Pinhole inspection system and method

Abstract

The present invention relates to a pinhole inspection system including an inspection device including an illumination, a camera, and a visual processing unit, the visual processing unit including: a first algorithm unit that performs a first feature extraction using a preset first algorithm, the first feature extraction specifying a plurality of first candidate regions in which estimated pinholes are located; a second algorithm unit that performs, using a preset second algorithm, a second feature extraction in which a plurality of second candidate regions in which estimated pinholes are located are specified, and a third feature extraction in which a plurality of second candidate regions in which the estimated pinholes are located are specified; designating a plurality of candidate regions screened by comparing the plurality of first candidate regions with the plurality of second candidate regions as a plurality of third candidate regions in which estimated pinholes are located; and a third algorithm unit for determining whether the pinholes are actually located in the plurality of third candidate regions by using a preset third algorithm.

Description

Technical Field

[0001] The invention relates to a pinhole inspection system and method. Background Art

[0002] Recently, as the world implements carbon reduction policies to combat climate change, the era of internal combustion engine vehicles is rapidly changing to the era of electric vehicles. As the downstream industry changes, the market for electric vehicle batteries will grow, and as a result, the demand for copper thin films as secondary battery electrode materials will increase significantly.

[0003] On the other hand, when preparing copper foil film, pinholes, which are tiny holes that penetrate the copper foil film, frequently occur during the preparation process of the copper foil film. In the past, workers manually identified the pinhole defects with their naked eyes to judge the defects of the copper foil film caused by the pinhole defects. For this existing pinhole inspection method, each worker has a different inspection standard, and due to repeated operations, the workers often feel dizzy and their recognition ability decreases. In this regard, according to the existing pinhole inspection method, the detection rate of pinholes and the reliability of the pinhole inspection results are reduced, so there is a problem that it is difficult to judge the defects of the copper foil film caused by pinhole defects. Summary of the invention

[0004] The present invention is used to solve the problems of the prior art. The purpose of the present invention is to provide a pinhole inspection system and method that are improved to automatically inspect pinholes in a copper foil film.

[0005] Furthermore, an object of the present invention is to provide a pinhole inspection system and method which are improved so as to accurately determine the position and size of the pinhole.

[0006] Furthermore, an object of the present invention is to provide a pinhole inspection system and method that are improved to eliminate various noises or errors generated during the pinhole inspection process.

[0007] A pinhole inspection system of a preferred embodiment of the present invention for solving the above-mentioned problem comprises an inspection device, wherein the inspection device comprises: an illumination device for irradiating inspection light to one side of an inspection object; a camera device for photographing the other side of the inspection object opposite to the one side to generate a photographed image of the inspection object; and a visual processing unit for analyzing the state of the inspection light penetrating the inspection object based on the photographed image to implement pinhole inspection on the inspection object, wherein the visual processing unit comprises: a first algorithm unit for implementing a first feature extraction using a preset first algorithm, in which a plurality of first candidate areas where the pinholes are estimated to be located are specified in the first feature extraction; a second algorithm unit for implementing a second feature extraction and a third feature extraction using a preset second algorithm, in which a plurality of second candidate areas where the pinholes are estimated to be located are specified in the second feature extraction, in which a plurality of candidate areas selected by comparing the plurality of the first candidate areas with the plurality of the second candidate areas are specified as a plurality of third candidate areas where the pinholes are estimated to be located in the third feature extraction; and a third algorithm unit for determining whether the pinholes are actually located in the plurality of the third candidate areas using a preset third algorithm.

[0008] The present invention relates to a pinhole inspection system and method, and the present invention has the following effects.

[0009] First, the present invention can utilize a preset algorithm to automatically and constantly perform pinhole inspection on the inspection object, which can reduce the time required for the pinhole inspection and improve the accuracy of the pinhole inspection.

[0010] Second, the present invention implements pinhole inspection by detecting the transmitted light area generated when the inspection light is transmitted through the inspection object in the dark environment of a darkroom. Thus, the present invention can suppress the noise caused by the surface reflection of the inspection object, improve the recognition of pinholes and other surface defects of the inspection object, reduce the complexity of the captured image and other images of the inspection object, and thus further improve the accuracy of pinhole inspection.

[0011] Thirdly, the present invention removes the diffusion area formed by the inspection light diffusing through the pinhole and the reflection area formed by the inspection light passing through the pinhole being reflected from the camera lens from the transmitted light area, thereby further improving the accuracy of the pinhole inspection.

[0012] Fourthly, the present invention analyzes and infers multiple candidate areas where pinholes are located according to a statistical method and removes ineffective candidate areas from the multiple candidate areas, thereby further improving the accuracy of pinhole inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A diagram for explaining a control system of a pinhole inspection system according to a preferred embodiment of the present invention.

[0014] Figure 2It is a top view showing the schematic structure of the transfer device.

[0015] Figure 3 It is a side view showing the schematic structure of the transfer device and the inspection device.

[0016] Figure 4 A top view of the tray.

[0017] Figure 5 It is a cross-sectional view showing a state in which the pallet is loaded with or unloaded with the inspection object.

[0018] Figure 6 This is a diagram showing a state where the inspection object is loaded on the tray at the loading position.

[0019] Figure 7 This is a diagram showing a state where the inspection object is moved to the inspection position along with the tray.

[0020] Figure 8 This is a diagram showing a state where the inspection object is moved to the unloading position along with the tray.

[0021] Fig. 9 It is a block diagram for explaining the control system of the inspection device.

[0022] Fig.10 This is a diagram showing a state where the inspection object is brought into the inner space of the darkroom along with the tray.

[0023] Fig.11 This is a diagram showing a state where a pinhole inspection is performed on a pallet in the inner space of a darkroom.

[0024] Fig.12 A diagram for explaining a control system of a vision processor.

[0025] Fig.13 A diagram for explaining a control system of a first algorithm unit.

[0026] Fig.14 is a diagram showing a first input image.

[0027] Fig.15 The flowchart is a flowchart showing a pixel value distribution state of a plurality of pixels included in a first input image.

[0028] Fig.16 A diagram for explaining a method of generating a first feature extraction result image by binarizing a plurality of pixels based on a first threshold.

[0029] Fig.17 A diagram for explaining a control system of a second algorithm unit.

[0030] Fig.18 is a diagram showing a second input image.

[0031] Fig.19 A diagram for explaining a diffusion region.

[0032] Fig. 20 A diagram for explaining a method of generating a second feature extraction result image by binarizing a plurality of pixels based on a second threshold.

[0033] Fig.21 is a diagram showing a third input image.

[0034] Fig. 22 It is a diagram for explaining a method of canceling the designation of the first candidate region designated with the diffusion region as the target.

[0035] Figure 23 to Figure 25 A diagram for explaining a method of canceling the designation of the first candidate area and the second candidate area designated with the reflection area as the target.

[0036] Fig.26 A diagram for explaining a method of generating a third feature extraction result image by specifying a third candidate region.

[0037] Fig. 27 A diagram for explaining a control system of a third algorithm unit.

[0038] Fig.28 A diagram for explaining a method of generating a pinhole inspection result image by specifying a final candidate region.

