Cognitive test failure determination system and cognitive test failure determination method
Patent Information
- Application Number
- KR1020250012961
- Authority / Receiving Office
- KR · KR
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-02-03
- Publication Date
- 2026-08-11
Smart Images
Figure PAT00002_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to a cognitive inspection defect detection system, and more specifically, to a cognitive inspection defect detection system capable of detecting defects with high inspection efficiency. Background Technology
[0002] In the production process of lightweight automotive die-casting parts based on the AL6*** alloy family, remote autonomous teaching and correction technologies for control programs of robots and automated equipment are required to enable autonomous modification and perform precise correction tasks. To this end, technology is required to provide the ability for robots and automated equipment to rapidly recognize 3D CAD models of lightweight automotive parts by learning them, and to precisely measure manufacturing tolerances for key post-processes of die-casting, such as trimming, coating (release agent spraying), deburring, and perception inspection. It is also essential to have the capability to ensure alignment between the 3D model and the actual environment in key processes through offline programming (OLP) of the robot, and to accurately estimate geometric information of the manufacturing environment. Furthermore, automatic correction technology for robot control programs capable of recognizing manufacturing system components and avoiding obstacles is also a critical requirement.
[0003] More specifically, in the perception inspection stage, to accurately determine whether a product is good or defective, the presence or absence of defects can be determined by analyzing images captured by a vision camera while moving the frame during the product loading / unloading process. In the past, there were problems such as reduced inspection efficiency and decreased productivity and management convenience of the process line because it was necessary to compare the overall information of the acquired image with the overall information of the standard image. Prior art literature
[0004] Republic of Korea Published Patent 10-2023-0082806 "Method for Detecting Transfer Clearance Abnormalities in Cold Forging Equipment Using a Machine Vision System" The problem to be solved
[0005] The present invention has been devised to solve the aforementioned problems, and the objective of the present invention is to provide a cognitive inspection defect determination system and a cognitive inspection defect determination method that include an interest region designation unit for designating an interest region for defect determination, and by having a defect determination unit determine whether there is a defect based on the designated interest region, the product inspection efficiency and inspection speed can be improved, and the productivity and management convenience of the process line can be increased. means of solving the problem
[0006] In order to solve the problem described above, a cognitive inspection defect determination system according to one embodiment of the present invention is a cognitive inspection defect determination system applied to a cognitive inspection robot that performs a cognitive inspection of a product, comprising an image acquisition unit that acquires an image of a product from a camera, an image processing unit that processes the image acquired by the image acquisition unit, and a defect determination unit that determines whether the product is defective based on the image processed by the image processing unit, wherein the image processing unit includes a region of interest designation unit, and the defect determination unit determines the defect based on the information of the region of interest.
[0007] In addition, the area of interest designation unit is characterized by designating at least one of the entire area of one surface of the product, or a hole or groove formed in the product, as an area of interest.
[0008] In addition, the image processing unit is characterized by including a comparison value receiving unit that receives information about the means of transportation to which the product is applied and a standard image of the product according to the type of means of transportation.
[0009] In addition, the defect determination unit is characterized by deriving a similarity by comparing a standard image received from a comparison value receiving unit with an acquired image of a product received from an image processing unit, and determining that a defect has occurred if the similarity is less than a predetermined threshold.
[0010] In addition, it is characterized by further including a calculation unit that calculates the value of the region of interest of an acquired image received from an image processing unit and transmits the calculated value of the region of interest to a defect determination unit.
[0011] In addition, the calculation unit is characterized by calculating at least one of the total size of the product, the size of a specific hole formed in the product, the spacing between specific holes formed in the product, and the size and depth of a groove formed in the product.
[0012] In addition, the defect determination unit compares the numerical value of the standard image received from the comparison value receiving unit with the numerical value of the region of interest of the acquired image, and determines that a defect has occurred if the numerical value of the region of interest of the acquired image differs from the numerical value of the standard image by more than a predetermined amount.
