inspection target object; and a variable diffuser 300 located between the light source and the inspection target object 10 and capable of adjusting the diffusion degree of light transmitted toward the inspection target object. The processor 110 may control the variable diffuser based on the characteristics of the inspection target object 10.

[0057] The processor 110 controls the variable diffuser based on the characteristics of the inspection target object 10, and the characteristics of the inspection target object 10 include the characteristics of the surface of the object (for example, texture, material, color, gloss, etc.) and the characteristics selected by the user. For example, the processor 110 can determine the transparency of the variable diffuser required for inspection from the image data related to the surface of the object. And, as another embodiment, the processor 110 can control the variable diffuser for an object having the characteristics selected by the user according to a predetermined control method or a predetermined series of control sequences. Specifically, for high-priced products whose sales costs exceed a specific value, in order to improve the defect detection rate, the processor can control the variable diffuser, thereby checking the low state and high state of the diffusion of light transmitted to the object through the variable diffuser. In this case, the characteristics selected by the user can be the product price. The characteristics described above are only illustrative, and the present disclosure is not limited to this.

[0058] In this specification, the fringe field is a term used to refer to the portion where the illumination intensity of the light irradiated onto the surface of the object to be inspected changes due to the light source image when the light source image is formed and irradiated onto the surface of the object to be inspected. The fringe field is a term defined in this specification to correspond to the dark field and light field used in an optical system, and is an area within a predetermined distance from the brightest center of the light source when the illumination light source image is formed and irradiated onto the surface of the object to be inspected.

[0059] Below, refer to Figure 2 The concept of fringe fields is also explained.

[0060] The preset distance d (reference numeral 12) of the above-mentioned fringe field includes an area wider than the light source area 11 that forms and illuminates the light source image on the surface of the inspection target object 10. That is, the above-mentioned fringe field may include the light source area 11 that is directly illuminated by the light source and has a higher illumination than other areas, and an area 13 with relatively low illumination outside the portion that is not directly illuminated by the light source. The light source area 11 that forms and illuminates the light source image refers to the light source shape in which the shape of the light source itself that transmits light according to the characteristics of the surface of the inspection target object is directly displayed on the surface of the inspection target object. The shape of the light source includes a circular light source shape, a linear light source shape, a polygonal light source shape, etc. according to the form of the light source, but the present disclosure is not limited thereto and includes all light sources with specific shapes.

[0061] In summary, the fringe field refers to a total area including a relatively low illuminance area 13 existing in an area spaced apart by a preset distance d (reference numeral 12) from the above-mentioned light source area 11 and the center of the light source.

[0062] The preset distance of the fringe field can be large or small according to the characteristics of the inspection object and the user's choice. For example, when the inspection object is not allowed to have any small scratches or dents, the processor 110 sets the preset distance of the fringe field to be small and inspects the surface of the inspection object with very dense transmitted light. On the contrary, in the case of an inspection process with relatively low importance or when it is sufficient to only inspect defects above a specified size, the processor 110 can set the preset distance of the fringe field to be larger. The description related to the range of the fringe field as described above is only for example, and the present disclosure is not limited thereto.

[0063] In this specification, “a region including a sharp change in illuminance” is used in the same meaning as “a region including at least one fringe field”.

[0064] The reason for introducing the fringe field concept will be described below with reference to FIGS. 3A , 3B, 3C, and 3D.

[0065] Generally, in the art, in order to observe the surface of the inspection object, a diffuser is attached to the lighting plate of the lighting system 200 to adjust the diffusion degree of the light transmitted to the inspection object 10 to be high in the state 300b, and the defects of the object are inspected (see FIG. 3B). That is, light with a predetermined illuminance is transmitted to the entire surface of the inspection object 10 (see FIG. 3D). This is because, generally, if there is a light source area 11 that directly forms a light source image on the surface of the inspection object 10, the illuminance of each area is different and the surface cannot be uniformly observed.

[0066] However, in the case of detecting appearance defects such as scratches, height differences, dents, etc. from the surface of the inspection object 10 that regularly reflects incident light at a specific angle, such as the surface of a shiny, glossy material, the surface of a metal frame, or a mirror, the visibility of the fringe field optical system (refer to FIG. 3A ) without a diffuser or with a high transparency of the diffuser is increased. That is, if the diffusion degree of the light transmitted to the inspection object 10 is adjusted to be low in the state 300a, the surface of the inspection object 10 is illuminated so that the light source area 11 and the dark area 14 are cross-illustrated, the surface defect 1 that is difficult to detect when the diffuser is present can be easily detected (refer to FIG. 3C ).

[0067] The reason why the defect inspection device 100 of an embodiment of the present disclosure controls the variable diffuser based on the characteristics of the inspection object 10 is that, when the fringe field optical system that does not diffuse the light transmitted to the inspection object (refer to FIG. 3A ) and the general optical system that diffuses the light transmitted to the inspection object (refer to FIG. 3B ) as described above are appropriately used according to the surface of the inspection object, there is an advantage of increasing the probability of surface defect detection.

[0068] In the variable diffuser 300 included in the lighting system 200 of the defect inspection device 100 according to an embodiment of the present disclosure, the transparency is controlled by the electrical signal of the processor, thereby adjusting the diffusion of the light passing through. The variable diffuser 300 included in the lighting system 200 of the defect inspection device 100 may include a liquid crystal film that adjusts the transparency according to the electrical signal. According to another embodiment of the present disclosure, the variable diffuser 300 included in the lighting system 200 of the defect inspection device 100 may adjust the diffusion of the light transmitted to the inspection target object 10 by adjusting the configuration state according to the control signal of the processor.

[0069] Figure 4 The defect inspection device 100 according to an embodiment of the present disclosure includes an illumination panel and a variable diffuser 300 of an illumination system 200 .

[0070] To adjust the diffusivity of light transmitted to the inspection target object 10, the variable diffuser 300 can be controlled by an electrical signal of the processor 110 to adjust the transparency. To adjust the diffusivity of light transmitted to the inspection target object 10, the variable diffuser 300 can be controlled by an electrical signal of the processor 110 to change the configuration.

[0071] In the variable diffuser 300 included in the illumination system 200 of the defect inspection apparatus 100 , transparency is changed according to an applied voltage, whereby the diffusion degree of the light passing therethrough can be adjusted.

[0072] Below, refer to Figure 5 and Figure 6 The following is an explanation. In the state 300a where a voltage is applied to the variable diffuser 300 and the diffusion degree of light transmitted to the inspection object is adjusted to be low, each liquid crystal molecule inside the liquid crystal of the variable diffuser can be arranged in parallel in the moving path of the light. In this case, the light transmitted from the lighting plate maintains the moving direction of the light when passing through the variable diffuser 300, thereby directly reaching the inspection object 10. That is, the light passes through the variable diffuser in a state where the transparency is high, and therefore the diffusion degree of the light is low. On the contrary, in the state 300b where the diffusion degree of light transmitted to the inspection object 10 is adjusted to be high without applying a voltage to the variable diffuser 300, each liquid crystal molecule inside the liquid crystal of the variable diffuser can be randomly arranged in an arbitrary direction. In this case, the diffusion degree of light transmitted from the lighting plate and passing through such liquid crystal molecules becomes high, and thus the light is expanded before reaching the inspection object 10. That is, the light passes through the variable diffuser in a state where the transparency is low, and therefore the diffusion degree of the light is high.

[0073] The variable diffuser 300 that adjusts the diffusion degree of light according to the above-mentioned electrical signal can be formed of a material whose transparency is adjusted by the electrical signal. For example, the variable diffuser can use a liquid crystal (LCD) film, a polymer dispersed liquid crystal film (PDLC film), etc. The material of the variable diffuser as described above is only an example, and the present disclosure is not limited thereto.

[0074] The variable diffuser 300 included in the illumination system 200 of the defect inspection device 100 according to an embodiment of the present disclosure can also be controlled in a manner of adjusting the configuration in the three-dimensional space according to the electrical signal of the processor 110 .

[0075] Below, refer to Figure 7 Provide explanation. Figure 7 This is a diagram showing an example of controlling the spatial arrangement of the variable diffuser 300 in a “sliding” manner.