[0039] Description of Reference Numerals

[0040] 1: Pinhole inspection system

[0041] 10: Controller

[0042] 20: Transfer device

[0043] 22: Pallet 24: Transfer track

[0044] 26: Shuttle

[0045] 30: Check the device

[0046] 31: Darkroom 32: Elevator

[0047] 33: Lighting 34: Camera

[0048] 35: Visual Processor

[0049] 36: Camera transfer machine

[0050] P: Inspection object Hp: Pinhole

[0051] A1: Loading position A2: Unloading position

[0052] A3: Check the location

[0053] Li: Check light

[0054] I1: The first input image I1a: The first feature extraction result image

[0055] I2: Second input image I2a: Second feature extraction result image

[0056] I3: The third input image I3a: The third feature extraction result image

[0057] I4: Pinhole inspection result image

[0058] Ac1: First candidate area Ac2: Second candidate area

[0059] Ac3: The third candidate area Ac4: The final candidate area

[0060] At: Transmitted light area An: Normal area

[0061] Ap: Pinhole area Ad: Diffusion area

[0062] Ar: Reflection area DETAILED DESCRIPTION

[0063] Figure 1 A diagram for explaining a control system of a pinhole inspection system according to a preferred embodiment of the present invention.

[0064] Reference Figure 1 The pinhole inspection system 1 of the preferred embodiment of the present invention is used to perform pinhole inspection on the inspection object P, and may include a controller 10, a transfer device 20, an inspection device 30, and the like.

[0065] Preferably, the inspection object P for which the pinhole inspection system 1 can be used to implement the pinhole inspection is a copper foil film as an electrode material of a secondary battery, but it is not limited thereto. That is, in addition to the copper foil film, the inspection object P can also be one of a variety of foil films that can cause pinhole defects.

[0066] First, the controller 10 is used to control various devices provided in the pinhole inspection system 1 .

[0067] The controller 10 may be a generally used personal computer, other devices for controlling various devices, and a detailed description thereof will be omitted.

[0068] Figure 2 A top view showing a simplified structure of the transfer device is shown below. Figure 3 A side view showing a simplified structure of a transfer device and an inspection device is shown in FIG. Figure 4 A top view of the tray. Figure 5 It is a cross-sectional view showing a state in which the pallet is loaded with or unloaded with the inspection object.

[0069] and, Figure 6 1 is a diagram showing a state where an inspection object is loaded onto a pallet at a loading position. Figure 7 1 is a diagram showing a state where the inspection object is moved to the inspection position as the tray is moved. Figure 8 This is a diagram showing a state where the inspection object is moved to the unloading position along with the tray.

[0070] Next, the transfer device 20 transfers the inspection object P along a preset transfer path to supply and recover the inspection object P.

[0071] The structure of the transfer device 20 is not particularly limited. Figure 2 and Figure 3 As shown, the transfer device 20 may include: a tray 22, which can load or unload the inspection object P; a transfer track 24, which extends along a preset transfer path; and a shuttle 26, which can use a clamp 26a to hold or release the tray 22, and is installed on the transfer track 24 in a movably manner to transfer the tray 22 and the inspection object P loaded on the tray 22 along the transfer path.

[0072] Furthermore, the transfer path is set so that the shuttle 26 can be used to reciprocate the pallet 22 between the preset loading position A1 and the unloading position A2, so that the inspection position A3 where the inspection device 30 is installed is located between the loading position A1 and the unloading position A2.

[0073] The structure of the tray 22 is not particularly limited. Figure 4 and Figure 5 As shown, the tray 22 may include: a loading slot 22a, which is recessed on the upper surface in a manner that allows loading or unloading of the inspection object P; and a guide hole 22b, which is formed through the loading slot 22a in a manner that allows the inspection light Li released from the lighting 33 described later to penetrate the inspection object P loaded in the loading slot 22a. In particular, preferably, the loading slot 22a is formed in a manner that allows the lower surface end of the inspection object P to be hung, and the guide hole 22b is formed in a manner that allows the lower surface of the inspection object P hung in the loading slot 22a to be exposed to the lower side direction of the tray 22 through the guide hole 22b, but it is not limited thereto.

[0074] like Figure 6 and Figure 7As shown, the transfer device 20 can transfer the pallet 22 loaded with the inspection object P that needs to be inspected for pinholes from the loading position A1 to the inspection position A3 and transfer the inspection object P to the inspection device 30. In this case, the inspection object P can be loaded into the pallet 22 in the loading position A1 by a worker or a separate automated device.

[0075] And, if Figure 8 As shown, the transfer device 20 can transfer the pallet 22 on which the inspection object P has been inspected for pinholes by the inspection device 30 from the inspection position A3 to the unloading position A2, where the inspection object P can be unloaded and recovered from the pallet 22. In this case, the inspection object P can be unloaded from the pallet 22 in the unloading position A2 by a worker or a separate automated device.

[0076] Fig. 9 is a block diagram for explaining a control system of an inspection device. Fig.10 This is a diagram showing a state where the inspection object enters the inner space of the darkroom along with the tray. Fig.11 This is a diagram showing a state where a pinhole inspection is performed on a pallet in the inner space of a darkroom.

[0077] Next, the inspection device 30 is used to perform a pinhole inspection on the inspection object P.

[0078] After the inspection device 30 irradiates the inspection light Li for pinhole inspection onto the inspection object P, the inspection device 30 can perform the pinhole inspection on the inspection object P by analyzing the captured image of the inspection object P captured while the inspection light Li passes through the inspection object P.

[0079] For example, Figure 7 and Fig. 9 As shown, the inspection device 30 may include a darkroom 31, a lift 32, lighting 33, a camera 34, a visual processor 35, and the like.

[0080] The darkroom 31 can provide a space for performing pinhole inspection in a dark environment without any light source other than the inspection light Li provided from the illumination 33 described later. Figure 7 As shown, the darkroom 31 may include: an internal space 31a, which is isolated in a manner to prevent light from entering from the outside; and an entrance 31b, which can allow the inspection object P or various devices required for pinhole inspection to enter the internal space 31a or lead them out of the internal space 31a. And, preferably, the entrance 31b is formed through the bottom of the darkroom 31 in a manner of communicating with the internal space 31a.

[0081] The darkroom 31 is provided so that the entrance 31b is separated from the tray 22 arriving at the inspection position A3 by a predetermined interval. For example, when the entrance 31b is formed at the bottom of the darkroom 31, the darkroom 31 is provided so that the entrance 31b is separated from the tray 22 arriving at the inspection position A3 by a predetermined reference distance upward.

[0082] The lift 32 allows the tray 22 arriving at the inspection position A3 and the inspection object P loaded on the tray 22 to enter the internal space 31a of the darkroom 31 through the entrance 31b of the darkroom 31 or to be led out from the internal space 31a.

[0083] To this end, the elevator 32 may include: a fixed bracket 32a, which is fixed with at least one bracket 32c for supporting the tray 22 from the inspection position A3; and a lifting component 32b, which reciprocates along the height direction to move the fixed bracket 32a and the tray 22 supported by the bracket 32c of the fixed bracket 32a.

[0084] Furthermore, the support 32c is provided on the lifting member 32b so as to support the tray 22 that has reached the inspection position A3 during the lifting process. In particular, preferably, the support 32c does not face the guide hole 22b and does not interfere with the shuttle 26.

