AI-based low-density foreign body detection system

Abstract

The present invention relates to an artificial intelligence-based low-density foreign object detection system, and more specifically, to an artificial intelligence-based low-density foreign object detection system capable of accurately distinguishing between scratches on the surface of a container, bubbles, and foreign objects inside a container based on continuous time-series captured frames of a rotating container. To this end, the present invention comprises: a container rotation device on which a container is mounted for detecting scratches on the surface or foreign substances in an aqueous solution contained therein, and which rotates the mounted container at a preset rotational speed; at least one vision device coupled to the container rotation device and which captures a rotating container (B) at regular time intervals to generate a plurality of captured frames; and a control unit that receives a plurality of captured frames from the vision device, preprocesses the corresponding captured frames, determines whether there are scratches on the surface of the container and determines whether there are foreign substances inside the container based on the preprocessed plurality of captured frames; wherein the control unit includes a scratch determination unit that determines whether there are scratches on the surface of the container based on the received plurality of captured frames, and a foreign substance determination unit that distinguishes between foreign substances and bubbles inside the container based on the received plurality of captured frames to determine whether there are foreign substances.

Description

Technology Field

[0001] The present invention relates to an artificial intelligence-based low-density foreign object detection system, and more specifically, to an artificial intelligence-based low-density foreign object detection system capable of accurately distinguishing between scratches on the surface of a container, bubbles, and foreign objects inside a container based on continuous time-series captured frames of a rotating container. Background Technology

[0003] When various types of liquids are contained within a container, the production facility performs a quality inspection to check for foreign substances after filling the container with the liquid.

[0004] A typical method for detecting foreign substances inside a transparent container containing liquid can be carried out by first rotating the container on a turntable and having an operator distinguish the rotating foreign substances inside the container through visual inspection.

[0005] Such foreign object detection methods are raising issues such as variations in inspection quality among operators, the possibility of human error, increased labor costs, and deterioration of software inspection quality; furthermore, due to insufficient accuracy and low reliability, the need for alternatives is becoming increasingly apparent.

[0006] Another method for inspecting foreign substances involves a conveyor-type foreign substance detection software. In this method, a container carried on a conveyor belt moves along a predetermined path, and a detector measures the intensity and amount of light passing through the liquid container. This information is then transmitted in real-time to a computer program, which can automatically detect foreign substances present within the liquid container using the foreign substance detection program.

[0007] However, in the case of such foreign substance inspection methods, in the case of containers containing liquid, bubbles may form within the liquid due to shaking of the internal liquid during transport on a conveyor belt, and in particular, bubbles frequently form within containers containing liquids such as pharmaceutical samples or carbonated beverages due to the beverage itself. There is a problem in that the method of detecting foreign substances by measuring changes in the inclination of the foreign substance inspection SW using an existing vision system cannot distinguish between foreign substances, floating particles, and external scratches that occur during the transport of liquid containers and bubbles.

[0008] In addition, existing foreign substance inspection software for transparent containers holding liquid may be suitable for inspecting high-density foreign substances, but there is a problem in that it fails to properly detect low-density foreign substances because it allows them to pass through.

[0009] Therefore, there is a need to develop an inspection device that can improve the stability and reliability of foreign object inspection within containers. Prior art literature

[0011] Korean Patent Publication No. 10-2023-0000477 (Publication Date: 2023.01.02) The problem to be solved

[0012] The present invention was devised to solve the above-mentioned problems and aims to provide an artificial intelligence-based low-density foreign matter detection system capable of accurately distinguishing between scratches on the container surface, bubbles, and foreign matter inside the container based on continuous time-series captured frames of a rotating container. means of solving the problem

[0014] The present invention has the following features to solve the above problem.