[0076] Figure 7 Part (a) is an example of changing the arrangement of the variable diffuser in order to maintain low the degree of diffusion of light transmitted through the variable diffuser 300 toward the inspection target object 10 by the electrical signal of the processor. Figure 7 Part (b) is an example diagram of providing a variable diffuser between a light source and an inspection target object in order to increase the degree of diffusion of light transmitted from the variable diffuser 300 to the inspection target object 10 by an electrical signal from a processor. In this case, the transparency of the variable diffuser 300 can be appropriately selected to be different depending on the inspection target object 10.

[0077] In the above Figure 7 The method for changing the spatial configuration of the variable diffuser described in the foregoing is merely an example. In order to adjust the diffusion degree of light transmitted to the inspection target object without limitation, the present disclosure includes a method for changing the spatial configuration of the variable diffuser 300 and an inspection device.

[0078] Generally, in the art, in order to inspect the prepared objects, an inspection device with a diffuser plate attached is used. However, as described above, in the case of an inspection target object having a luminous surface characteristic and an inspection target object having a matte surface characteristic, in order to detect surface defects, the need for a diffuser may be different. However, if the user directly assembles and disassembles the diffuser according to whether the diffuser is used for each product produced, it requires expensive costs and time. Therefore, the present disclosure can actively control the diffusion degree of light through the variable diffuser 300, thereby having the effect of saving production costs and time. Furthermore, the processor can control the transparency of the variable diffuser in real time, and various production products can be inspected by using a defect inspection device of the present disclosure, thereby having an economic advantage.

[0079] The variable diffuser 300 included in the illumination system 200 of the defect inspection device 100 according to an embodiment of the present disclosure can adjust the diffusion degree of light transmitted to different areas on the surface of the target object through the control signal of the processor 110 to be different.

[0080] In the case of a variable diffuser 300 in which the processor 110 controls the transparency using an electrical signal in order to adjust the diffusion degree of light transmitted to the inspection target object 10, the transparency of the variable diffuser 300 can be controlled to be different for each region. Figure 8 As shown, the processor 110 can control the variable diffuser 300 by region by distinguishing a region in which the diffusion of light transmitted to the inspection object 10 is adjusted to be low 300a and another region in which the diffusion of light transmitted to the inspection object 10 is adjusted to be high 300b. The processor 110 controls the variable diffuser 300 by region, which means that the transparency of at least one region and other regions of the variable diffuser 300 can be controlled to be different. The number of regions controlled with a transparency different from that of the at least one region can be plural. In the process of controlling the variable diffuser 300 by region, the processor 110 can control the variable diffuser 300 in a manner that makes the variable diffuser 300 have a specific pattern. The specific pattern can be a horizontal line, a vertical line, a diagonal line, a circle or a polygon. In order to control the specific pattern, the variable diffuser 300 can also be formed by a combination of multiple individual variable diffusers. The pattern by region is only an embodiment, and the present disclosure is not limited thereto. The variable diffuser can be controlled by any pattern to adjust the diffusion of light transmitted to the desired surface of the inspection object. According to the above-described control of the transparency of the variable diffuser 300 for each region, the required transparency state of the variable diffuser 300 can be configured at once for each region according to the surface of the inspection target object 10, thereby having the advantage of increasing the inspection speed.

[0081] When the processor 110 adjusts the configuration state of the variable diffuser in order to adjust the diffusivity of light transmitted to the inspection target object 10, the processor may position the variable diffuser 300 only between the light source and a portion of the inspection target object. If the variable diffuser is only positioned between the light source and the portion, light with high diffusivity may be transmitted to the portion, and light with low diffusivity may be transmitted to other regions.

[0082] The defect inspection device 100 according to an embodiment of the present disclosure may include a photographing unit for acquiring image data. The processor determines the diffusion degree of light transmitted to the inspection object based on the object identified by identifying the surface characteristics of the inspection object through the image data acquired from the photographing unit, thereby controlling the variable diffuser 300.

[0083] Fig. 9A variable diffuser 300 formed of a liquid crystal film included in the defect inspection apparatus 100 according to an embodiment of the present disclosure is shown. Fig. 9 Part (a) of FIG. 3 illustrates a state 300a in which the degree of diffusion of light transmitted toward the inspection target object 10 is adjusted to be low. Fig. 9 Part (b) of FIG. 3 illustrates a state 300 b in which the degree of diffusion of light transmitted toward the inspection target object 10 is adjusted to be high.

[0084] Hereinafter, an example of determining the diffusion degree of light transmitted to the inspection object based on the identified object is described with reference to FIGS. 10A and 10B. Part (a) of FIG. 10A shows a case where the processor 110 adjusts the diffusion degree of light transmitted to the inspection object 10 to be low by controlling the variable diffuser 300. On the contrary, part (b) of FIG. 10A shows a case where the processor 110 adjusts the diffusion degree of light transmitted to the inspection object 10 to be high by controlling the variable diffuser 300. 3a and 3b of FIG. 10A respectively refer to the same lines on the surface of the same inspection object. FIG. 10B shows an example of the illumination distribution related to the dark field on the surface of the inspection object measured along the above-mentioned line. That is, if the illuminance distribution is measured along the line segments at the same position of the dark area of ​​a region of the inspection object, it can be seen that the illuminance distribution is measured along the portion 3a displayed in the image data when the diffusion of the light transmitted to the inspection object 10 is adjusted to be low (refer to the portion (a) of FIG. 10B ) is more uniform than the illuminance distribution is measured along the portion 3b displayed in the image data when the diffusion of the light transmitted to the inspection object 10 is adjusted to be high (refer to the portion (b) of FIG. 10B ). The uniform distribution of illuminance means that the region is subdivided accordingly, which improves the image quality of the image data related to the dark region and ultimately increases the efficiency of appearance defect detection.

[0085] In one embodiment of the present disclosure, the processor 110 identifies the inspection object from the image data obtained by the photographing unit, and determines the diffusion degree of light transmitted to the inspection object 10 based on the reflectivity of light on the surface of the inspection object, thereby controlling the variable diffuser.

[0086] The reflectivity of light on the surface of the object used as a basis by the above-mentioned processor may include a surface of a reflective object formed of a smooth material or a glossy material and reflecting most of the incident light, a surface formed of a matte material and reflecting a portion of the incident light and washing away the remaining light, and a surface formed of a transparent material and passing most of the incident light.

[0087] Below, refer to Figures 11 to 16 Embodiments of the invention are described.

[0088] Figures 11 to 16FIG. 1 is a diagram showing image data acquired according to the diffusion degree of light when the diffusion degree of light is adjusted by a variable diffuser according to an embodiment of the present disclosure. Figures 11 to 16 In the figure, the left image of each figure (part (a) of the figure) is an example of image data taken in an environment where the diffusion of light transmitted to the inspection object 10 is adjusted to be low by the variable diffuser 300, and the right image of each figure (part (b) of the figure) is an example of image data taken in an environment where the diffusion of light transmitted to the inspection object 10 is adjusted to be high by the variable diffuser 300. As observed in the exemplary image data, by adjusting the diffusion of light of the variable diffuser, the recognition possibility of the defective part shown in the form of a box in each figure can be different. The defect inspection device 100 of an embodiment of the present disclosure can control the variable diffuser according to the surface characteristics of the inspection object, the defect characteristics of the inspection object, etc., so that the recognition possibility of the defect becomes higher, thereby adjusting the diffusion of light transmitted to the inspection object.

[0089] Fig.11 Part (a) of Fig.12 Part (a) of Fig.13 Part (a) of Fig.14 Part (a) of Fig.15 Part (a) of Fig.16 Part (a) shows the case where the transparency of the variable diffuser is controlled high. Fig.11 Part (b) of Fig.12 Part (b) of Fig.13 Part (b) of Fig.14 Part (b) of Fig.15 Part (b) of Fig.16 Part (b) shows the case where the transparency of the variable diffuser is controlled to be low.

[0090] Fig.11 This figure shows an example of how the transmitted light differs depending on the degree of diffusion when the surface of the inspection target object is formed of a glossy material. Fig.11 The portion (a) of FIG. 1 has the following advantages, that is, if the marked portion 3a is observed, a part of the scratch as a surface defect can be clearly observed, and the portion 3a of FIG. 1 can be clearly observed when the marked portion 3a is observed. Fig.11 In the case of the portion 3b marked in part (b), there is an advantage that the entire shape of the scratch can be confirmed.