[0013] In addition, the interest region designation unit designates two or more interest regions, and the defect determination unit compares the numerical value of the standard image and the numerical value of the interest region of the acquired image for each interest region, and is characterized by determining that a defect has occurred if at least one numerical value of the interest region of the acquired image differs from the numerical value of the standard image by a predetermined amount.
[0014] In addition, the image processing unit is characterized by further including a tone adjustment unit that adjusts the acquired image of the product to a gray tone and binarizes it.
[0015] In addition, the method is characterized by including: (a) a step in which an image acquisition unit acquires an image of a product to be identified as defective; (b) a step in which an image processing unit processes the image acquired in step (a); (c) a step in which a region of interest designation unit designates a region of interest; (d) a step in which a calculation unit calculates a value of the region of interest; and (e) a step in which a defect identification unit identifies a defect by comparing the acquired image and the value of the region of interest with a standard image.
[0016] Additionally, step (b) is characterized by including (b1) a step in which a comparison value receiver receives information of a means of transportation to which the product is applied, and (b2) a step in which a comparison value receiver receives a standard image of the product according to the type of means of transportation.
[0017] Additionally, step (b) is characterized by further including step (b3) in which a tone adjustment unit adjusts the tone of the image to a gray tone and binarizes it.
[0018] In addition, in step (c), the region of interest is characterized as being at least one of the entire surface area of the product, a hole or a groove formed in the product.
[0019] In addition, in step (d), the calculation unit is characterized by calculating at least one of the total size of the product, the size of a specific hole formed in the product, the spacing between specific holes formed in the product, and the size and depth of a groove formed in the product.
[0020] Additionally, step (e) includes (e1) a step of acquiring images of a standard image and an acquired image from an image processing unit, (e2) a step of loading the coordinates of a region of interest specified in step (c) onto the acquired image of the product and the standard image, and (e3) a step of deriving a similarity by comparing the standard image and the acquired image of the product received from the image processing unit, and the defect determination unit is characterized by determining that a defect has occurred if the similarity derived in step (e2) is less than a predetermined threshold.
[0021] Additionally, step (e) includes (e4) receiving a numerical value of a standard image and a numerical value of a region of interest of an acquired image from a calculation unit, and (e5) comparing the numerical value of a standard image with the numerical value of a region of interest of an acquired image of a product received from the calculation unit, and the defect determination unit determines that a defect has occurred if the numerical value of the region of interest of the acquired image differs from the numerical value of the standard image by more than a predetermined amount.
[0022] In addition, at step (c), two or more regions of interest are designated, and at step (e5), the numerical value of the standard image and the numerical value of the region of interest of the acquired image are compared for each region of interest, and if at least one numerical value of the region of interest of the acquired image differs from the numerical value of the standard image by a predetermined amount, it is determined that a defect has occurred. Effects of the invention
[0023] The cognitive inspection defect determination system and cognitive inspection defect determination method of the present invention, configured as described above, include an area of interest designation unit for designating an area of interest for defect determination, and by the defect determination unit determining whether there is a defect based on the designated area of interest, it has the effect of improving product inspection efficiency and inspection speed, and increasing the productivity and management convenience of the process line. Brief explanation of the drawing
[0024] Figure 1 is a schematic diagram illustrating the production process of a product. FIG. 2 is a block diagram illustrating the cognitive test defect determination system of the present invention. FIGS. 3 and 4 are schematic diagrams illustrating examples of areas of interest of the present invention for each product. FIG. 5 is a block diagram illustrating the image processing unit, calculation unit, and defect determination unit of the present invention. FIG. 6 is a schematic diagram illustrating the defect detection algorithm of the present invention. FIG. 7 is a flowchart illustrating the method for determining cognitive test defects according to the present invention. FIG. 8 is a flowchart illustrating the detailed steps of the image processing step in the cognitive test defect determination method of the present invention. FIG. 9 is a flowchart illustrating the detailed steps of the defect determination step in the cognitive test defect determination method of the present invention. Specific details for implementing the invention
[0025] Hereinafter, the technical concept of the present invention will be explained in more detail using the attached drawings. Prior to this, terms and words used in this specification and claims should not be interpreted as being limited to their ordinary or dictionary meanings, and should be interpreted in a meaning and concept consistent with the technical concept of the present invention, based on the principle that the inventor can appropriately define the concept of the terms to best describe his invention.