[0085] Furthermore, the lifting component 32b may be one of a plurality of devices that enables the fixing bracket 32a to be lifted and lowered in the height direction. For example, the lifting component 32b may be a cylinder device.

[0086] When the tray 22 reaches the inspection position A3, the lift 32 uses the lifting member 32b to lift the fixed bracket 32a so that the tray 22 is supported by the bracket 32c. Therefore, the clamp 26a of the shuttle 26 releases the grip of the tray 22 and transfers it to the bracket 32c, so that the tray 22 can be lifted and lowered along the fixed bracket 32a while being supported by the bracket 32c.

[0087] And, if Fig.10 As shown, when the tray 22 is transferred to the bracket 32c, the elevator 32 uses the lifting component 32b to lift the tray 22 so that the tray 22 enters the internal space 31a of the darkroom 31 through the entrance 31b to a preset reference depth.

[0088] When the pinhole inspection of the inspection object P loaded on the tray 22 is completed in the internal space 31a of the darkroom 31, the lift 32 lowers the tray 22 from the inspection position A3 to the installation height of the waiting shuttle 26 by means of the lifting member 32b, so that the tray 22 is led out of the internal space 31a of the darkroom 31 through the entrance 31b. Therefore, the gripper 26a of the shuttle 26 can hold the tray 22 and retrieve it from the lift 32.

[0089] Furthermore, when the pallet 22 is recovered by the clamp 26 a of the shuttle 26 , the lift 32 lowers the fixed bracket 32 ​​a using the lifting member 32 b so that the bracket 32 ​​c is spaced apart from the pallet 22 by a preset distance.

[0090] When the inspection object P arrives at the inspection position A3 along with the tray 22, the lighting 33 may irradiate the inspection light Li for pinhole inspection to the lower surface of the inspection object P exposed through the guide hole 22b of the tray 22. Figure 7 As shown, the lighting 33 can be fixedly installed on the fixed bracket 32a so as to face the lower surface of the inspection object P that arrives at the inspection position A3 along with the tray 22 through the guide hole 22b. Therefore, the lighting 33 can be raised and lowered together with the tray 22 by the lifting member 32b in a state where it is spaced downward from the inspection object P by a distance corresponding to the height of the bracket 32c. In this regard, preferably, the bracket 32c has a height that allows the inspection light Li released from the lighting 33 to be uniformly irradiated to the entire lower surface area of ​​the inspection object P.

[0091] like Fig.11 As shown, when the tray 22 enters the inner space 31a of the darkroom 31 through the entrance 31b to a preset reference depth, the lighting 33 can release the inspection light Li toward the lower surface of the inspection object P to irradiate the inspection light Li to the lower surface of the inspection object P.

[0092] Preferably, the inspection light Li is one of white light, red light, blue light, and green light in the commonly used visible light range. However, the present invention is not limited thereto, and the inspection light Li may also be light in the ultraviolet (UV) range or infrared (IR) range that is not in the visible light range.

[0093] The illumination 33 may irradiate the inspection light Li having a preset reference illuminance range toward the inspection object P. For example, the reference illuminance range may be 100000 Lux to 600000 Lux.

[0094] The camera 34 is disposed in the inner space 31a of the darkroom 31 so as to photograph the other side of the inspection object P opposite to the side of the inspection object P irradiated with the inspection light Li by the illumination 33 when the inspection object P and the illumination 33 enter the preset reference depth inside the darkroom 31 by the lift 32. For example, when the illumination 33 is disposed in such a manner that the inspection light Li can irradiate the lower surface of the inspection object P, the camera 34 is disposed in the inner space 31a of the darkroom 31 so as to photograph the upper surface of the inspection object P.

[0095] There is no particular limitation on the type of camera that can be used as the camera 34. For example, the camera 34 may be an area camera.

[0096] Generally, the camera may have a preset shooting area, and a plurality of shooting regions having an area corresponding to the shooting area may be set on the upper surface of the inspection object P. In this case, the inspection device 30 may further include a camera transfer machine 36 for transferring the camera 34 in at least one of the X direction and the Y direction, so that the camera 34 can move along the preset path and sequentially shoot the plurality of shooting regions.

[0097] When the inspection light Li is irradiated to the one side of the inspection object P through the lighting 33, the camera 34 can capture the other side of the inspection object P, thereby generating a captured image of the inspection object P showing the state in which the inspection light Li penetrates the inspection object P.

[0098] Fig.12 is a diagram for explaining a control system of a visual processor. Fig.13 is a diagram for explaining a control system of a first algorithm unit. Fig.14 is a diagram showing a first input image, Fig.15 1 is a flowchart showing a pixel value distribution state of a plurality of pixels included in a first input image. Fig.16 A diagram for explaining a method of generating a first feature extraction result image by binarizing a plurality of pixels based on a first threshold.

[0099] The vision processor 35 analyzes the state of the inspection light Li transmitting the inspection object P based on the captured image of the inspection object P transmitted from the camera 34 , so that the pinhole inspection of the inspection object P can be performed.

[0100] The structure of the visual processor 35 is not particularly limited. Fig.12 As shown, the visual processor 35 may include: a first algorithm unit 35a, which uses a preset first algorithm to analyze the captured image to implement a first feature extraction, in which a plurality of first candidate areas Ac1 where the pinhole Hp is estimated to be located are specified; a second algorithm unit 35b, which uses a preset second algorithm to analyze the captured image to implement a second feature extraction and a third feature extraction, in which a plurality of second candidate areas Ac2 where the pinhole Hp is estimated to be located are specified, and in the third feature extraction, a plurality of first candidate areas Ac1 and a plurality of second candidate areas Ac2 are compared and analyzed, and a plurality of first candidate areas Ac1 and a plurality of second candidate areas Ac2 selected according to a preset method from the plurality of first candidate areas Ac1 and the plurality of second candidate areas Ac2 are specified as third candidate areas Ac3; and a third algorithm unit 35c, which uses a preset third algorithm to analyze a plurality of third candidate areas Ac3 to implement a pinhole or not judgment for determining whether the pinhole Hp is actually located in each third candidate area Ac3.

[0101] The first algorithm unit 35a may perform the first feature extraction by presetting a first algorithm to specify a plurality of first candidate areas Ac1 where the pinholes Hp are estimated to be located in the entire area of ​​the inspection object P displayed in the captured image of the inspection object P.

[0102] For this reason, Fig.13 As shown, the first algorithm unit 35a may include: a first input-output unit 35a1, used to input and output data related to the first feature extraction and other data; a first processor 35a2, using the first algorithm to implement the first feature extraction; and a first memory 35a3, etc., for storing data related to the first feature extraction and other data.

[0103] Furthermore, the first algorithm of the first processor 35a2 is composed of an artificial neural network deep learning algorithm (e.g., YOLOv7), but is not limited thereto.

[0104] Hereinafter, a method of performing the first feature extraction using the first algorithm unit 35 a will be described with reference to the drawings.

[0105] First, the first input / output unit 35 a 1 inputs a captured image transmitted from the camera 34 to the first processor 35 a 2 .