[0091] Fig.12 This figure shows an example of how the transmitted light differs according to the degree of diffusion when the object to be inspected is made of a transparent material. Fig.12 The part 3a marked in part (a) of Fig.12The portion 3b marked at the same position as the image data captured in part (b) can more clearly confirm the scratches existing on the surface of the transparent inspection object.

[0092] Fig.13 This is image data of a transparent glue sticking to an object to be inspected during the manufacturing process. Fig.13 The portion 3a marked with the portion (a) shows the glue present on the surface of the object to be inspected. Fig.13 In the part (b) of FIG. 3b marked at the same position, it can be seen that compared with the low diffusion Fig.13 Part (a) of the image above makes it easier to identify surface defects.

[0093] Fig.14 1 is an exemplary image data when there is a dent mark defect on the surface of the inspection target object 10. It can be confirmed that the image data captured in a state where the diffusion degree of the transmitted light is low (refer to Fig.14 The change of light around the defect of the portion 3a marked in (a) of FIG. 1 is greater than the image data taken when the diffusion of the transmitted light is high (refer to FIG. 1 ). Fig.14 The same marked part 3b of (b) is more serious.

[0094] Fig.15 The following is an example of a state where a defect with an irregular curvature exists on the surface of the inspection target object 10. Image data captured in a state where the diffusion degree of the transmitted light is low (see Fig.15 The portion 3a marked in (a) of FIG. 1 is larger than the image data captured when the diffusion degree of the transmitted light is high (refer to FIG. 1 ). Fig.15 The same marked portion 3b of part (b) more clearly acquires image data related to the defective portion.

[0095] Fig.16 The example below shows a case where a defect of a specific form exists on the surface of the inspection target object 10. Image data captured in a state where the diffusion degree of the transmitted light is low (see Fig.16 The portion 3a marked in (a) of FIG. 1 is larger than the image data captured when the diffusion degree of the transmitted light is high (refer to FIG. 1 ). Fig.16 The same marked portion 3b in part (b) of FIG. 1 is used to obtain higher resolution image data related to the defective portion.

[0096] In the above Figures 11 to 16 The illustrated example is merely an example, and the present invention is not limited thereto.

[0097] As described above, the degree of light diffusion required for effective inspection varies depending on the surface state of the inspection object 10. The defect inspection device disclosed in the present invention has the advantage that the transparency of the variable diffuser can be actively controlled in real time by adjusting the transparency of the variable diffuser by the processor in consideration of the reflectivity of the light on the inspection object surface.

[0098] In one embodiment of the present disclosure, when the inspection object 10 is identified and the diffusion degree of light required in at least one area on the surface of the inspection object is different from that in other remaining areas, the processor 110 can control the variable diffuser 300 according to the area. Alternatively, when the inspection object 10 is identified and the diffusion degree of light required in at least one area on the surface of the inspection object is different from that in other remaining areas, the processor 110 can control the variable diffuser 300 in a manner that has at least two different transparencies. In this case, in order to control the variable diffuser 300 according to the area, the variable diffuser can be formed by a combination of multiple individually controlled variable diffusers. For example, in the case where the area on the surface of the inspection object identified by the processor 110 is a mobile phone screen with a glossy material and a mobile phone frame with a matte material, in an area of ​​the inspection object corresponding to the screen, the transparency of the variable diffuser is increased to reduce the diffusion degree of light, and in the remaining area of ​​the inspection object corresponding to the frame, the transparency of the variable diffuser is reduced to increase the diffusion degree of light, thereby enabling rapid and accurate inspection. The above description is only illustrative, and the present disclosure is not limited thereto.

[0099] The processor 110 identifies the inspection object, and then determines the transparency state required for inspection among the low transparency state, high transparency state, or any transparency state between the two of the variable diffuser, and can control the variable diffuser. The transparency state required for the above inspection may include more than one state. Specifically, the processor 110 can inspect defects on the surface of the inspection object in only one transparency state according to the transparency state required for the inspection. In addition, in the processor 110, when it is determined that it is necessary to compare images after shooting in the low transparency state and the high transparency state according to the transparency state required for the inspection, the image data shot in the different transparency states of the variable diffuser can be compared. In addition, the processor performs the action of controlling the variable diffuser 300 to be one of the low transparency state, the high transparency state, or any transparency state between the two for multiple times, and can compare the image data shot in each state.

[0100] The processor 110 may determine whether there are defects in the inspection object using image data including the surface of the inspection object. The processor 110 processes the image data related to the inspection object using a model based on deep learning, thereby determining whether there are defects in the inspection object. The processor 110 inputs the image data to the model based on deep learning, and may determine whether there are defects in the inspection object using the model based on deep learning. In addition, the processor 110 may include the result of classifying defects existing on the surface of the inspection object from the image data using the model based on deep learning.

[0101] In one embodiment of the present disclosure, the deep learning-based model stored in the memory 120 may be a model learned to perform classification, detection, segmentation, etc. on input image data. The learning of the model may be performed in a supervised learning or unsupervised learning manner related to the image data.

[0102] The processor 110 inputs image data through the above-mentioned deep learning-based model, and can input whether there are defects in the input image of the inspection object. The processor 110 can also use the deep learning-based model to classify whether there are defects and the type of defects. For example, the type of defect can be classified as scratches or dent marks. The processor 110 outputs whether there are defects. At the same time, if it is judged that there are defects, it can be displayed in a bounding box by detecting the corresponding defect position. The processor 110 can use the deep learning-based model learned for providing segmentation results to determine whether there are defects. At the same time, if it is judged that there are defects, the pixels of the area with defects and the pixels of other areas are distinguished and the defect position is marked in the output image data.

[0103] As described above, the processor 110 inputs image data of the inspection object into the deep learning model and obtains output related to whether there are defects. The present disclosure can ensure the visibility of surface defects of the inspection object regardless of the complexity of the image, thereby increasing the defect recognition rate.

[0104] In one embodiment of the present disclosure, for one or more image data according to the transparency of the variable diffuser, the processor 110 may use one or more sub-models of the deep learning-based model to determine defects of the inspection object included in the image data. Specifically, the one or more sub-models refer to deep learning-based models that are learned separately according to the state of the variable diffuser. That is, the diffusion degree of light transmitted to the inspection object is different according to the various transparencies of the variable diffuser, and therefore, a sub-model based on deep learning optimized for each diffusion degree may be used. The processor may use the above sub-models to output whether each model is defective according to the transparency of the variable diffuser. After the processor uses the above sub-models to calculate whether there is a defect, it may also mark the image data of the inspection object at the same time to generate an output data. In a specific embodiment, when the transparency of the variable diffuser does not diffuse the transmitted light, the processor inputs the acquired non-diffused light image data to the first sub-model of the deep learning-based model to determine the defects of the inspection object included in the above image data; and when the transparency of the variable diffuser diffuses the transmitted light, the processor may input the acquired diffused light image data to the second sub-model of the deep learning-based model to determine the defects of the inspection object included in the above image data.

[0105] As described above, the fringe field optical system (refer to FIG. 3B ) is useful for surface defects of the inspection object 10 having a smooth surface, but the reason why the fringe field optical system has not been used in previous industrial sites is that the image is highly complex and it is not easy to derive the presence or absence of defects using the previous image processing system. Therefore, the present invention has the effect of quickly and easily inspecting defects on the surfaces of more kinds of objects by combining image data related to the fringe field and a model based on deep learning. Below, the processor 110 of the defect inspection device 100 of an embodiment of the present disclosure controls the transparency of the above-mentioned variable diffuser 300 included in the lighting system 200 and maintains the diffusion degree of the light passing through in a low state (refer to Figure 5 Part (a) of Figure 6 Part (a) of Figure 7 In part (a) of FIG. 1 , that is, when there is a region including a dramatic change in illumination on the surface of the inspection target object 10 through which light is transmitted by the lighting system 200, additional control of the processor 110 and the effect thereof are described.