[0026] The basic configuration and usage environment of the cognitive inspection defect determination system (1000) of the present invention will be described below with reference to FIGS. 1 and 2.
[0027] As illustrated in FIG. 1, the recognition inspection defect determination system (1000) of the present invention can be applied to the production process of lightweight automotive die-casting parts of the A16*** alloy series. More specifically, the above-described part production process can be placed from a product (P) loading tray (T) onto a conveyor belt via a recognition inspection robot (R1) after the die-casting and trimming processes for forming the product (P). An image of the placed part can be captured by a recognition inspection vision camera (C) and transmitted to the recognition inspection robot (R1). Based on the transmitted image, the recognition inspection defect determination system (1000) of the present invention determines whether a defect has been detected. If a defect is detected, the part can be transferred to a defective loading tray (T), etc., via an extraction robot (R3) and discharged to the outside. Additionally, good products (P) that are not detected as defective can be transferred to a deburring robot (R2) to proceed with the deburring process.
[0028] More specifically, as illustrated in FIG. 2, the cognitive inspection defect determination system (1000) of the present invention is applied to the above-described cognitive inspection robot (R1) and may include an image acquisition unit (100) that acquires an image of a product (P) from a camera (C), an image processing unit that processes the image acquired by the image acquisition unit (100), and a defect determination unit (300) that determines whether the product (P) is defective based on the image processed by the image processing unit. At this time, the image processing unit includes an interest region designation unit (210), and the defect determination unit (300) can determine defects based on information of the interest region (ROI). Accordingly, the defect status of the product (P) can be detected by focusing on a specific interest region, and ultimately, the product (P) inspection efficiency and inspection speed can be improved, and the productivity and management convenience of the process line can be increased.
[0029] Hereinafter, the region of interest designation part (210) of the present invention will be explained in more detail with reference to FIGS. 3 and 4.
[0030] As illustrated in FIGS. 3 and 4, the region of interest designation unit (210) can designate at least one of the entire surface area of the product (P), or a hole or groove formed in the product (P), as the region of interest (ROI). More specifically, the region of interest designation unit (210) can designate the region of interest (ROI) based on an image of the product (P) normalized by the tone adjustment unit (230) described later. At this time, a center hole formed in the product (P) can be designated as the region of interest (ROI), and a phase fixing hole formed in the product (P) can be designated as the region of interest (ROI). In addition, in one example, the region of interest can be designated as a part prone to defects due to the mechanical limitations of the processing machine. Accordingly, the product (P) inspection efficiency can be further maximized.
[0031] At this time, the area of interest designation unit (210) may receive the location of the area of interest (ROI) from the operator, or the area of interest designation unit (210) may designate the area of interest (ROI) according to arbitrary rules. To receive the location of the area of interest from the operator, the area of interest designation unit (210) may transmit an image with the background removed to the operator, and the coordinates obtained when designating the area of interest thereafter may also use the coordinates with the background removed. The area of interest information designated in this way may be stored in a database as a single data set, including the vehicle type, index, starting coordinates (x, y), width, and height (ROIw, ROIh) received from the comparison value receiving unit (220) to be described later.
[0032] Hereinafter, the image processing unit and defect determination unit (300) of the present invention will be described in more detail with reference to FIGS. 5 and 6.