[0106] Preferably, the first feature extraction is performed using a black-and-white image of the inspection object P. To this end, the first input / output unit 35a1 directly inputs the black-and-white image of the inspection object P transmitted from the camera 34 to the first processor 35a2, or converts the color image of the inspection object P transmitted from the camera 34 into a black-and-white image and then inputs it to the first processor 35a2, depending on the type of the camera 34. Fig.14 As shown, the first input / output unit 35a1 names the black-and-white image of the inspection object P input to the first processor 35a2 based on the captured image of the inspection object P transmitted from the camera 34 for the first feature extraction as a first input image I1.

[0107] Next, the first processor 35a2 performs a first feature extraction to specify the first candidate area Ac1 in the entire area of ​​the inspection object P displayed in the first input image I1.

[0108] Generally, when the inspection light Li is irradiated to the inspection object P, in the pinhole Hp in the entire area of ​​the inspection object P, a larger amount of the inspection light Li can pass through the inspection object P, and in the normal area An without the pinhole Hp or other defects, the inspection light Li is blocked by the inspection object P or only a very small amount of the inspection light Li can pass through the inspection object P. For the convenience of explanation, hereinafter, among the inspection light Li irradiated to the inspection object P, the inspection light Li that passes through the pinhole Hp or other areas of the inspection object P and passes through the inspection object P is referred to as transmitted light.

[0109] like Fig.15 As shown, among the plurality of pixels included in the first input image I1, when the plurality of pixels representing the inspection object P are classified based on the size of the pixel value representing the pixel brightness (for example, 0 to 255), in the entire area of ​​the inspection object P, the pixel values ​​of most of the pixels of the first input image I1 representing the normal area An are distributed near the minimum value (for example, 0) at which the pixels appear as dark spots, and in the entire area of ​​the inspection object P, only a very small number of pixels have high pixel values ​​that make the pixels appear brightly in the form of bright spots due to the transmitted light that transmits the inspection object P. For the convenience of explanation, hereinafter, in the entire area of ​​the inspection object P, the area having a higher pixel value than the pixel values ​​of the plurality of pixels belonging to the normal area An due to the transmitted light is referred to as the transmitted light area At.

[0110] However, depending on the size and shape of the pinhole Hp, when the amount of light transmitted through the pinhole Hp is small, the transmitted light area At may not be clear, and when the amount of light transmitted through the pinhole Hp is large, the transmitted light area At appears brightly in the form of a complete bright spot. Fig.14 and Fig.16 As shown, in order to clearly identify the weak transmitted light of the transmission inspection object P, the first processor 35a2 can implement binarization by converting the pixel values ​​of multiple pixels having pixel values ​​above the first threshold to the maximum value (for example, 255) and converting the pixel values ​​of multiple pixels having pixel values ​​less than the first threshold to the minimum value (for example, 0) based on the first feature extraction result data accumulated in the first memory 35a3. Preferably, the first threshold is determined by the average pixel value (for example, 5) of multiple pixels located at the boundary of the normal area An and the transmitted light area At, but it is not limited to this. Such a first threshold can be stored in the first memory 35a3 and can be updated to an appropriate value according to the pinhole inspection result.

[0111] According to the binarization, a plurality of pixels having pixel values ​​above the first threshold value are all brightly displayed as completely bright spots, and a plurality of pixels having pixel values ​​less than the first threshold value are displayed as dark spots. Through such binarization, the first processor 35a2 can make the transmitted light area At that is weakly displayed in the first input image I1 due to relatively low pixel values ​​brightly displayed as completely bright spots among the plurality of transmitted light areas At, thereby clearly distinguishing the plurality of transmitted light areas At having a plurality of pixels having pixel values ​​above the first threshold value, the normal area An having a plurality of pixels having pixel values ​​below the first threshold value, and the remaining area of ​​the other inspection object P. That is, through such binarization, the first processor 35a2 can clearly identify the plurality of transmitted light areas At having a plurality of pixels having pixel values ​​above the first threshold value.

[0112] Furthermore, when it is determined that a large amount of noise corresponding to the transmitted light region At formed due to other reasons exists in the first input image I1, rather than due to the pinhole Hp, the first processor 35a2 may perform noise processing (pre-processing / post-processing) such as blurring, sharpening, contrast enhancement, and filtering in at least one period before and after binarization to remove the noise from the first input image I1. For example, after binarization, the first processor 35a2 regards the transmitted light region At composed of pixels less than a preset reference number (for example, 2) as noise and removes it from the first input image I1.

[0113] The method of removing noise from the first input image I1 is not particularly limited. For example, the first processor 35a2 can remove noise by transforming the pixel values ​​of the plurality of pixels corresponding to the noise into minimum values ​​in such a way that the plurality of pixels corresponding to the noise appear as dark spots.

[0114] On the other hand, there is a situation where excessive transmitted light is transmitted through the inspection object P due to tearing of the inspection object P or other reasons, so that the transmitted light area At is excessively displayed in the first input image I1. In this regard, when an area of ​​a preset ratio or more (for example, 50%) corresponds to the transmitted light area At in the entire area of ​​the inspection object P displayed in the first input image I1, the first processor 35a2 can implement an exception process to judge the inspection object P as defective. Preferably, when such an exception process is implemented, the second feature extraction and other subsequent inspection processes of the inspection object P judged as defective are interrupted. When such an exception process is performed, a part of the operation can be selectively added in order to improve the accuracy of the defect judgment of the inspection object P and the efficiency of the pinhole inspection.

[0115] Furthermore, the first processor 35a2 specifies the plurality of transmitted light areas At displayed as bright spots in the first input image I1 as the first candidate areas Ac1 where the pinhole Hp is estimated to be located, while performing binarization, noise removal, and exception processing. However, the plurality of transmitted light areas At may have complex contours of irregular shapes, and it is difficult to accurately derive the coordinates and sizes of the plurality of transmitted light areas At. Fig.16 As shown, after the first processor 35a2 specifies a plurality of first candidate areas Ac1 in a box-shaped manner in which each of the plurality of transmitted light areas At is included in the interior thereof, the plurality of first candidate areas Ac1 specified as above can be displayed on the first input image I1. Hereinafter, the first input image I1 in which the plurality of first candidate areas Ac1 are specified in a state where binarization, noise removal, and exception processing are performed is named as a first feature extraction result image I1a.

[0116] Afterwards, the first memory 35a3 can store the first feature extraction result image I1a generated as above together with the position information (x coordinate, y coordinate), size information (width, height, area) and other data of the plurality of first candidate regions Ac1.

[0117] Fig.17 is a diagram for explaining a control system of a second algorithm unit. Fig.18 To illustrate a diagram of a second input image, Fig.19 is a diagram for explaining the concept of diffusion area. Fig. 20 A diagram for explaining a method of generating a second feature extraction result image by binarizing a plurality of pixels based on a second threshold.

[0118] Fig.21 1 is a diagram showing a state where a plurality of first candidate regions and a plurality of second candidate regions are displayed on a third input image. Fig. 22 FIG. 1 is a diagram for explaining a method of canceling the designation of the first candidate region designated as a diffusion region. Figure 23 to Figure 25 A diagram for explaining a method of canceling the designation of the first candidate area and the second candidate area designated with the reflection area as the object, Fig.26 A diagram for explaining a method of generating a third feature extraction result image by specifying a third candidate region.