[0106] The processor 110 may acquire image data including an illumination pattern transmitted through the illumination system 200 on the surface of the object. The processor 110 may acquire the image data from the imaging unit 140 that images the surface of the inspection target object. Alternatively, the processor 110 may acquire image data stored in the memory 120 as described above, and may also acquire image data received from an external memory through the network 130. The image data may include a dramatic change in illumination in at least one area of ​​the surface of the object through the transmitted illumination pattern. Fig.17 The above-mentioned lighting system 200 used in the defect inspection device of one embodiment of the present disclosure is described. As one embodiment of the present disclosure, the lighting system includes a plurality of light sources, and the plurality of light sources can be arranged at predetermined intervals. In addition, the plurality of light sources of the lighting system can also be arranged in a predetermined shape. The above-mentioned light sources include incandescent lamps, halogen lamps, fluorescent lamps, electrodeless lamps, high-pressure sodium lamps, light-emitting diode (LED) lamps, etc., and also include, without limitation, light sources that light up according to electrical signals.

[0107] The predetermined interval refers to the distance between the light sources included in the lighting system. The predetermined interval can be determined according to the distance between the edge fields to be formed on the surface of the object to be inspected. The predetermined shape refers to the shape shown by the configuration of at least two light sources included in the lighting system. The predetermined interval and the predetermined shape can be determined according to the characteristics of the object to be inspected. As a specific embodiment, the predetermined shape of the multiple light sources can be a chessboard arrangement 200a in which at least one other light source exists above, below, left, and right of a specific light source, at least one light source forms a circle or an ellipse at a specified interval, and a plurality of the above-mentioned circles or ellipse lines form a concentric circle relationship to be scattered on the same surface 200b. In order to reflect the predetermined shape of the multiple light sources, in the process of configuring the light sources, the predetermined interval can be removed according to the required distance between the edge fields. Although not shown in the drawings, as another embodiment, multiple light sources are configured so that another light source only exists on the diagonal line of a light source, thereby forming a lattice structure. Furthermore, as one embodiment of the invention, in order to make a plurality of light sources project a fringe field onto the surface of the inspection object according to a predetermined interval and shape, the plurality of light sources may be distributed on a surface forming a sphere, a cuboid, a side surface of a cylinder, a cone, etc. in a three-dimensional coordinate system. The predetermined interval and the predetermined shape as described above are examples of the present disclosure, which includes the configuration of light sources for forming a fringe field on the surface of the inspection object. As described above, in the case where the lighting system includes a plurality of light sources according to a predetermined shape and a predetermined interval, the lighting pattern required for inspecting the inspection object can be formed with a minimum number, thereby having the effect of saving costs.

[0108] The lighting system 200 of one embodiment of the present disclosure includes a plurality of light sources, and different lighting patterns can be transmitted by controlling the actions of the plurality of light sources. Specifically, any one of the plurality of light sources in the lighting system 211a, 221a, 211b, 221b can be individually turned on and off. This means that a single light source can be controlled independently of the configuration of the light source, and further, in addition to the standardized pattern (e.g., polygon, circle, straight line, diagonal line, etc.), an atypical pattern (e.g., a curve pattern with different curvature radii) or a pattern composed of a plurality of patterns (e.g., simultaneously projecting a quadrilateral lighting pattern and a circular lighting pattern onto an object) can be formed by considering the characteristics of the surface of the object or the shape of the object, which is obvious to ordinary technicians through the content of this specification. Moreover, the individual control of the plurality of light sources as described above has the following effect: even if the shape of the object to be inspected continues to change, regardless of it, an edge field can be formed in a desired area on the surface of the object by controlling the light source of the lighting system 200, thereby inspecting the appearance defect.

[0109] The lighting system 200 of an embodiment of the present disclosure may be composed of a plurality of light sources, and the plurality of light sources may be composed of at least two groups. For example, the light source groups 210a and 220a may be formed in a straight line. Alternatively, when the light sources are connected by virtual lines, a plurality of light sources forming a circle may mutually form groups 210b and 220b. The above-mentioned group of light sources is only one embodiment, and the present disclosure is not limited thereto. As described above, when the light sources are controlled in groups, a lighting pattern of a desired shape can be projected onto an object, and it is also possible not to control each light source individually, thereby having the effect of reducing the cost for control.

[0110] Furthermore, in a defect inspection device according to an embodiment of the present disclosure, the lighting system includes a plurality of light sources, and when the light sources emit light, the number of fringe fields per unit area generated in a region on the surface of the inspection object can be different from that in other regions. In other words, the density of the light source region 11 can be different according to the region. Alternatively, for light sources in groups or for each individual light source, the control related to the plurality of light sources is made different, and a lighting pattern can be generated in such a way that an illumination change region (in the present disclosure, a fringe field) including a drastic change in illumination covers at least a portion of the surface of the inspection object.

[0111] As described above, the reason for adjusting the density of the fringe field or generating the fringe field on only a portion of the surface of the object to be inspected is that, depending on the surface characteristics of the object to be inspected, the fringe field may be required or not. For example, in the defect inspection device 100, when one area on the surface of the object is formed of a glossy material, but another area is formed of a matte material, the glossy portion forms a fringe field to improve the inspection efficiency, and on the other hand, for the other matte area, a diffuser or the like is used to transmit uniform light, or, depending on the situation, the light source may be turned off to prevent light from being transmitted to the matte area. As another example, when adjusting the density of the fringe field or generating the fringe field on only a portion of the surface of the object to be inspected, it may be necessary to further inspect a specific portion of the surface of the object to be inspected. That is, for the remaining area, defect inspection has been completed, and a defect occurs in a specific area on the surface of the object and needs to be irradiated again.

[0112] As described above, the lighting system transmits fringe fields only to a part of the area or is capable of operating the lighting system in a manner that adjusts the number of fringe fields per unit area. This has the effect of reducing the total cost associated with control costs and electricity costs at the industrial site, concentrating resources in areas that need to be concentrated, thereby improving inspection efficiency.

[0113] The lighting system 200 can be controlled to transmit a lighting pattern with different illuminances for each region to the surface of the inspection object. Hereinafter, the lighting system 200 used in the defect inspection apparatus according to an embodiment of the present disclosure will be described to describe a lighting pattern formed on the surface of the inspection object 10 .

[0114] Below, refer to Fig.18 The above-mentioned multiple light sources are arranged at predetermined intervals, and the above-mentioned illumination pattern can be transmitted to the surface of the inspection object 10 by turning on / off the intermediate light sources. In the illumination system in which multiple light sources are arranged uniformly at the same intervals, at least one light source is controlled according to a line of a linear shape, a polygonal shape, a circular shape, etc., so that the above-mentioned illumination pattern can be transmitted to the surface of the inspection object 10.

[0115] exist Fig.18In the example of , the dark area and the bright area are examples of groups of light sources used to form an illumination pattern. The illumination pattern may be a linear vertical line 410 or a horizontal line related to the inspection object that is cross-projected by the light source 411 of the first group and the light source 412 of the second group. In order to generate an illumination pattern on the surface of the inspection object, the light source 411 of the first group and the light source 412 of the second group may be controlled by the processor to mutually change the light emitting state of the light source. As an additional example of the illumination pattern, the light source 421 of the first group and the light source 422 of the second group can show a repeated pattern in the form of a diagonal line 420. Alternatively, it may also be a polygon 430, a circle 450, etc., and when the group of light sources is divided into a grid shape, an illumination pattern such as the reference numeral 440 may be irradiated to the inspection object. However, the above-mentioned examples are only examples, and the present disclosure includes the above-mentioned examples, thereby, a variety of patterns can be formed without restriction.

[0116] The width or interval of the illumination pattern as described above can be reflected on the surface of the inspection object by controlling multiple light sources individually or in groups. In addition, the width or interval of the illumination pattern can be reflected by changing the distance between the light source and the inspection object.

[0117] In the present disclosure, one of the causes of defects such as scratches, height differences, dents, etc. that can be inspected using a fringe field on a surface with a shiny, glossy material, a surface of a metal frame, or an object with a mirror surface is as follows: the defective surface as described above reflects light at a reflection angle different from that of a normal surface relative to the incident angle, and therefore, defects can be inspected by detecting light scattered with different reflection angles as described above.