[0033] As illustrated in FIG. 5, the image processing unit may include a comparison value receiving unit (220) that receives information on the means of transportation to which the product (P) is applied and a standard image (I2) of the product (P) according to the type of means of transportation. More specifically, the comparison value receiving unit (220) may receive the type of vehicle to which the product (P) is applied, the standard image (I2) (target value) of the product (P) that differs therefrom, and the target value of the region of interest (ROI) value. Additionally, the region of interest designation unit (210) may adjust and designate the region of interest (ROI) based on the type of means of transportation received from the comparison value receiving unit (220). More specifically, among the shapes of the product (P) applied for each type of vehicle, a part where defects frequently occur or a characteristic part may be designated as the region of interest (ROI).
[0034] Additionally, the image processing unit may further include a tone adjustment unit (230) that adjusts the acquired image (I1) of the product (P) to a gray tone. The tone adjustment unit (230) may use a single threshold setting on the acquired image (I1) adjusted to a gray tone, convert the RGB Model to an HSV Model to minimize interference caused by lighting, and then perform Likeliho and normalization. Afterward, this can be displayed as a histogram graph. Subsequently, a Film Filter may be applied to acquire the inspection target in order to remove noise caused by lighting reflected from the background. At this time, the reference coordinates of the inspection target may be directly entered by the operator or may be numerical values stored in memory.
[0035] At this time, as illustrated in FIG. 6, the defect determination unit (300) may request coordinate information and a reference image from the database of the comparison value receiving unit (220) using vehicle type information as a search condition. Upon request, the reference image, index, starting coordinate, width, and height of the corresponding vehicle type may be combined into a single data set, and multiple data sets (equal to the number of regions of interest) may be received. Subsequently, the defect determination unit (300) may load the region of interest onto the standard image (I2) received from the comparison value receiving unit (220) and the acquired image (I1) of the product (P) received from the image processing unit (200), and derive a similarity by comparing the loaded image of the region of interest. If the similarity is less than a predetermined threshold, it may be determined that a defect has occurred.
[0036] At this time, five calculation formulas were used to derive similarity, and for the formulas other than the correlation coefficient method, the formulas provided by OpenCV can be used. More specifically, the correlation coefficient method, the TM-SQDIFF_NORMED method, and the TM_CCORR_NORMED method can produce results close to 1.0 when similarity is high, the TM_SQDIFF method can produce results close to 0 when similarity is high, and the TM_CCORR method can produce results of very large values when similarity is high.
[0037] Additionally, the cognitive inspection defect determination system (1000) of the present invention may further include a calculation unit (400) that calculates the value of the region of interest of an acquired image (I1) received from an image processing unit and transmits the calculated value of the region of interest to a defect determination unit (300). The calculation unit (400) can calculate at least one of the overall size of the product (P), the size of a specific hole formed in the product (P), the spacing between specific holes formed in the product (P), and the size and depth of a groove formed in the product (P).
[0038] At this time, the defect determination unit (300) compares the numerical value of the standard image (I2) received from the comparison value receiving unit (220) with the numerical value of the region of interest of the acquired image (I1) of the product (P) received from the calculation unit (400), and if the numerical value of the region of interest of the acquired image (I1) differs from the numerical value of the standard image (I2) by more than a predetermined error range, it can determine that a defect has occurred. At this time, the threshold value considering the error range can be directly input by the operator.
[0039] Furthermore, the interest area designation unit (210) can designate two or more interest areas, and the defect determination unit (300) compares the numerical value of the standard image (I2) and the numerical value of the interest area of the acquired image (I1) for each interest area, and if at least one numerical value of the interest area of the acquired image (I1) differs from the numerical value of the standard image (I2) by more than a predetermined error range, it can determine that a defect has occurred.
[0040] The method for determining cognitive test defects according to the present invention will be explained in more detail below with reference to FIGS. 7 to 9.
[0041] As illustrated in FIG. 7, the method for determining defects in a cognitive inspection according to the present invention may include: (a) a step in which an image acquisition unit (100) acquires an image of a product (P) to be determined as a defect; (b) a step in which an image processing unit (200) processes the image acquired in step (a); (c) a step in which a region of interest designation unit (210) designates a region of interest (ROI); (d) a step in which a calculation unit (400) calculates a value of the region of interest (ROI); and (e) a step in which a defect determination unit (300) determines a defect by comparing the acquired image and the value of the region of interest (ROI) with a standard image (I2).