[0015] The present invention comprises: a container rotation device on which a container is mounted for detecting scratches on the surface or foreign substances in an aqueous solution contained therein, and which rotates the mounted container at a preset rotational speed; at least one vision device coupled to the container rotation device and which captures a rotating container (B) at regular time intervals to generate a plurality of captured frames; and a control unit that receives a plurality of captured frames from the vision device, preprocesses the corresponding captured frames, determines whether there are scratches on the surface of the container and determines whether there are foreign substances inside the container based on the preprocessed plurality of captured frames. The control unit includes a scratch determination unit that determines whether there are scratches on the surface of the container based on the received plurality of captured frames, and a foreign substance determination unit that distinguishes between foreign substances and bubbles inside the container based on the received plurality of captured frames to determine whether there are foreign substances.

[0016] Here, the vision device generates 6 to 10 shooting frames per second and transmits them to the control unit.

[0017] In addition, the scratch detection unit arranges a plurality of received captured frames in chronological order and compares neighboring images to generate a binarization result value based on whether there is an image change regarding a specific target within the image.

[0018] In addition, the foreign object identification unit includes a scraper unit that arranges a plurality of received shooting frames in chronological order, creates a layer stack that considers each shooting frame as a single layer and stacks them in a three-dimensional space, extracts the coordinates of a target determined to be a foreign object or a bubble from each layer, and calculates the amount of change in coordinates for each target by comparing the coordinates of the target extracted from an adjacent layer, and a determination unit that identifies the target as either a foreign object or a bubble based on the information on the amount of change in coordinates of the target over time received from the scraper unit.

[0019] In addition, the above-mentioned judgment unit determines, based on the hourly coordinate change information of the target, that a target that continues to rise or disappears after rising is a bubble, and that a target that continues to descend or gradually rises and then descends is a foreign object.

[0020] In addition, the scratch detection unit and the foreign matter detection unit include a machine learning model or a deep learning model that has been trained in advance, and input a plurality of captured frames into the machine learning model or the deep learning model in chronological order to obtain result information regarding bubbles and foreign matter. Effects of the invention

[0022] According to the present invention, there is an effect that scratches on the surface of the container and foreign substances and bubbles inside the container can be identified more accurately. Brief explanation of the drawing

[0024] FIG. 1 is a block diagram showing the schematic internal configuration of a foreign object inspection system according to an embodiment of the present invention. FIG. 2 is a perspective view showing a container rotation device according to an embodiment of the present invention. Figure 3 is a side view of Figure 2. FIG. 4 is a drawing showing a rotating module according to an embodiment of the present invention. FIG. 5 is a drawing showing a container pressurization module according to an embodiment of the present invention. FIG. 6 is a drawing showing a vision fixing module according to an embodiment of the present invention. FIG. 7 is a drawing showing a sensor module according to an embodiment of the present invention. FIG. 8 is a drawing showing the scratch determination unit determining a scratch according to an embodiment of the present invention. FIG. 9 is a drawing showing the distinction between foreign substances and bubbles in a foreign substance detection unit according to an embodiment of the present invention. Specific details for implementing the invention

[0025] The advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the accompanying drawings. However, the present invention is not limited to the embodiments disclosed below but may be implemented in various different forms. These embodiments are provided merely to ensure that the disclosure of the present invention is complete and to fully inform those skilled in the art of the scope of the present invention, and the present invention is defined only by the scope of the claims.

[0026] The terms used herein are for describing the embodiments and are not intended to limit the invention. In this specification, the singular form includes the plural form unless specifically stated otherwise in the text. As used herein, "comprises" and / or "comprising" do not exclude the presence or addition of one or more other components in addition to the components mentioned.

[0027] Throughout the specification, the same reference numerals refer to the same components, and "and / or" includes each of the mentioned components and all combinations of one or more. Although "first," "second," etc. are used to describe various components, these components are not limited by these terms.

[0028] These terms are used merely to distinguish one component from another. Therefore, it goes without saying that the first component mentioned below may be the second component within the technical scope of the present invention.

[0029] Unless otherwise defined, all terms used herein (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.