[0118] Therefore, as described above, in the inspection device of the present disclosure that can form a plurality of illumination patterns with different illumination intensities according to regions, the angle of the incident light is deformed into a plurality of angles and utilized, so that different reflection angles reflected on the defect surface can be inspected at as many angles as possible. This has the effect of increasing the performance of the defect inspection device.

[0119] On the other hand, generally, defects on the surface of an object occur in random areas, and therefore, the probability of detecting a defect increases when the fringe fields are made to overlap with each other as little as possible and as much surface area as possible can be covered with the fringe fields.

[0120] Therefore, in the lighting system of the defect inspection device of an embodiment of the present disclosure, at least two different lighting patterns are transmitted to the inspection object so that at least a part of the illumination change area including the drastic change of illumination does not overlap, and for the at least two different lighting patterns, the camera unit can obtain multiple image data. Afterwards, the processor can input the multiple different image data into the deep learning-based model to check whether there are defects on the surface of the inspection object.

[0121] As described above, the fringe field is transmitted so that it can cover most of the surface area of ​​the inspection target object 10, thereby increasing the inspectable area and improving the inspection efficiency.

[0122] Below, refer to Fig.19 and Fig. 20 An exemplary structure of an illumination system used in a defect inspection device according to an embodiment of the present disclosure and an exemplary illumination pattern thereof are described.

[0123] In an embodiment of the present disclosure, the lighting system of the defect inspection device is composed of multiple lighting panels equipped with multiple light sources. The multiple lighting panels can be in the form of a punched area so that at least a portion of the light transmitted from the lighting panels located in the upper layer reaches the surface of the inspection object.

[0124] For the sake of specific example, if there are two lighting panels, there is Fig.19 (a) the upper lighting panel 510 and Fig.19 In this case, the multiple light sources 511 of the upper lighting panel and the multiple light sources 521 of the lower lighting panel are separated from each other so that they do not overlap each other when the two lighting panels are overlapped, and can be located inside the lighting panels. In addition, the area 522 corresponding to the light source of the upper lighting panel of the lower lighting panel has a punched shape so that the lower lighting panel 520 does not block the light source 511 of the upper lighting panel. Fig.19 In FIG. 5 , the cross section 513 of the upper lighting plate and the cross section 523 of the lower lighting plate are used to illustrate the cross sections when the upper lighting plate and the lower lighting plate are overlapped. Fig. 20 A side view relating to a cross section of the overlapping lighting panels is disclosed. Fig. 20 It is confirmed that the corresponding part of the lower lighting plate is formed by the punched area 522, so that the lower lighting plate does not block the light source 511 of the upper lighting plate. And, when the inspection device is used, in order to form an illumination pattern with different illumination according to the area on the surface of the inspection object, it can be used in the following form, that is, when the light source of the upper lighting plate is turned on, the lower lighting plate is turned off, and when the light source of the upper lighting plate is hung on the wall, the lower lighting plate is turned on. The lattice-shaped punching pattern shown in the above-mentioned lower lighting plate is only an example, and various patterns including linear, circular, polygonal, etc. can be manufactured without limitation by overlapping multiple lighting plates according to the deformation of the punching pattern, which is obvious to ordinary technicians.

[0125] As described above, the effect of the invention obtained by overlapping a plurality of lighting panels is characterized as follows: instead of controlling a specific light source group or controlling a plurality of light sources separately, only each lighting panel is controlled to form a lighting pattern having different illuminations according to regions on the surface of the object to be inspected. In other words, a plurality of light sources included in individual lighting panels can be uniformly controlled to form different lighting patterns, thus having the effect of low control cost and easy operation.

[0126] The processor 110 can obtain image data related to each different illumination pattern transmitted through the illumination system. For example, for an inspection object, there can be multiple image data according to different illumination patterns. The processor 110 can determine the defects of the inspection object based on the multiple image data including different illumination patterns.

[0127] The processor 110 may determine whether there are defects in the inspection object using image data including the surface of the inspection object. The processor 110 processes the image data related to the inspection object using a model based on deep learning, thereby determining whether there are defects in the inspection object. The processor 110 inputs the image data to the model based on deep learning, and may determine whether there are defects in the inspection object using the model based on deep learning. In addition, the processor 110 may include the result of classifying defects existing on the surface of the inspection object from the image data using the model based on deep learning.

[0128] Fig.21 1 is an example diagram of the processor 110 detecting a defect existing in the inspection object 10 from image data including the surface of the inspection object 10 and marking the defect position. In the case where the inspection object 10 has a defect, the processor 110 not only outputs the determination of the existence of the defect but also outputs the defect position that serves as the basis for the determination. Fig.21 In part (a), when the surface of the inspection object 10 has a defect 1 in the form of a scratch, as shown in FIG. Fig.21 In part (b), the processor 110 may utilize a deep learning-based model to distinguish the defect 1 location from other normal parts using visual expressions (e.g., color, etc.).

[0129] In one embodiment of the present disclosure, the processor 110 may extract only the edge field portion from the image data and input it to the deep learning-based model. The processor 110 may divide the image data into patches and extract the edge field based on the pixel values ​​of the patches. For example, the processor 110 obtains the illuminance value of each pixel by referring to the RGB value of each pixel from the acquired image data. This can obtain the absolute value by comparing the actual RGB value with the color table, and the average and dispersion of the RGB value of the entire pixel can be derived to obtain the relative value. In addition, the processor 110 can extract only the edge field portion from the image data based on the meta-information representing the edge field portion.

[0130] In order to filter out multiple pixels that become the center of the fringe field based on the illumination value of each pixel mentioned above, when the pixel located at the exact center of the 3×3 matrix has an illumination value higher than that of the surrounding pixels, the processor 110 can use the pixel that becomes the center of the fringe field to filter it. In addition, such a pixel can be used as the center of the light source, and the fringe field can be extracted using a preset distance of the fringe field. In another embodiment, the pixel filtering process that becomes the center of the fringe field mentioned above can also be performed by the following process, that is, on the two-dimensional image data, the illumination value of each pixel is used as an additional axis, and the maximum value of the illumination value is found for filtering. In addition, instead of the maximum value, pixels with illumination values ​​that meet a specific threshold can also be filtered, and the maximum value can be selected in a wider N×N matrix.

[0131] In another embodiment, after extracting a plurality of fringe field patches, the processor 110 compares the areas included in each patch, and when the overlapping area of ​​the patches is larger than a predetermined area, only a portion of the patches may be selected.

[0132] In one embodiment of the present disclosure, the processor 110 may input image data and metadata to the deep learning-based model, and in the metadata, in order to emphasize the part where the drastic change in illumination occurs, information related to the part where the drastic change in illumination occurs may be included. The processor 110 may input the position information and image data of the edge field to the deep learning-based model. Specifically, the processor 110 may obtain the illumination value of each pixel by referring to the RGB value of each pixel from the acquired image data. In addition, the processor may calculate the difference in illumination value between pixels separated by any distance in a direction orthogonal to the boundary line direction of the illumination pattern transmitted to the inspection object. If the illumination pattern is a circle 450, the difference in illumination value between pixels separated by a prescribed distance along a line extending radially from the center of the light source area located at the center of the image data may be calculated. The difference in illumination value may be a simple subtraction, or may be the difference in illumination value between pixels divided by the slope of the prescribed distance. The processor may screen pixels in an area showing a drastic change in illumination based on the difference in illumination value. In addition, the pixel information and the image data may be input to the deep learning model. The above-mentioned lighting patterns and the orthogonal directions thereof are merely illustrative, and the present disclosure is not limited thereto.

[0133] As described above, when only the edge field portion is extracted from the image data and input into a deep learning-based model, or when metadata emphasizing an area showing a drastic change in illumination and the above-mentioned image data are input into a deep learning-based model, the processor does not need to explore the entire image data, but only needs to perform operations to determine whether there is a defect or not on the extracted edge field or near the emphasized pixel. Therefore, it has the advantages of consuming less computing resources and increasing the computing processing speed.