[0042] Additionally, as illustrated in FIG. 8, step (b) may include (b1) a step in which a comparison value receiving unit (220) receives information about a means of transportation to which the product (P) is applied, and (b2) a step in which a comparison value receiving unit (220) receives a standard image (I2) of the product (P) according to the type of means of transportation. More specifically, the comparison value receiving unit (220) may receive the type of vehicle to which the product (P) is applied, the standard image (I2) (target value) of the product (P) that differs therefrom, and the target value of the region of interest (ROI) value. Based on this, in step (c) to be described later, the region of interest designation unit (210) may adjust and designate the region of interest (ROI) based on the type of means of transportation received from the comparison value receiving unit (220). More specifically, among the shapes of the product (P) applied for each type of vehicle, a part where defects frequently occur or a characteristic part may be designated as the region of interest (ROI).
[0043] Additionally, step (b) may further include a step (b3) in which a tone adjustment unit (230) adjusts the tone of the image to a gray tone. The tone adjustment unit (230) may use a single threshold setting on the acquired image (I1) adjusted to a gray tone, convert the RGB Model to an HSV Model to minimize interference caused by lighting, and then perform Likeliho and normalization. Afterward, this can be displayed as a histogram graph. Subsequently, a Flod Fil may be applied to acquire the inspection target to remove noise caused by lighting reflected from the background. At this time, the reference coordinates of the inspection target may be directly entered by the operator or may be numerical values stored in memory.
[0044] At this time, in step (c), the region of interest may be at least one of the entire surface area of the product (P), or a hole or groove formed in the product (P). More specifically, the region of interest designation unit (210) may designate the region of interest (ROI) based on an image of the product (P) normalized by the tone adjustment unit (230) described later. At this time, a center hole formed in the product (P) may be designated as the region of interest (ROI), and a phase fixing hole formed in the product (P) may be designated as the region of interest (ROI). In addition, in one example, the region of interest may be designated as a part prone to defects due to the mechanical limitations of the processing machine. Accordingly, the inspection efficiency of the product (P) can be further maximized.
[0045] At this time, the area of interest designation unit (210) may receive the location of the area of interest (ROI) from the operator, or the area of interest designation unit (210) may designate the area of interest (ROI) according to arbitrary rules. To receive the location of the area of interest from the operator, the area of interest designation unit (210) may transmit an image processed to remove blank spaces to the operator, and the coordinates obtained when designating the area of interest thereafter may also use coordinates with blank spaces removed. The area of interest information designated in this way may be stored in a database as a single data set, including the vehicle type, index, starting coordinates (x, y), width, and height (ROIw, ROIh) received from the comparison value receiving unit (220) to be described later.
[0046] Additionally, in step (d), the calculation unit (400) preferably calculates at least one of the total size of the product (P), the size of a specific hole formed in the product (P), the spacing between specific holes formed in the product (P), and the size and depth of a groove formed in the product (P).
[0047] Additionally, as illustrated in FIG. 9, step (e) may include (e1) acquiring images of a standard image (I2) and an acquired image (I1) from an image processing unit (200), (e2) loading the coordinates of a region of interest (ROI) onto the acquired image (I1) and the standard image (I2) of the product (P). More specifically, in step (e1), the defect determination unit (300) may request coordinate information and a reference image from the database of the comparison value receiving unit (220) using vehicle type information as a search condition. Upon request, a plurality of data (a number equal to the number of regions of interest) may be received by combining the reference image, index, starting coordinates, width, and height of the corresponding vehicle type into a single data. Subsequently, in step (e2), the defect determination unit (300) loads a region of interest onto a standard image (I2) received from the comparison value receiving unit (220) and an acquired image (I1) of the product (P) received from the image processing unit (200), and can derive a similarity by comparing the loaded image of the region of interest. If the similarity is less than a predetermined threshold, it can be determined that a defect has occurred.