[0030] Spatially relative terms such as "below," "beneath," "lower," "above," and "upper" may be used to facilitate the description of the relationship between one component and other components as illustrated in the drawings. Spatially relative terms should be understood as encompassing different orientations of components during use or operation, in addition to the orientations depicted in the drawings. For example, if a component depicted in a drawing is inverted, a component described as "below" or "beneath" of another component may be placed "above" of that component. Therefore, the exemplary term "below" may encompass both the lower and upper directions. Components may also be oriented in other directions, and accordingly, spatially relative terms may be interpreted according to the orientation.

[0031] The terms “part” or “module” as used in the specification refer to hardware components such as software, FPGAs, or ASICs, and the “part” or “module” performs certain roles. However, the meaning of “part” or “module” is not limited to software or hardware. The “part” or “module” may be configured to reside in an addressable storage medium or may be configured to run one or more processors.

[0032] Accordingly, as an example, a "part" or "module" includes components such as software components, object-oriented software components, class components, and task components, as well as processes, functions, attributes, procedures, subroutines, segments of program code, drivers, firmware, microcode, circuits, data, databases, data structures, tables, arrays, and variables.

[0033] The functions provided within the components and "parts" or "modules" may be combined into a smaller number of components and "parts" or "modules," or further separated into additional components and "parts" or "modules."

[0034] Unless otherwise defined, all terms used herein (including technical and scientific terms) may be used in a meaning commonly understood by those skilled in the art to which the present invention pertains. Additionally, terms defined in commonly used dictionaries are not to be interpreted ideally or excessively unless explicitly and specifically defined otherwise.

[0035] FIG. 1 is a block diagram showing the schematic internal configuration of a foreign object inspection system according to an embodiment of the present invention, FIG. 2 is a perspective view showing a container rotation device according to an embodiment of the present invention, FIG. 3 is a side view of FIG. 2, FIG. 4 is a drawing showing a rotation module according to an embodiment of the present invention, and FIG. 5 is a drawing showing a container pressurization module according to an embodiment of the present invention.

[0036] In addition, FIG. 6 is a drawing showing a vision fixing module according to an embodiment of the present invention, FIG. 7 is a drawing showing a sensor module according to an embodiment of the present invention, FIG. 8 is a drawing showing a scratch detection unit determining a scratch according to an embodiment of the present invention, and FIG. 9 is a drawing showing a foreign substance detection unit distinguishing between a foreign substance and a bubble according to an embodiment of the present invention.

[0037] Referring to the drawings, a foreign substance inspection system (4000) according to one embodiment of the present invention is largely composed of: a container (B) for detecting scratches on the surface or foreign substances in a solution contained therein, a container rotation device (1000) that rotates the container (B) at a preset rotation speed, at least one vision device (2000) coupled to the container rotation device (1000) and capturing the rotating container (B) at regular time intervals to generate a plurality of captured frames, and a control unit (3000) that receives a plurality of captured frames from the vision device (2000), preprocesses the corresponding captured frames, determines whether there are scratches on the surface of the container (B) and determines whether there are foreign substances inside the container (B) based on the preprocessed plurality of captured frames.

[0038] Here, the container rotation device (1000) is configured to have a container (B) placed thereon to detect scratches on the surface or foreign substances in the solution contained therein, and to rotate the placed container (B) at a preset rotational speed. According to one embodiment of the present invention, such a container rotation device (1000) can be configured as shown in FIG. 2.

[0039] Referring to FIG. 2, the container rotation device (1000) is described as follows: the container rotation device (1000) according to one embodiment of the present invention comprises a base plate (100), a rotation module (200) that rotates a container (B) positioned on the base plate (100) and seated thereon, a container pressurizing module (300) that is coupled to the base plate (100) and installed so as to be adjustable to move up and down on the upper side of the rotation module (200) and supports rotation by accommodating a certain portion of the upper side of the container (B) seated on the rotation module (200), and a vision fixing module (500) that is coupled to the base plate (100) and coupled to the at least one vision device (2000) so that the vision device (2000) is fixedly supported.