[0134] In one embodiment of the present disclosure, in order to cover an area larger than a specified width in the entire area of ​​the surface of the inspection object with the fringe field, the processor 110 may perform additional operations. That is, when the sum of the areas of multiple different areas covered by the fringe field in the total area of ​​the surface of the inspection object is less than a threshold value, in order to generate a fringe field in the previously uncovered area by controlling the lighting system, the processor 110 may transmit an illumination pattern for inducing a drastic change in illumination to other areas of the surface of the inspection object. For example, in the case where the threshold set to 100% is set in order to inspect the entire surface of the inspection object, if it is determined that the sum of the areas of the fringe field areas extracted after performing the operation of extracting the fringe field cannot cover the entire surface of the inspection object, the processor 110 controls the lighting system to transmit the illumination pattern to different areas, and may receive image data from the imaging unit 140 again to extract the fringe field again. The re-input and re-extraction process described above may be performed multiple times according to the set threshold value. As another example, when controlling the illumination system in such a way as to transmit an illumination pattern different from a previous pattern, the processor may control the illumination system so as to have a drastic change in illumination at other areas of the inspection target object surface than the area covered with the fringe field.

[0135] Generally, appearance defects occur in any area of ​​the surface of the object to be inspected. Therefore, as described above, the processor 110 has the advantage that if at least a coverage threshold can be detected on the surface of the object to be inspected, the probability of detecting surface defects can be increased.

[0136] In one embodiment of the present disclosure, the processor 110 transmits different illumination patterns to the inspection target object to obtain multiple image data, which can be input into a deep learning-based model to determine defects on the surface of the inspection target object. In a specific embodiment, after the processor 110 overlaps multiple image data according to each pixel and generates one image data, it can be input into the deep learning-based model. In another embodiment, after multiple image data are sequentially input into the same model and the positions with defects are sequentially confirmed, they can all be finally displayed in one output image. In an additional embodiment, among the multiple image data, image data including illumination patterns corresponding to sub-models learned according to each illumination pattern are input, thereby, the defect positions under each illumination pattern can be additionally output.

[0137] Depending on the surface of the object to be inspected, when the incident angle of the transmitted light forms a specific angle, surface defects can be better detected. For example, if the surface of the object to be inspected is a material including a plurality of circular grooves with very narrow and thin widths and a glossy material, a circular lighting pattern is more effective for detecting scratches on such a surface. Therefore, as described above, if defects on the surface of the object to be inspected are inspected using different lighting patterns, the incident angle of the light can be changed to be different to inspect the defects in appearance, thereby increasing the probability of detection. In addition, when the processor inspects by overlapping multiple image data according to different lighting patterns, the inspection speed is increased. On the contrary, if output data is obtained according to an additional model or by sequentially inputting multiple image data into the same model, the accuracy of defect inspection can be improved.

[0138] As described above, the processor 110 inputs image data of the inspection object including the edge field into the deep learning-based model to obtain output having multiple forms. Thus, regardless of the complexity of the image, the present disclosure can ensure the visibility of surface defects of the inspection object and increase the defect recognition rate.

[0139] The processor 110 included in the defect inspection device of the embodiment of the present disclosure may include a process of re-extracting the illumination change area including the drastic change in illumination from the image data. The pre-processing process of extracting the illumination change area from the input image data as described above can reduce the error of the processor in judging the existence of defects from the area outside the object in the image, that is, the background image, and only extract the focus area from the entire image data to perform inspection, so that the data processing speed can be increased.

[0140] Throughout this specification, models based on deep learning, computational models, neural networks, network functions, and neural networks can be used with the same meaning. A neural network can be composed of a collection of connected computing units, usually referred to as nodes. Such nodes can also be referred to as neurons. A neural network includes at least one node. The nodes (or neurons) constituting a neural network can be connected by more than one link.

[0141] In a neural network, one or more nodes connected by a link can relatively form a relationship of input node and output node. The concepts of input node and output node are relative, and any node that is in an output node relationship with a node can be in an input node relationship with other nodes, and vice versa. As described above, the relationship between input nodes and output nodes can be generated with links as the center. An input node can be connected to more than one output node through a link, and vice versa.

[0142] In a relationship between an input node and an output node connected by a link, the data of the output node can determine its value based on the data input to the input node. The link connecting the input node to the output node can have a weight. The weight is variable and can be changed by the user or the algorithm in order to perform the desired function of the neural network. For example, in the case where an output node is connected to more than one input node by respective links, the output node can determine the output node value based on the value input to the input node connected to the above output node and the weight set in the link corresponding to each input node.

[0143] As described above, in a neural network, one or more nodes are connected by one or more links to form an input node and an output node relationship in the neural network. In a neural network, the characteristics of the neural network can be determined according to the number of nodes and links, the correlation between the nodes and the links, and the value of the weighted values ​​assigned to the plurality of links. For example, in the case where there are two neural networks having the same number of nodes and links but different weighted values ​​of the links, the two neural networks can be identified as being different from each other.

[0144] A neural network may be composed of a collection of more than one nodes. A partial collection of nodes constituting a neural network may constitute a layer. A portion of the nodes constituting a neural network may constitute a layer based on the distance from an initial input node. For example, a collection of nodes whose distance from an initial input node is n may constitute n layers. The distance from the initial input node may be defined by the minimum number of links required to reach the corresponding node from the initial input node. However, this definition of layers is arbitrarily defined for the sake of explanation, and the order of layers in a neural network may be defined by a method different from that described above. For example, a layer of a node may also be defined by the distance from a final output node.

[0145] The initial input node may mean one or more nodes in the neural network that directly input data without linking the relationship with other nodes. Alternatively, it may mean a node that does not have other input nodes connected by links in the relationship between nodes based on links in the neural network. Similarly, the final output node may mean one or more nodes that do not have output nodes in the relationship with other nodes in the neural network. Furthermore, the hidden node may mean a node that constitutes the neural network that is not an initial input node or a final output node.

[0146] The neural network of one embodiment of the present disclosure may be a neural network of the following form, that is, the number of nodes in the input layer may be the same as the number of nodes in the output layer, and as moving from the input layer to the hidden layer, the number of nodes first decreases and then increases again. Furthermore, the neural network of yet another embodiment of the present disclosure may be a neural network of the following form, that is, the number of nodes in the input layer may be less than the number of nodes in the output layer, and as moving from the input layer to the hidden layer, the number of nodes decreases. Furthermore, the neural network of another embodiment of the present disclosure may be a neural network of the following form, that is, the number of nodes in the input layer may be more than the number of nodes in the output layer, and as moving from the input layer to the hidden layer, the number of nodes increases. The neural network of yet another embodiment of the present disclosure may be a neural network of the form of a combination of the neural networks described above.

[0147] A deep neural network (DNN) may mean a neural network including multiple hidden layers in addition to an input layer and an output layer. If a deep neural network is used, the latent structure of the data can be grasped. That is, the latent structure of photos, texts, videos, sounds, and music (for example, what is the object in the photo, what is the content and emotion of the text, what is the content and emotion of the sound, etc.) can be grasped. A deep neural network may include a convolutional neural network (CNN), a recurrent neural network (RNN), an auto encoder, a generative adversarial network (GAN), a restricted Boltzmann machine (RBM), a deep belief network (DBN), a Q network, a U network, a Siamese network, etc. The above-mentioned deep neural network is recorded only as an example, and the present disclosure is not limited thereto.

[0148] In one embodiment of the present disclosure, the network function may further include an autoencoder. The autoencoder may be a type of artificial neural network for outputting output data similar to the input data. The autoencoder may include at least one hidden layer, and an odd number of hidden layers may be configured between the input layer and the output layer. The number of nodes in each layer may be reduced to an intermediate layer called a bottleneck layer (encoding) when the number of nodes in the input layer is reduced, and may be symmetrical with the reduction and expanded to the output layer (symmetrical with the input layer) at the bottleneck layer. The autoencoder may perform nonlinear dimensionality reduction. The number of input layers and output layers may correspond to the dimension after preprocessing of the input data. In the autoencoder structure, the number of hidden layer nodes included in the encoder may have a structure that decreases as the distance from the input layer increases. In the case where the number of nodes in the bottleneck layer (the layer with the least nodes between the encoder and the decoder) is too small, sufficient information may not be transmitted, and it may also be more than a specific number (for example, more than half of the input layer, etc.).

[0149] The neural network may learn by at least one of supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning. Learning of the neural network may be a process of applying knowledge used by the neural network to perform a specific action to the neural network.