[0119] The second algorithm unit 35b can utilize the preset second algorithm to implement the second feature extraction and the third feature extraction. In the second feature extraction, a plurality of second candidate areas Ac2 where the pinholes Hp are estimated to be located are specified in the entire area of ​​the inspection object P displayed in the captured image of the inspection object P. In the third feature extraction, a specific plurality of first candidate areas Ac1 and a specific plurality of second candidate areas Ac2 screened by comparing and analyzing the plurality of first candidate areas Ac1 and the plurality of second candidate areas Ac2 as well as other preset reference methods are designated as three candidate areas Ac3.

[0120] For this reason, Fig.17 As shown, the second algorithm unit 35b may include: a second input-output unit 35b1, used to input and output data related to the second feature extraction, data related to the third feature extraction, and other data; a second processor 35b2, using the second algorithm to implement the second feature extraction and the third feature extraction; and a second memory 35b3, storing data related to the first feature extraction, data related to the second feature extraction, and other data.

[0121] And, preferably, the second algorithm of the second processor 35b2 is composed of an artificial neural network deep learning algorithm (e.g., YOLOv7), but is not limited thereto.

[0122] Hereinafter, a method of performing the second feature extraction using the second algorithm unit 35b will be described with reference to the drawings.

[0123] First, the second input / output unit 35 b 1 inputs the captured image transmitted from the camera 34 to the second processor 35 b 2 .

[0124] Preferably, the second feature extraction is performed using a black-and-white image of the inspection object P. In this regard, the second input / output unit 35b1 directly inputs the black-and-white image of the inspection object P transmitted from the camera 34 to the second processor 35b2, or converts the color image of the inspection object P transmitted from the camera 34 into a black-and-white image and then inputs it to the second processor 35b2, depending on the type of the camera 34. Fig.18 As shown, hereinafter, the black-and-white image of the inspection object P inputted by the second input-output unit 35b1 to the second processor 35b2 for the second feature extraction based on the captured image of the inspection object P transmitted from the camera 34 is named as a second input image I2.

[0125] Next, the second processor 35b2 performs a second feature extraction to specify the second candidate area Ac2 in the entire area of ​​the inspection object P displayed in the second input image I2.

[0126] Generally, when the inspection light Li passes through the pinhole Hp, a portion of the inspection light Li passing through the pinhole Hp may diffuse to the surrounding area of ​​the pinhole Hp. This diffusion phenomenon becomes more serious as the size of the pinhole Hp increases or the brightness of the illumination 33 increases. Fig.19 As shown, when the diffusion phenomenon of the inspection light Li occurs in the peripheral area of ​​the pinhole Hp, in the captured image of the inspection object P, the first input image I1 based on the captured image of the inspection object P, the second input image I2, and other various images, the diffusion area Ad formed by the inspection light Li diffusing to the peripheral area of ​​the pinhole Hp moves along the formation direction of the pinhole Hp without diffusion, and the inspection light Li directly passing through the pinhole Hp is merged with the pinhole area Ap formed in the area where the pinhole Hp actually exists, thereby displaying a transmitted light area At. In this regard, the following error may exist in the first feature extraction result image I1a, in which the first candidate area Ac1 is specified with the transmitted light area At merged into one by the pinhole area Ap and the diffusion area Ad as the object, and the peripheral area of ​​the pinhole Hp formed with the diffusion area Ad is mistakenly identified as a part of the pinhole Hp. For the convenience of explanation, the inspection light Li diffused to the peripheral area of ​​the pinhole Hp in the process of passing through the pinhole Hp in the inspection light Li is named as diffuse light.

[0127] The diffused light has the property of spreading at a wide diffusion angle, and the brightness of the diffused area Ad is relatively darker than that of the pinhole area Ap. Fig. 20 As shown, in order to clearly distinguish the diffusion area Ad from the pinhole area Ap, the second processor 35b2 implements binarization for the second input image I2, in which the feature qualities of multiple pixels having pixel values ​​above the second threshold are all converted into the maximum value (e.g., 255) in a manner having a pixel value higher than the first threshold, and the pixel values ​​of multiple pixels having pixel values ​​less than the second threshold are all converted into the minimum value (e.g., 0). Preferably, the second threshold is determined by the average pixel value (e.g., 180) of multiple pixels located at the boundary of the pinhole area Ap and the diffusion area Ad based on the second feature extraction result data accumulated in the second memory 35b3, but is not limited thereto. Such a second threshold can be stored in the second memory 35b3 and can be updated to an appropriate value according to the pinhole inspection result.

[0128] According to the binarization, a plurality of pixels having pixel values ​​above the second threshold value are brightly displayed in the form of complete bright spots, and a plurality of pixels having pixel values ​​below the second threshold value are displayed in the form of dark spots. Through such binarization, the second processor 35b2 displays the pinhole region Ap, which can be estimated as a region having pixel values ​​of a plurality of pixels above the second threshold value, in the entire region of each transmitted light region At as a complete bright spot, and displays the diffusion region Ad, which can be estimated as having pixel values ​​of a plurality of pixels below the second threshold value, as a dark spot. That is, the second processor 35b2 displays and removes the normal region An, where the pinhole Hp does not actually exist, but is included in the transmitted light region At as the diffusion region Ad and is mistakenly regarded as a part of the pinhole Hp, as a dark spot.

[0129] Furthermore, when it is determined that a large amount of noise corresponding to the transmitted light region At formed due to other reasons exists in the second input image I2, rather than due to the pinhole Hp, the second processor 35b2 may perform noise processing (pre-processing / post-processing) such as blurring, sharpening, contrast enhancement, and filtering in at least one period before and after binarization to remove the noise from the second input image I2. For example, after binarization, the second processor 35b2 regards the transmitted light region At composed of pixels less than a preset reference number (for example, 2) as noise and removes it from the second input image I2.

[0130] The method of removing noise from the second input image I2 is not particularly limited. For example, the second processor 35b2 may transform the pixel values ​​of the plurality of pixels corresponding to the noise into minimum values ​​so that the plurality of pixels corresponding to the noise appear as dark spots to remove the noise.

[0131] Furthermore, the second processor 35b2 can designate the plurality of transmitted light areas At that appear as bright spots in the second input image I2 as the second candidate areas Ac2 where the pinhole Hp is estimated to be located, while binarizing and removing noise. However, the plurality of transmitted light areas At may have complex contours of irregular shapes, so it is difficult to derive the coordinates and sizes of the plurality of transmitted light areas At. Fig. 20 As shown, after the second processor 35b2 specifies a plurality of second candidate areas Ac2 in a box-shaped manner having a polygon (e.g., a quadrilateral) each including one of the plurality of transmitted light areas At inside, the plurality of second candidate areas Ac2 specified as above can be displayed on the second input image I2. Hereinafter, the second input image I2 in which the plurality of second candidate areas Ac2 are specified in a state of binarization and noise removal is named as a second feature extraction result image I2a.

[0132] Afterwards, the second memory 35b3 can store the second feature extraction result image I2a generated as above together with the position information (x coordinate, y coordinate), size information (width, height, area) and other data of multiple second candidate areas Ac2.

[0133] Hereinafter, the third feature extraction method implemented by the second algorithm unit 35b will be described with reference to the drawings.

[0134] First, the second input / output unit 35b1 inputs the captured image of the inspection object P transmitted from the camera 34 and the first feature extraction result image I1a transmitted from the first input / output unit 35a1 of the first algorithm unit 35a to the second processor 35b2.