[0040] Here, a rotation module (200), a container pressurization module (300), and a vision fixing module (500) are installed on the base plate (100).

[0041] It is preferable that the rotation module (200), container pressurizing module (300), and vision fixing module (500) installed on the base plate (100) have a support frame (310) of the container pressurizing module (300) installed between the upper sides centered on the rotation module (200), and that the vision fixing module (500) be installed at a position spaced apart toward the front.

[0042] Additionally, the rotation module (200) is positioned on the base plate (100) and is configured to rotate a container (B) that is seated thereon. This rotation module (200) comprises a cylindrical body (210) that supports the container (B), a support body (220) that accommodates the cylindrical body (210), a rotation motor (230) that provides rotational force to enable the support body (220) to rotate, and a rotation support assembly (240) that is coupled to the base plate (100) and rotates the support body (220).

[0043] Here, the above-mentioned cylinder (210) is composed of a rotating disc (211) that supports a container (B) and a rotating cylinder (212) that is coupled to the lower side of the rotating disc (211), has a hollow formed in the center, and has an open lower side.

[0044] As such, the reason for rotating the container (B) in the present invention is that when the container (B) rotates, bubbles are generated inside the container, and if foreign matter is contained inside the container, the foreign matter rises or falls while rotating inside the container. By photographing these bubbles, foreign matter, and scratches formed on the surface of the container through the vision device (2000), it becomes possible to distinguish between the bubbles, foreign matter, and scratches through the analysis of the photographic information.

[0045] In other words, conventional foreign object inspection devices inspected foreign objects by photographing a stationary container from various angles using multiple cameras; however, the photographic information obtained from such stationary containers makes it difficult to distinguish between foreign objects and scratches on the container surface, and there is a problem in that air bubbles generated during the container's movement may be mistaken for foreign objects.

[0046] Accordingly, the foreign object inspection device according to the present invention can clearly detect foreign objects inside a container and scratches on the surface of a container by photographing a container rotating in place and analyzing the photographic information to clearly distinguish between such foreign objects, bubbles, and scratches.

[0047] In addition, to more clearly distinguish foreign substances, bubbles, and scratches when photographing a rotating container (B), a lighting module (800) is provided in the present invention, and it is preferable that such a lighting module (800) be installed in the axial direction of the rotating container (B).

[0048] According to one example of the present invention, the lighting module (800) is provided at the bottom of the container (B), and for this purpose, the lighting module (800) is coupled with the rotational support assembly (240) of the rotational module (200).

[0049] In addition, it is preferable that the rotating disc (211) be formed of a transparent material so that light irradiated from the lighting module (800) can smoothly reach the container (B), and it is preferable that the rotating cylinder (212) have a hollow formed in the center and an open lower side.

[0050] In addition, it is preferable that the support body (220) accommodating the above-mentioned cylinder (210) also has an open lower central side so that light irradiated from the lighting module (800) reaches the container side via the rotating cylinder (212) and the rotating disc (211).

[0051] In addition, the rotational support assembly (240) is provided to be coupled to the base plate (100) to rotately support the support member (220). The rotational support assembly (240) is composed of a first assembly (241) which is coupled through the base plate (100) and accommodates the support member (220) inwardly, and is bearing-coupled to the support member (220), and a second assembly (242) which is coupled to the lower side of the base plate (100), accommodates the first assembly (241) inside, and supports the rotational motor (230) as the rotational motor (230) is coupled through the bottom surface inwardly.

[0052] In addition, it is preferable that the aforementioned lighting module (800) be coupled to the lower side of the first assembly (241) and the bottom surface of the second assembly (242).