[0150] A neural network can learn in a direction that minimizes the error of the output. The process of learning a neural network is a process that repeatedly inputs learning data to the neural network, calculates the error between the output of the neural network related to the learning data and the target, backpropagates the error of the neural network from the output layer to the input layer of the neural network in the direction for reducing the error, and updates the weighted values ​​of each node of the neural network. In the case of supervised learning, learning data (i.e., labeled learning data) with the correct answer marked in each learning data is used, and in the case of unsupervised learning, the correct answer may not be marked in each learning data. That is, for example, the learning data in the case of supervised learning related to data classification may be data with a category marked in each learning data. The labeled learning data is input to the neural network, and the output (category) of the neural network is compared with the label of the learning data, thereby calculating the error. As another example, in the case of unsupervised learning related to data classification, the error can be calculated by comparing the learning data as input with the output of the neural network. The calculated error can be backpropagated in the reverse direction (i.e., from the output layer to the input layer) of the neural network, and the connection weighted values ​​of each node of each layer of the neural network can be updated according to the backpropagation. The amount of change in the connection weight value of each node updated can be determined according to the learning rate. The calculation of the neural network related to the input data and the back propagation of the error can constitute a learning cycle (epoch). The learning rate can be applied differently according to the number of repetitions of the learning cycle of the neural network. For example, a high learning rate is used in the early stage of learning of the neural network so that the neural network quickly ensures the performance of a specified level to improve efficiency. In the later stage of learning, a low learning rate can be used to improve accuracy.

[0151] In the process of learning a neural network, usually, the learning data can be a partial set of actual data (i.e., data to be processed by the learned neural network), so the error related to the learning data can be reduced, but the learning cycle with increased errors with actual data. Overfitting is the phenomenon that the error related to the actual data increases due to excessive learning of learning data as described above. For example, the phenomenon that a neural network that learns cats by showing yellow cats does not recognize cats other than yellow as cats is a kind of overfitting. Overfitting can be a cause of increasing the error of the machine learning algorithm. In order to prevent such overfitting, a variety of optimization methods can be used. In order to prevent overfitting, methods such as increasing learning data or regularization, disabling signal loss (dropout) of a part of the network nodes during the learning process, and using a batch normalization layer can be applied.

[0152] Fig. 22 is a simplified, general diagram relating to an exemplary computing environment in which embodiments of the present disclosure may be implemented.

[0153] The present disclosure is generally described as being embodied by a computing device; however, a person of ordinary skill in the art will appreciate that the present disclosure may be combined with computer executable instructions and / or other program modules that may be executed on more than one computer and / or may be embodied by a combination of hardware and software.

[0154] Generally, program modules include routines, programs, components, data structures, and others that perform specific tasks or embody specific abstract data types. Moreover, as long as a person of ordinary skill in the art can know, the method of the present disclosure can be implemented not only by a single processor or multi-processor computing system, a small computer, or a large computer structure, but also by a personal computer, a handheld computing device, a microprocessor-based or programmable home appliance, and others (these can be connected to more than one related device to act).

[0155] The described embodiments of the present disclosure may also be practiced in a distributed computing environment where certain tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote memory storage devices.

[0156] Typically, a computer includes a variety of computer-readable media. Any medium accessible to a computer may be a computer-readable medium, including volatile and non-volatile media, transitory and non-transitory media, removable and non-removable media. As a non-limiting example, a computer-readable medium may include a computer-readable storage medium and a computer-readable transmission medium. Computer-readable storage media include volatile and non-volatile media, transitory and non-transitory media, removable and non-removable media embodied in any method or technology for storing information such as computer-readable instructions, data structures, program modules or other data. Computer-readable storage media include random access memory, read-only memory, electrically erasable programmable read-only memory, flash memory or other storage technology, CD-ROM, digital versatile disc (DVD, digital video disk) or other optical disk storage device, tape cassette, tape, magnetic disk storage device or other magnetic storage device or any other medium that can be accessed by a computer and used to store the desired information, but is not limited thereto.

[0157] Typically, computer-readable transmission media embody computer-readable instructions, data structures, program modules or other data in a modulated data signal such as a carrier wave or other transport mechanism and includes all information delivery media. The term "modulated data signal" refers to a signal that is arranged in a manner to encode information in the signal or one or more of its signal characteristics. As non-limiting examples, computer-readable transmission media include wired media such as a wired network or a direct-wired connection and wireless media such as audio, radio frequency (RF), infrared, and other wireless media. Any combination of the above media is also included in the scope of computer-readable transmission media.

[0158] An exemplary environment 1100 for embodying various aspects of the present disclosure is shown, including a computer 1102, which includes a processing device 1104, a system memory 1106, and a system bus 1108. The system bus 1108 connects system components including the system memory 1106 (but not limited to this) to the processing device 1104. The processing device 1104 can be one of a variety of commercially available processors. Dual processors and other multi-processor structures can also be used as the processing device 1104.

[0159] The system bus 1108 may be one of several types of bus structures that can be connected in addition to a local bus using any of a memory bus, a peripheral device bus, and a variety of commercial bus structures. The system memory 1106 includes a read-only memory 1110 and a random access memory 1112. The basic input-output system (BIOS) is stored in a non-volatile memory 1110 such as a read-only memory, an erasable programmable read-only memory (EPROM), or an electrically erasable programmable read-only memory. The basic input-output system includes basic routines that help transfer information between structural elements in the computer 1102 while starting. The random access memory 1112 may also include a high-speed random access memory such as a static random access memory for caching data.

[0160] The computer 1102 also includes an internal hard disk drive (HDD) 1114 (e.g., enhanced integrated drive electronics (EIDE), serial advanced technology attachment (SATA)), a disk drive (FDD) 1116 (e.g., reads from or writes to a removable disk 1118), and an optical drive 1120 (e.g., reads from a CD-ROM disk 1122 or reads from or writes to other high-capacity optical media such as a digital versatile disk). The internal hard disk drive 1114 can also be configured for external use in a suitable chassis (not shown). The hard disk drive 1114, the disk drive 1116, and the optical drive 1120 can be connected to the system bus 1108 via a hard disk drive interface 1124, a disk drive interface 1126, and an optical drive interface 1128, respectively. The interface 1124 for embodying an external drive includes at least one or both of a universal serial bus (USB) and an IEEE 1394 interface technology.

[0161] These drives and computer-readable media associated therewith provide non-volatile storage of data, data structures, computer-executable instructions, and the like. In the case of computer 1102, the drives and media correspond to storing any data in an appropriate digital form. In the above, the description related to computer-readable media mentions hard disk drives, removable disks, and removable optical media such as optical disks or digital versatile disks. As long as a person of ordinary skill in the art knows, other types of computer-readable media such as zip drives, magnetic tape cartridges, flash memory cards, cassette tape drives, and the like can also be used in the exemplary operating environment, and any of the above media may include computer-executable instructions for performing the methods disclosed herein.

[0162] A plurality of program modules including an operating system 11130, one or more application programs 1132, other program modules 1134, and program data 1136 may be stored in the drives and random access memory 1112. All or a portion of the operating system, application programs, modules, and / or data may also be cached in random access memory 1112. As is well known, the present disclosure may be embodied in various commercially available operating systems or a combination of multiple operating systems.

[0163] The user can input commands and information to the computer 1102 through one or more wired / wireless input devices, such as a keyboard 1138 and a pointing device such as a mouse 1140. Other input devices (not shown) include microphones, infrared (IR) remote controls, joysticks, game pads, styluses, touch screens, and the like. These and other input devices are usually connected to the processing unit 1104 through an input device interface 1142 connected to the system bus 1108, and may also be connected through other interfaces such as a parallel port, an IEEE 1394 serial port, a game port, a universal serial bus port, an infrared interface, and the like.

[0164] A display 1144 or other type of display device may also be connected to the system bus 1108 via an interface such as a display adapter 1146. In addition to the display 1144, computers typically include other peripheral output devices (not shown) such as speakers, printers, etc.