[0135] Preferably, the third feature extraction is performed using a black-and-white image of the inspection object P. In this regard, the second input / output unit 35b1 directly inputs the black-and-white image of the inspection object P transmitted from the camera 34 to the second processor 35b2, or converts the color image of the inspection object P transmitted from the camera 34 into a black-and-white image and then inputs it to the second processor 35b2, depending on the type of the camera 34. Fig.21 As shown, hereinafter, the second input-output unit 35b1 names the black-and-white image of the inspection object P input to the second processor 35b2 based on the captured image of the inspection object P transmitted from the camera 34 for the third feature extraction as a third input image I3.

[0136] Next, the second processor 35b2 implements a third feature extraction for designating the third candidate area Ac3, in which the specific multiple first candidate areas Ac1 and the specific multiple second candidate areas Ac2 screened by comparative analysis of the multiple first candidate areas Ac1 based on the first feature extraction and the multiple second candidate areas Ac2 based on the second feature extraction and other methods are inferred to be the area where the pinhole Hp is located.

[0137] The second processor 35b2 specifies a plurality of first candidate regions Ac1 and a plurality of second candidate regions Ac2 in the third input image I3 based on information about the plurality of first candidate regions Ac1 included in the first feature extraction result image I1a and information about the plurality of second candidate regions Ac2 included in the second feature extraction result image I2a. Fig.21 As shown, when multiple first candidate areas Ac1 and multiple second candidate areas Ac2 are displayed in a box shape in the first feature extraction result image I1a and the second feature extraction result image I2a, the second processor 35b2 may display multiple first candidate areas Ac1 and multiple second candidate areas Ac2 in a box shape in the third input image I3.

[0138] Furthermore, the second processor 35b2 compares the plurality of first candidate regions Ac1 with the plurality of second candidate regions Ac2 to screen the plurality of first candidate regions Ac1 and the plurality of second candidate regions Ac2 that are designated as overlapping each other. Preferably, when the distance between a center point of the plurality of first candidate regions Ac1 and a center point of the plurality of second candidate regions Ac2 is less than a preset reference interval, the second processor 35b2 determines that one of the first candidate regions Ac1 and one of the second candidate regions Ac2 are designated as overlapping each other, but is not limited thereto.

[0139] Furthermore, the second processor 35b2 may selectively cancel the designation of the plurality of first candidate regions Ac1 among the plurality of first candidate regions Ac1 and the plurality of second candidate regions Ac2 designated in a manner overlapping with each other. Fig. 22 As shown, when multiple first candidate areas Ac1 and multiple second candidate areas Ac2 are displayed in a box shape in the third input image I3, the second processor 35b2 can remove the box representing the first candidate area Ac1 from the multiple first candidate areas Ac1 and multiple second candidate areas Ac2 displayed overlapping each other from the third input image I3.

[0140] As described above, in the first feature extraction result image I1a, there may be the following error: multiple first candidate areas Ac1 are designated based on the transmitted light area At formed by merging the pinhole area Ap and the diffusion area Ad, and the peripheral area of ​​the pinhole Hp formed with the diffusion area Ad is mistakenly recognized as a part of the pinhole Hp. In this regard, when the size of the pinhole Hp is calculated based on the multiple first candidate areas Ac1, there may be an error that the calculated size of the pinhole Hp is larger than the actual size of the diffusion area Ad. However, in the second feature extraction result image, multiple second candidate areas Ac2 are designated based on the diffusion area Ad in which multiple pixels with pixel values ​​less than the second threshold are removed and the multiple transmitted light areas At are corrected in a manner that only the pinhole area Ap remains. In this regard, in order to prevent the diffusion area Ad from being mistakenly recognized as a part of the pinhole Hp, when the multiple first candidate areas Ac1 and the multiple second candidate areas Ac2 are designated to overlap each other, the second processor 35b2 selectively cancels the designation of the multiple first candidate areas Ac1.

[0141] However, depending on the size and shape of the pinhole Hp, when the amount of light transmitted through the pinhole Hp is small, the transmitted light area At may appear weakly. In response to this, in the first feature extraction result image I1a, the weak multiple transmitted light areas At with pixel values ​​of multiple pixels being greater than the first threshold and less than the second threshold can be designated as multiple first candidate areas Ac1. On the contrary, in the second feature extraction result image I2a, the weak multiple transmitted light areas At with pixel values ​​of multiple pixels being greater than the first threshold and less than the second threshold cannot be designated as multiple second candidate areas Ac2. Therefore, after the operation of selectively releasing the designation of multiple first candidate areas Ac1 that overlap with multiple second candidate areas Ac2 is performed, the designation of multiple first candidate areas Ac1 with pixel values ​​of multiple pixels being greater than the first threshold and less than the second threshold is not released, but is retained. Among them, most of the first candidate areas Ac1 in which the pixel values ​​of multiple pixels are above the first threshold and less than the second threshold are caused by small-sized pinholes Hp that rarely occur in the diffusion area of ​​the transmitted light. The transmitted light area At only includes a very small-sized diffusion area Ad, and there is no or very little error that the calculated size of the pinhole Hp is larger than the actual size due to the diffusion area Ad.

[0142] On the other hand, when the camera 34 is used to photograph the other side of the inspection object P, when the transmitted light reflected by the lens of the camera 34 after passing through the pinhole Hp is irradiated to the other side of the inspection object P, a reflection area Ar formed by the transmitted light reflected by the lens of the camera 34 may be formed in the photographed image and the input images I1, I2, I3, and other images generated based on the photographed image. Usually, the pixel values ​​of the multiple pixels included in the reflection area Ar are greater than the pixel values ​​of the multiple pixels included in the normal area An. In other words, the pixel values ​​may be greater than the first threshold value. In this regard, if Fig.23 and Fig.24 As shown, in the third input image I3, there may be an error in specifying the first candidate area Ac1 or the second candidate area Ac2 based on the reflection area Ar because the reflection area Ar is mistakenly identified as the pinhole area Ap. For the convenience of explanation, the transmitted light reflected by the lens of the camera 34 is named as reflected light below.

[0143] However, the reflected light loses a part of the light amount in the process of being reflected by the lens of the camera 34, and then irradiates the other side of the inspection object P at a state of a reflection angle of the lens of the camera 34 with a wider propagation range. In this regard, the area of ​​the reflection region Ar is larger than the area of ​​the pinhole region Ap, and the brightness is darker than the brightness of the pinhole region Ap. At the same time, the brightness of the general pinhole region Ap gradually becomes brighter toward the center, and the general reflection region Ar has a uniform brightness as a whole.

[0144] Taking into account the characteristics of such a reflection area Ar, the second processor 35b2 analyzes the transmitted light areas At respectively included in the plurality of first candidate areas Ac1 and the plurality of second candidate areas Ac2 displayed in the third input image I3 and regards the reflection area Ar as the object to designate a specific plurality of first candidate areas Ac1 and a specific plurality of second candidate areas Ac2, wherein the average pixel value of the plurality of pixels included in the transmitted light area At is lower than a preset reference value or the transmitted light area At only includes a plurality of pixels having pixel values ​​outside the preset reference range. And, preferably, based on the third feature extraction result data accumulated in the second memory 35b3, the reference value is determined by the average pixel value of the plurality of pixels included in the plurality of pinhole areas Ap, and the reference range is determined to be above the lowest pixel value and below the highest pixel value of the plurality of pixels included in the plurality of pinhole areas Ap, but is not limited thereto. Such a reference value and a reference range can be stored in the second memory 35b3, and can be updated to appropriate values ​​according to the pinhole inspection result.