[0048] Additionally, step (e) may include a step of deriving similarity by comparing (e3) a standard image (I2) with an acquired image (I1) of a product (P) received from an image processing unit (200). At this time, there are five calculation formulas used to derive similarity, and the formulas provided by OpenCV can be used for all formulas except the correlation coefficient method. More specifically, the method using the correlation coefficient, the TM-SQDIFF_NORMED method, and the TM_CCORR_NORMED method can produce a result close to 1.0 when similarity is high, the TM_SQDIFF method can produce a result close to 0 when similarity is high, and the TM_CCORR method can produce a result with a very large value when similarity is high. At this time, if the similarity derived in step (e3) of the defect determination unit (300) is less than a predetermined threshold, it can be determined that a defect has occurred.
[0049] Additionally, step (e) may include (e4) receiving the numerical value of the standard image (I2) and the numerical value of the region of interest (ROI) of the acquired image (I1) from the calculation unit (400), and (e5) comparing the numerical value of the standard image (I2) with the numerical value of the region of interest of the acquired image (I1) of the product (P) received from the calculation unit (400). At this time, the defect determination unit (300) may determine that a defect has occurred if the numerical value of the region of interest of the acquired image (I1) differs from the numerical value of the standard image (I2) by more than a predetermined amount. At this time, the threshold value considering the error range may be directly input by the operator.
[0050] Furthermore, in step (c), two or more regions of interest may be designated, and in step (e4), the numerical value of the standard image (I2) and the numerical value of the region of interest of the acquired image (I1) are compared for each region of interest, and if at least one numerical value of the region of interest of the acquired image (I1) differs from the numerical value of the standard image (I2) by more than a predetermined amount, it can be determined that a defect has occurred.
[0051] The technical concept of the present invention should not be interpreted as being limited to the above-described embodiments. Not only is the scope of application diverse, but various modifications are possible at the level of a person skilled in the art without departing from the essence of the invention claimed in the claims. Accordingly, such improvements and modifications fall within the scope of protection of the present invention insofar as they are obvious to a person skilled in the art. Explanation of the symbols
[0052] 1000 : Cognitive Test Defect Detection System 100 : Image acquisition unit 200 : Image processing unit 210 : Area of Interest Designation 220 : Comparison value receiver 230 : Tone adjustment section 300 : Defect detection unit 400 : Operation unit P : Product ROI: Area of Interest ROI1: Region of Interest ROI2: Region of Interest ROI3: Third Region of Interest ROI4: Region of Interest 4 ROI5: The 5th Region of Interest C: Camera R1: Cognitive testing robot R2 : Deburing Robot R3: Extraction robot T : Tray I1: Acquired Image I2: Standard image
Claims
Claim 1 A cognitive inspection defect determination system applied to a cognitive inspection robot that performs a cognitive inspection of a product, comprising: an image acquisition unit that acquires an image of a product from a camera; an image processing unit that processes the image acquired by the image acquisition unit; and a defect determination unit that determines whether the product is defective based on the image processed by the image processing unit; wherein the image processing unit includes a region of interest designation unit, and the defect determination unit determines the defect based on the information of the region of interest. Claim 2 A recognition inspection defect determination system according to claim 1, wherein the area of interest designation unit designates at least one of the entire area of one side of the product, a hole or a groove formed in the product as an area of interest. Claim 3 A cognitive inspection defect determination system according to claim 1, wherein the image processing unit includes a comparison value receiving unit that receives information on the means of transportation to which the product is applied and a standard image of the product according to the type of means of transportation. Claim 4 In claim 3, the defect determination unit is characterized by deriving a similarity by comparing the standard image received from the comparison value receiving unit with the acquired image of the product received from the image processing unit, and determining that a defect has occurred if the similarity is less than a predetermined threshold. Claim 5 A cognitive inspection defect determination system characterized by further including, in claim 4, a calculation unit that calculates the numerical value of the region of interest of the acquired image received from the image processing unit and transmits the calculated numerical value of the region of interest to the defect determination unit. Claim 6 A cognitive inspection defect