[0053] Meanwhile, the container pressurizing module (300) is coupled to the base plate (100) and installed on the upper side of the rotation module (200) so as to be adjustable to move up and down, and is configured to support rotation by accommodating a certain portion of the upper side of the container (B) seated on the rotation module (200). This container pressurizing module (300) comprises a '∏' shaped support frame (310) with its lower ends respectively coupled to both sides of the rotation module (200) on the base plate (100), a length adjustment rod (320) that penetrates the upper center of the support frame (310) and is coupled so as to be rotatable, with a screw thread (322) formed on the lower outer edge for a certain length, a lifting plate (330) coupled to the lower screw thread (322) of the length adjustment rod (320) so as to be able to move up and down according to the rotation of the length adjustment rod (320), and a lower plate (330) coupled to one side of the lower surface of the lifting plate (330). It consists of a container pressurizing part (340) that protrudes to the lower side.

[0054] Here, as shown in FIGS. 1 and 2, the length adjustment rod (320) is preferably configured to have a rotating handle (321) formed at the upper end to facilitate rotation, and the length adjustment rod (320) is coupled to the support frame (310) so as to be rotatable in a fixed position.

[0055] Accordingly, when the length adjustment rod (320) rotates, the screw thread (322) formed on the lower outer edge of the length adjustment rod (320) rotates, and the lifting / lowering plate (330) that is screw-coupled with the screw thread (322) moves up and down.

[0056] In addition, the container pressurizing part (340) is coupled to one side of the lowering plate (330) and moves up and down together. This container pressurizing part (340) is composed of a pressurizing fixing body (341) that is coupled to one side of the lowering plate (330) and protrudes downward, and a bushing body (342) that is coupled to the lower end of the pressurizing fixing body (341) and has a receiving groove (342a) formed at the lower center that corresponds in shape to the upper side of the container (B) to receive a certain portion of the upper side of the container (B) and rotately support it.

[0057] At this time, according to an example of the present invention, a spring receiving tube (341a) is formed on the outside of the pressure fixing body (341), and a spring (341b) can be received within the spring receiving tube (341a).

[0058] At the bottom of this spring (341b), an elastic pressure bushing (343) capable of moving up and down according to pressure is provided, and this elastic pressure bushing (343) is coupled with the bushing body (342).

[0059] In addition, at this time, the bushing body (342) is coupled to the pressure fixing body (341) with a certain margin length in the upper and lower directions, and it is known that it has a critical position where it is no longer separated from the pressure fixing body (341) even if the pressure of the spring (341b) increases.

[0060] Accordingly, the container pressurizing part (340) pressurizes the upper side of the container (B) that rotates elastically, thereby enabling stable pressurization of the container. This prevents problems such as overheating and damage caused by excessive contact between the container (B) and the container pressurizing part (340), and problems such as excessive bubble generation caused by the container (B) rotating in the correct position shaking due to excessive contact force.

[0061] Meanwhile, the vision fixing module (500) is provided to fix and support a vision device (2000) that generates imaging information by photographing a rotating container (B) positioned apart from the container (B). This vision fixing module (500) is composed of a rail frame (510) having a longitudinal direction that is joined to a pair of each other facing each other on a base plate (100), a movable plate (520) installed to be movable along the longitudinal direction on the rail frame (510), an arc-shaped curvature frame (530) with one end joined to the upper side of the movable plate (520), and a vision module support part (540) that is movable along the longitudinal direction of the curvature frame (530) and fixed after moving to a selected position.

[0062] As shown in FIG. 8, the movable configuration between the movable plate (520) and the rail frame (510) is achieved by positioning a rotating screw rod (550) that penetrates the movable plate (520) between a pair of rail frames (510), rotatably fixing both ends of the rotating screw rod to separate fixed brackets, and screwing the rotating screw rod (550) and the movable plate (520) together so that the movable plate (520) can move according to the rotation of the rotating screw rod (550).

[0063] Depending on the movement of the moving plate (520), the distance between the vision module (400) and the container (B) can be adjusted.

[0064] In addition, the vision module support (540) can be configured to be fixed at a desired position after moving by forming a moving groove (531) on the central side along the longitudinal direction on the above-mentioned curvature frame (530).

[0065] Meanwhile, as illustrated in FIG. 7, a sensor module (700) may be provided on the rotation module (200) to detect whether the container (B) is normally positioned at a preset position.