[0145] Furthermore, the second processor 35b2 may cancel the designation of the plurality of first candidate areas Ac1 and the plurality of second candidate areas Ac2 designated with the reflection area Ar as the object. Fig.25As shown, when multiple first candidate areas Ac1 and multiple second candidate areas Ac2 are displayed in a box shape in the third input image I3, the second processor 35b2 can remove the boxes representing the multiple first candidate areas Ac1 and multiple second candidate areas Ac2 specified with the reflection area Ar as the object from the third input image I3.

[0146] Furthermore, the second processor 35b2 may designate the plurality of first candidate areas Ac1 and the plurality of second candidate areas Ac2 remaining in the third input image I3 as the third candidate area Ac3 where the estimated pinhole Hp is located, while canceling the designation of the plurality of first candidate areas Ac1 and the plurality of second candidate areas Ac2 designated for the diffusion area Ad and the reflection area Ar. Fig.26 As shown, the second processor 35b2 can display the plurality of third candidate areas Ac3 in a box shape on the third input image I3. Hereinafter, the third input image I3 in which the plurality of third candidate areas Ac3 are designated in a state where the plurality of first candidate areas Ac1 and the plurality of second candidate areas Ac2 designated with the diffusion area Ad and the reflection area Ar as objects are released is named as the third feature extraction result image I3a.

[0147] Afterwards, the second memory 35b3 can store the third feature extraction result image I3a generated as above together with the position information (x coordinate, y coordinate), size information (width, height, area) and other data of multiple third candidate areas Ac3.

[0148] Fig. 27 : is a diagram for explaining a control system of a third algorithm unit. Fig.28 A diagram for explaining a method of generating a pinhole inspection result image by specifying a final candidate region.

[0149] The third algorithm unit 35c can use a preset third algorithm to analyze multiple third candidate areas Ac3 included in the third feature extraction result image I3a to determine whether the pinhole Hp is actually located in the multiple third candidate areas Ac3 (hereinafter referred to as "pinhole judgment").

[0150] For this reason, Fig. 27 As shown, the third algorithm unit 35c may include: a third input-output unit 35c1, used to input and output data used for pinhole judgment and other data; a third processor 35c2, using the third algorithm to implement pinhole judgment; and a third memory 35c3, storing data related to the pinhole judgment data and other data.

[0151] And, preferably, the third algorithm of the third processor 35c2 is composed of an artificial neural network deep learning algorithm (e.g., YOLOv7), but is not limited thereto.

[0152] Hereinafter, a method of determining whether or not a pinhole is present using the third algorithm unit 35c will be described with reference to the drawings.

[0153] First, the third input-output unit 35c1 inputs the third feature extraction result image I3a transmitted from the second input-output unit 35b1 of the second algorithm unit 35b to the third processor 35c2.

[0154] Next, the third processor 35c2 analyzes the third feature extraction result image I3a to perform pinhole determination on multiple third candidate areas Ac3.

[0155] When the inspection object P has extremely sensitive material properties such as copper foil film, there may be a plurality of reflection areas Ar with various patterns formed by the surface roughness and other conditions of the inspection object P in the captured image of the inspection object P and the images I1, I2, and I3 generated based on the captured image. As a result, the third feature extraction result image I3a may have an error in specifying the third candidate area Ac3 based on the reflection area Ar that has not been found in the second feature extraction process. In addition, the third candidate area Ac3 with state information such as size and brightness (pixel value) may exist in the third feature extraction result image I3a, and the state information such as size and brightness belongs to the range that cannot effectively judge the transmitted light area A included in the third candidate area Ac3 as the actual pinhole Hp according to the hardware characteristics of the inspection device 30 (for example, the resolution of the camera 34). In this regard, the third processor 35c2 uses an artificial neural network deep learning algorithm and other statistical methods to analyze data related to status information of multiple third candidate areas Ac3, including the size and brightness (pixel value) of the transmitted light area At in each third candidate area Ac3, in order to determine whether multiple third candidate areas Ac3 have pinholes.

[0156] Furthermore, the third processor 35c2 can cancel the designation of the third candidate area Ac3 (hereinafter referred to as "invalid candidate area") where the transmitted light area At corresponds to the reflected area Ar or where the pinhole Hp is not actually located because the state (size, brightness, etc.) of the transmitted light area At exceeds the preset reference state, according to the data analysis result of the plurality of third candidate areas Ac3. Fig.28 As shown, the third processor 35c2 may convert the pixel values ​​of the plurality of pixels included in the ineffective candidate region into the minimum value, so that the ineffective candidate region appears as a dark spot to remove the ineffective candidate region. However, the present invention is not limited thereto, and when the plurality of third candidate regions Ac3 are displayed in the form of boxes on the third feature extraction result image I3a, the third processor 35c2 may remove only the boxes representing the ineffective candidate region from the third feature extraction result image I3a.

[0157] Furthermore, the third processor 35c2 cancels the designation of the invalid candidate area, designates the remaining third candidate area Ac3 as the final candidate area Ac4, and determines the transmitted light area At located in the final candidate area Ac4 as the pinhole Hp. For the convenience of explanation, the third feature extraction result image I3a after the pinhole determination is completed in the state of removing the invalid candidate area is named as the pinhole detection result image I4.

[0158] Afterwards, the third memory 35c3 can store the pinhole inspection result image I4 generated as above together with the position information (x coordinate, y coordinate) of the pinhole Hp, the size information (width, height, area) of the pinhole Hp and other data related to the pinhole Hp.

[0159] Hereinafter, a pinhole inspection method using the pinhole inspection system 1 will be described.

[0160] First, if Figure 6 As shown, at the loading position A1 , the inspection object P is loaded on the pallet 22 held by the shuttle 26 .

[0161] Then, if Figure 7 As shown, the shuttle 26 is used to move the pallet 22 from the loading position A1 to the inspection position A3.

[0162] Afterwards, if Fig.10 As shown, the elevator 32 of the inspection device 30 is used to allow the lighting 33 and the tray 22 to enter the inner space 31 a of the darkroom 31 to a preset depth through the entrance 31 b of the darkroom 31 .

[0163] Then, if Fig.11 As shown, while the inspection light Li is irradiated to one surface of the inspection object P loaded on the tray 22 by the lighting 33, the camera 34 is transferred along a preset shooting path by the camera transfer machine 36, and the other side of the inspection object P opposite to the said one side is photographed by the camera 34.

[0164] Thereafter, the visual processor 35 analyzes the image of the inspection object P captured by the camera 34 to implement pinhole inspection for collecting information on the pinholes Hp formed in the inspection object P, such as the number, position, and size of the pinholes Hp.

[0165] Then, if Figure 7 As shown, after the lighting 33 and the pallet 22 are brought out from the inner space 31 a of the darkroom 31 through the entrance 31 b of the darkroom 31 by the lift 32 of the inspection device 30 , the pallet 22 is transferred to the shuttle vehicle 26 .