[0165] The computer 1102 can operate in a networked environment by using logical connections to one or more remote computers, such as remote computer(s) 1148, via wired and / or wireless communications. The remote computer(s) 1148 can be a workstation, computing device computer, router, personal computer, portable computer, microprocessor-based entertainment device, peer device, or other conventional network node, and typically includes most or all of the structural elements described for the computer 1102, with only a memory storage device 1150 being shown for simplicity. The illustrated logical connections include wired / wireless connections in a local area network 1152 and / or a larger network, such as a wide area network 1154. Such local area network and wide area network networking environments are common in offices and companies, facilitating enterprise-wide computer networks such as intranets, which can all be connected to a global computer network, such as the Internet.

[0166] When used in a LAN networking environment, the computer 1102 is connected to the local network 1152 through a wired and / or wireless communication network interface or adapter 1156. The adapter 1156 can facilitate wired or wireless communication with the LAN 1152, and the LAN 1152 includes a wireless access point disposed thereon, thereby communicating with the wireless adapter 1156. When used in a WAN networking environment, the computer 1102 may include a modem 1158 or be connected to a communication computing device in the WAN 1154, or have other means for setting up communication through the WAN 1154, such as the Internet. The modem 1158, which may be internal or external and wired or wireless, may be connected to the system bus 1108 via a serial port interface 1142. In a networked environment, the various program modules described for the computer 1102 or portions thereof may be stored in the remote memory / storage device 1150. The illustrated network connections are merely exemplary, and other means for setting up a communication link between computers may be used as is well known.

[0167] The computer 1102 communicates with any wireless device or entity configured to operate via wireless communication, such as printers, scanners, desktop and / or portable computers, portable data assistants (PDAs), communication satellites, any device or location associated with wirelessly detectable tags, and telephones. This includes at least Wireless Fidelity (Wi-Fi) and Bluetooth wireless technologies. Thus, the communication can be a predefined structure or simply ad hoc communication between at least two devices as in conventional networks.

[0168] Wireless Fidelity (Wi-Fi) allows connections to the Internet, etc. without wires. Wireless Fidelity is a wireless technology that allows devices such as computers, such as mobile phones, to transmit and receive data indoors and outdoors, that is, anywhere within the communication range of a base station. Wireless Fidelity networks can be secure and reliable, and use wireless technologies called IEEE 802.11 (a, b, g, others) to provide high-speed wireless connections. Wireless Fidelity can be used to connect computers and to the Internet and wired networks (using IEEE 802.3 or extranets). Wireless Fidelity networks operate in the unlicensed 2.4 and 5 GHz wireless bands at data rates such as 11 Mbps (802.11a) or 54 Mbps (802.11b), or can operate in products that include two bands (dual-band).

[0169] It will be understood by those skilled in the art that any two or more different technologies and methods may be used to express information and signals. For example, the data, instructions, commands, information, signals, bits, symbols, and chips cited above may be expressed by voltage, current, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any combination thereof.

[0170] It will be understood by those skilled in the art of the present disclosure that the various exemplary logic blocks, modules, processors, units, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be embodied in electronic hardware, (referred to herein as software for ease of description) various forms of programs or design codes, or combinations thereof. In order to clearly illustrate this interchangeability of hardware and software, various exemplary components, blocks, modules, circuits, and steps and their functions have been described above. Whether such functions are embodied in hardware or software depends on the design constraints imposed on the specific application and the entire system. Those skilled in the art of the present disclosure may embody the functions described in various ways for each specific application, but such determination is within the scope of the present disclosure.

[0171] The various embodiments disclosed herein may be embodied as methods, devices, or articles of manufacture that are programmed and / or engineered using standard techniques. The term "article of manufacture" includes a computer program, carrier wave, or medium that can be accessed from any computer-readable storage device. For example, computer-readable storage media include magnetic storage devices (e.g., hard disks, floppy disks, magnetic strips, etc.), optical disks (e.g., optical disks, digital versatile disks, etc.), smart cards, and flash memory (e.g., electrically erasable programmable read-only memories, cards, strips, key drives, etc.), but are not limited thereto. Furthermore, the various storage media proposed herein include one or more devices and / or other machine-readable media for storing information.

[0172] It should be understood that the specific order or hierarchy of steps in the processor presented is an example of an exemplary approach. Based on design priorities, the specific order or hierarchy of steps in the processor may be rearranged within the scope of the present disclosure. The claimed scope provides elements of various steps in a sample order, but is not meant to be limited to the specific order or hierarchy presented.

[0173] Descriptions related to the proposed embodiments are provided so that any person of ordinary skill in the art of the present disclosure may utilize or implement the present disclosure. Various variations related to such embodiments will be apparent to a person of ordinary skill in the art of the present disclosure, and the general principles defined herein may be applied to other embodiments without exceeding the scope of the present disclosure. Therefore, the present disclosure is not limited to the embodiments disclosed herein and must be interpreted in the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A defect inspection device, characterized in that: include: The lighting system is designed to transmit a lighting pattern having different illumination intensities according to regions to the surface of the object to be inspected; A photographing unit, used for acquiring image data of the inspection object; One or more processors for processing the image data; and Memory, which stores deep learning-based models, In one or more of the above processors, By controlling the lighting system, the lighting system transmits a lighting pattern having different illumination intensities according to regions to the surface of the inspection object. The image data acquired by the camera unit is input into the deep learning-based model, the image data including at least one fringe field generated in at least one region of the surface of the object due to the transmitted illumination pattern, the fringe field being a region within a preset distance from the center of the light source region when a light source image is formed on the surface of the inspection object, The defects on the surface of the inspection object are determined using the deep learning-based model and based on the edge field.

2. The defect inspection device according to claim 1, characterized in that: The lighting system includes a plurality of light sources, and the plurality of light sources are arranged at predetermined intervals.

3. The defect inspection device according to claim 1, characterized in that: The illumination system includes a plurality of light sources, and an illumination pattern having different illumination intensities according to regions is transmitted to the surface of the object to be inspected by controlling the operation of each of the plurality of light sources.

4. The defect inspection device according to claim 1, characterized in that: The above-mentioned lighting system includes a plurality of light sources composed of at least two combinations.

5. The defect inspection device according to claim 1, characterized in that: The illumination system includes a plurality of light sources, and when the plurality of light sources emit light, the number of fringe fields per unit area included in a region on the surface of the inspection target object is at least different from that in other regions.

6. The defect inspection device according to claim 1, characterized in that: The illumination pattern is configured such that an illumination change region where the illumination changes drastically covers at least a portion of the surface of the inspection target object.

7. The defect inspection device according to claim 1, characterized in that: The processor controls the lighting system so that the lighting system transmits a lighting pattern in a linear, polygonal or circular pattern.

8. The defect inspection device according to claim 1, characterized in that: The processor determines defects on the surface of the inspection target object by inputting a plurality of image data acquired under different lighting patterns into a deep learning-based model.

9. The defect inspection device according to claim 1, characterized in that: The lighting system is composed of a plurality of lighting panels equipped with a plurality of light sources. The lighting plate is formed in a region punched so that at least a portion of light transmitted through the lighting plate located in the upper layer reaches the surface of the inspection target object.

10. The defect inspection device according to claim 1, characterized in that: The processor further extracts, from the image data, an illumination change region where the illumination changes dramatically.

11. A computer-readable storage medium storing a computer program, characterized in that: When the above computer program is executed on one or more processors, the following actions are performed to check for defects: The above actions include: Controlling the lighting system so that the lighting system designed to transmit different lighting patterns transmits different lighting patterns according to regions to the surface of the object to be inspected; Inputting image data acquired by a camera into a deep learning-based model, the image data including at least one fringe field generated in at least one region of a surface of an object due to the transmitted illumination pattern; and Using the deep learning-based model and judging the defects on the surface of the inspection object according to the edge field, The fringe field is a region within a predetermined distance from the center of the light source region when a light source image is formed on the surface of the inspection target object.

12. A method for inspecting whether a surface of an object to be inspected has defects, which is performed in a defect inspection device, characterized in that: The steps include: Controlling the lighting system so that the lighting system designed to transmit different lighting patterns transmits different lighting patterns according to regions to the surface of the object to be inspected; Inputting image data acquired by a camera into a deep learning-based model, the image data including at least one fringe field generated in at least one region of a surface of an object due to the transmitted illumination pattern; and Using the deep learning-based model and judging the defects on the surface of the inspection object according to the edge field, The fringe field is a region within a predetermined distance from the center of the light source region when a light source image is formed on the surface of the inspection target object.

Images