[0066] The sensor module (700) is composed of a support bracket (710) coupled to one side of the support frame (310) and a detection sensor (720) coupled to the support bracket (710) to detect the presence of a container (B) located on the rotation module (200).

[0067] Such a detection sensor (720) can be, for example, an infrared sensor or an ultrasonic sensor, and through such a detection sensor (720), it can be determined that the container (B) is positioned in the correct position.

[0068] Meanwhile, at least one vision device (2000) is provided to generate a plurality of captured frames by photographing a rotating container (B) at regular time intervals, which is coupled to the container rotation device (1000). According to an example of the present invention, this vision device (2000) captures 6 to 10 captured frames per second.

[0069] The interval for generating shooting frames of such a vision device (2000) is preferably set in relation to the rotation speed of the container rotation device (1000). For example, when the container rotation device (1000) rotates at 400 rpm, it is preferable to generate 6 frames per second, and when it rotates at 600 rpm, it is preferable to generate 10 frames per second.

[0070] That is, in the present invention, it is preferable that each frame is generated when the container (B) rotates once on a specific plane and returns to its original position when viewed from the vision device (2000) toward the container (B).

[0071] In this way, the position of foreign matter or bubbles changes due to the generation of a frame by rotation of the container (B), but the scratches on the surface of the container (B) do not change, so the detection of scratches on the surface of the container (B) can be performed very accurately.

[0072] Of course, depending on the settings, even if the container (B) rotates at 400 rpm, it can be configured to generate 6 frames per second for scraping detection and then generate an additional 10 frames per second to more accurately determine the change in coordinates of bubbles and foreign matter.

[0073] Such a vision device (2000) transmits a plurality of captured frames to a control unit (3000), and the vision device (2000) and the control unit (3000) can be connected via wired or wireless connection.

[0074] For example, the vision device (2000) may communicate with the control unit (3000) using Bluetooth communication, BLE (Bluetooth Low Energy) communication, Near Field Communication unit, WLAN (Wi-Fi) communication, Zigbee communication, infrared (IrDA, infrared Data Association) communication, WFD (Wi-Fi Direct) communication, UWB (ultra wideband) communication, Ant+ communication, and WIFI communication methods, but the present invention is not limited thereto.

[0075] Meanwhile, the control unit (3000) receives a plurality of shooting frames from the vision device (2000), preprocesses the corresponding shooting frames, determines whether there are scratches on the surface of the container (B) and determines whether there are foreign substances inside the container (B) based on the preprocessed plurality of shooting frames. The control unit (3000) is composed of a scratch determination unit (3100) that determines whether there are scratches on the surface of the container (B) based on the received plurality of shooting frames, and a foreign substance determination unit (3200) that distinguishes between foreign substances and bubbles inside the container (B) based on the received plurality of shooting frames to determine whether there are foreign substances.

[0076] Here, the scratch determination unit (3100) arranges a plurality of received shooting frames in chronological order and compares neighboring images to generate a binarization result value based on whether there is an image change regarding a specific target within the image.

[0077] If there is no change in the image, a binarization result value of 0 is generated according to an example of the present invention, and if a change in the image occurs, a binarization result value of 255 is generated.

[0078] Accordingly, specific target objects generated with a binarization result of 0 are identified as scratches on the surface of the container, and specific target objects generated with 255 are identified as either bubbles or foreign matter.

[0079] Through this scratch detection unit (3100), the scratch portion within the shooting frame captured in chronological order is primarily identified, and for a specific target identified as a bubble or foreign object, the foreign object detection unit (3200) identifies each target as either a bubble or a foreign object.

[0080] In addition, the foreign substance detection unit (3200) is configured to determine the presence of foreign substances by distinguishing between foreign substances and bubbles within the container (B) based on a plurality of received shooting frames. As described above, the target is identified as either a bubble or a foreign substance through the information on the time-dependent coordinate change amount of a specific target determined to be a bubble or a foreign substance via the scratch detection unit (3100).