determination system according to claim 5, wherein the calculation unit calculates at least one of the total size of the product, the size of a specific hole formed in the product, the spacing between specific holes formed in the product, and the size and depth of a groove formed in the product. Claim 7 A cognitive inspection defect determination system characterized in that, in claim 5, the defect determination unit compares the numerical value of the standard image received from the comparison value receiving unit with the numerical value of the region of interest of the acquired image, and determines that a defect has occurred if the numerical value of the region of interest of the acquired image differs from the numerical value of the standard image by a predetermined amount or more. Claim 8 A cognitive inspection defect determination system according to claim 7, wherein the interest region designation unit designates two or more interest regions, and the defect determination unit compares the numerical value of the standard image with the numerical value of the interest region of the acquired image for each interest region, and determines that a defect has occurred if at least one numerical value of the interest region of the acquired image differs from the numerical value of the standard image by a predetermined amount. Claim 9 A cognitive inspection defect determination system according to claim 1, wherein the image processing unit further includes a tone adjustment unit that adjusts the acquired image of the product to a gray tone and binarizes it. Claim 10 A method for determining defects in a cognitive inspection using a defect determination system of claim 1, comprising: (a) a step in which an image acquisition unit acquires an image of a product to be determined as a defect; (b) a step in which an image processing unit processes the image acquired in step (a); (c) a step in which a region of interest designation unit designates a region of interest; (d) a step in which a calculation unit calculates a numerical value of the region of interest; and (e) a step in which a defect determination unit determines a defect by comparing the acquired image and the numerical value of the region of interest with a standard image. Claim 11 A method for determining a cognitive inspection defect according to claim 10, wherein step (b) comprises: (b1) a step in which a comparison value receiving unit receives information of a means of transportation to which the product is applied; and (b2) a step in which the comparison value receiving unit receives a standard image of the product according to the type of means of transportation. Claim 12 A method for determining cognitive test defects according to claim 11, wherein the above step (b) further includes (b3) a step in which a tone adjustment unit adjusts the tone of an image to a gray tone and binarizes it. Claim 13 A method for determining defects in a cognitive inspection according to claim 10, wherein in step (c) above, the area of interest is at least one of the entire area of one side of the product, a hole or a groove formed in the product. Claim 14 A method for determining a defect in a cognitive inspection according to claim 10, wherein, in step (d) above, the calculation unit calculates at least one of the total size of the product, the size of a specific hole formed in the product, the spacing between specific holes formed in the product, and the size and depth of a groove formed in the product. Claim 15 In claim 10, the above step (e) comprises: (e1) acquiring images of a standard image and an acquired image from the above image processing unit; (e2) loading the coordinates of the region of interest specified in step (c) onto the acquired image of the product and the standard image; and (e3) deriving a similarity by comparing the standard image with the acquired image of the product received from the above image processing unit, wherein the defect determination unit determines that a defect has occurred when the similarity derived in step (e2) is less than a predetermined threshold. Claim 16 In claim 15, the above step (e) comprises (e4) receiving a numerical value of a standard image and a numerical value of a region of interest of an acquired image from the above calculation unit, and (e5) comparing the numerical value of the standard image with the numerical value of the region of interest of the acquired image of the product received from the above calculation unit, and the defect determination unit determines that a defect has occurred when the numerical value of the region of interest of the acquired image differs from the numerical value of the standard image by a predetermined amount or more. Claim 17 A method for determining defects in a cognitive inspection according to claim 16, wherein in step (c) above, two or more regions of interest are designated, and in step (e5) above, the numerical value of the standard image and the numerical value of the region of interest of the acquired image are compared for each region of interest, and if the numerical value of at least one of the regions of interest of the acquired image differs from the numerical value of the standard image by a predetermined amount or more, it is determined that a defect has occurred.