[0166] Afterwards, if Figure 8 As shown, the pallet 22 is transferred from the inspection position A3 to the unloading position A2 by the shuttle 26 .

[0167] Next, at the unloading position A2 , the inspection object P is unloaded from the tray 22 .

[0168] The pinhole inspection system 1 can automatically and constantly perform pinhole inspection on the inspection object P using a preset algorithm, thereby reducing the time required for the pinhole inspection and improving the accuracy of the pinhole inspection.

[0169] Furthermore, the pinhole inspection system 1 performs pinhole inspection by detecting the transmission light area At generated by the inspection light Li penetrating the inspection object P in the dark environment of the darkroom 31. Thus, the pinhole inspection system 1 can suppress the noise caused by the surface reflection of the inspection object P, improve the recognition of the pinhole Hp and other surface defects of the inspection object P, reduce the complexity of the captured image of the inspection object P and other images, and thus further improve the accuracy of the pinhole inspection.

[0170] Furthermore, the pinhole inspection system 1 can remove the diffusion area Ad formed by the inspection light Li diffusing through the pinhole Hp and the reflection area Ar formed by the inspection light passing through the pinhole Hp being reflected from the lens of the camera 34 from the transmitted light area At, thereby further improving the accuracy of the pinhole inspection.

[0171] Furthermore, the pinhole inspection system 1 can analyze and infer multiple candidate areas where the pinhole Hp is located according to a statistical method and remove ineffective candidate areas from the multiple candidate areas, thereby further improving the accuracy of the pinhole inspection.

Claims

1. A pinhole inspection system, characterized in that: The pinhole inspection system comprises an inspection device, The inspection device comprises: Illumination, shining inspection light on one side of the inspection object; a camera, for photographing another side of the inspection object opposite to the one side, so as to generate a photographed image of the inspection object; and a visual processing unit that analyzes a state in which the inspection light penetrates the inspection object based on the captured image to perform a pinhole inspection on the inspection object, The visual processing unit includes: A first algorithm unit performs a first feature extraction using a preset first algorithm, wherein a plurality of first candidate regions where the pinholes are estimated to be located are specified in the first feature extraction; a second algorithm unit, using a preset second algorithm to perform a second feature extraction and a third feature extraction, in which a plurality of second candidate regions where the pinholes are estimated are specified in the second feature extraction, and in which a plurality of candidate regions selected by comparing the plurality of first candidate regions with the plurality of second candidate regions are specified as a plurality of third candidate regions where the pinholes are estimated in the third feature extraction; and The third algorithm unit uses a preset third algorithm to determine whether the pinhole is actually located in the plurality of third candidate areas.

2. The pinhole inspection system according to claim 1, characterized in that: The first algorithm unit takes the first input image generated based on the captured image as an object, classifies the pixel values ​​of multiple pixels representing the inspection object based on a preset first threshold value, specifies multiple transmitted light areas formed by the inspection light transmitting the inspection object in the entire area of ​​the inspection object, and specifies multiple first candidate areas in a manner that each includes one of the multiple transmitted light areas.

3. The pinhole inspection system according to claim 2, characterized in that: The first threshold is determined by an average pixel value of a plurality of pixels located at a boundary between a normal region of the inspection object and the transmitted light region.

4. The pinhole inspection system according to claim 2, characterized in that: The first algorithm unit takes the first input image as an object, and implements binarization to convert the pixel values ​​of multiple pixels having pixel values ​​greater than the first threshold into a maximum value and convert the pixel values ​​of multiple pixels having pixel values ​​less than the first threshold into a minimum value, so as to identify the multiple transmitted light areas.

5. The pinhole inspection system according to claim 2, characterized in that: The second algorithm unit takes a second input image generated based on the captured image as an object, and classifies pixel values ​​of multiple pixels representing the inspection object based on a preset second threshold value so that they are higher than the first threshold value. As a part of the inspection light diffuses in the process of passing through the pinhole, the multiple transmitted light areas are corrected in a manner to remove the diffused area formed around the pinhole, and the multiple second candidate areas are specified in a manner to respectively include one of the multiple corrected transmitted light areas.

6. The pinhole inspection system according to claim 5, characterized in that: The second threshold is determined by an average pixel value of a plurality of pixels located at a boundary between a pinhole region formed in a region where the pinhole formed by the inspection light actually exists and the diffusion region.

7. The pinhole inspection system according to claim 5, characterized in that: The second algorithm unit performs binarization on the second input image, converting pixel values ​​of multiple pixels having pixel values ​​greater than the second threshold to a maximum value and converting pixel values ​​of multiple pixels having pixel values ​​less than the second threshold to a minimum value, so as to remove the diffusion area.

8. The pinhole inspection system according to claim 5, characterized in that: The second algorithm unit specifies a plurality of the first candidate regions and a plurality of the second candidate regions in a third input image generated based on the captured image, and selectively cancels the designation of a plurality of first candidate regions among a plurality of first candidate regions and a plurality of second candidate regions that are specified to overlap with each other.

9. The pinhole inspection system according to claim 8, characterized in that: When a distance between a center point of one of the plurality of first candidate regions and a center point of one of the plurality of second candidate regions is equal to or smaller than a preset reference interval, the second algorithm unit determines that one of the first candidate regions and one of the second candidate regions are designated to overlap each other.

10. The pinhole inspection system according to claim 8, characterized in that: The second algorithm unit regards a transmitted light area whose average pixel value of multiple pixels is lower than a preset reference value or only includes multiple pixels with pixel values ​​outside a preset reference range as a reflection area formed by irradiating the other side of the inspection object with the inspection light reflected by the camera, and releases the designation of multiple first candidate areas and multiple second candidate areas designated with the reflection area as the object.

11. The pinhole inspection system according to claim 10, characterized in that: The reference value is determined by an average pixel value of a plurality of pixels included in a pinhole region formed in a region where the pinhole formed by the inspection light actually exists, The reference range is determined to be above the lowest pixel value and below the highest pixel value of a plurality of pixels included in the plurality of pinhole areas.

12. The pinhole inspection system according to claim 10, characterized in that: The second algorithm unit designates, as the plurality of third candidate areas, the plurality of first candidate areas overlapping with the plurality of second candidate areas and the plurality of first candidate areas and the plurality of second candidate areas remaining after the plurality of first candidate areas and the plurality of second candidate areas designated with the reflection area as objects are released.

13. The pinhole inspection system according to claim 12, characterized in that: The third algorithm unit analyzes state information of the transmitted light region located in each of the third candidate regions to designate the remaining third candidate regions other than the ineffective candidate region as final candidate regions among the plurality of third candidate regions, and determines the transmitted light region located in the final candidate region as a pinhole.

14. The pinhole inspection system according to claim 1, characterized in that: The pinhole inspection device also includes: Darkrooms are formed in such a way that the interior space is separated from the exterior; and The elevator allows the inspection object to enter the internal space or lead it out of the internal space.

15. The pinhole inspection system according to claim 14, characterized in that: The lighting enters the inner space together with the inspection object or is led out from the inner space, and is combined with the lift in such a way that the inspection light can be irradiated to the one side of the inspection object. The camera is disposed in the internal space so as to be able to photograph the other side of the inspection object entering the internal space.

Images