[0081] The foreign object identification unit (3200) is composed of a scraper unit (3210) that arranges a plurality of received shooting frames in chronological order, creates a layer stack by considering each shooting frame as a single layer and stacking them in a three-dimensional space, extracts the coordinates of a target determined to be a foreign object or a bubble from each layer, and calculates the amount of change in coordinates for each target by comparing the coordinates of the target extracted from an adjacent layer, and a judgment unit (3220) that determines whether the target is a foreign object or a bubble based on the information on the amount of change in coordinates of the target over time received from the scraper unit (3210).

[0082] At this time, the above judgment unit (3220) may be configured to determine, based on the time-dependent coordinate change information of the target, that a target that continues to rise or disappears after rising is a bubble, and that a target that continues to fall or gradually rises and then falls is a foreign substance.

[0083] The above-mentioned scratch detection unit (3100) and foreign substance detection unit (3200) may include a machine learning model or a deep learning model that has been pre-trained by inputting a large amount of prior data, that is, a large amount of captured frames.

[0084] Here, the machine learning learning model can be generated as a supervised learning model, an unsupervised learning model, and a reinforcement learning model, and the deep learning learning model can be generated as an ANN (Artificial Neural Network), RNN (Recurrent Neural Network), and CNN (Convolutional Neural Network), etc.

[0086] Although preferred embodiments of the present invention have been described above, the present invention is not limited to the specific embodiments described above. That is, those skilled in the art to which the present invention pertains can make numerous changes and modifications to the present invention without departing from the spirit and scope of the appended claims, and all such appropriate changes and modifications should be deemed to fall within the scope of the present invention as equivalents.

[0087] This application was prepared with the support of the Korea Information and Communication Technology Industry Promotion Agency for the purpose of supporting the growth of regional digital enterprises and the commercialization of regional software services. The information contained herein was prepared by the applicant, and it is stated that all rights and responsibilities related to the contents of this application belong to the applicant. Explanation of the symbols

[0089] 1000 : Foreign object inspection device 100 : Base plate 200 : Rotating module 210: Cylindrical body 211: Rotating disc 212: Rotating cylinder 220: Support body 230: Rotary motor 240: Rotary support assembly 241: First assembly 242: Second assembly 300: Container pressurization module 310: Support frame 320: Length adjustment rod 321: Rotating knob 322 : Screw thread 330 : Lifting / lowering plate 340: Container pressurizing part 341: Pressurizing fixing body 341a : Spring receiving tube 341b : Spring 342 : Bushing body 342a : Receiving groove 343: Elastic pressure bushing 500: Vision fixed module 510 : Rail frame 520 : Moving plate 530 : Curvature frame 531 : Move home 540: Vision module support 550: Rotary screw rod 700: Sensor module 710: Support bracket 720: Detection sensor 800: Lighting module 2000: Vision device 3000 : Control unit 3100: Scratch detection unit 3200: Foreign object detection unit 3210 : Scraper section 3220 : Judgment section 4000 : Foreign Object Inspection System B: Container

Claims

Claim 1 A container (B) for detecting scratches on the surface or foreign matter in the solution contained inside is mounted, and a container rotation device (1000) that rotates the mounted container (B) at a preset rotation speed; at least one vision device (2000) coupled to the container rotation device (1000) and capturing the rotating container (B) at regular time intervals to generate a plurality of captured frames; The invention includes a control unit (3000) that receives a plurality of captured frames from the vision device (2000), preprocesses the captured frames, determines whether there are scratches on the surface of the container (B) and determines whether there are foreign substances inside the container (B) based on the preprocessed plurality of captured frames; wherein the control unit (3000) includes a scratch determination unit (3100) that determines whether there are scratches on the surface of the container (B) based on the received plurality of captured frames, and a foreign substance determination unit (3200) that distinguishes between foreign substances and bubbles inside the container (B) based on the received plurality of captured frames to determine whether there are foreign substances; wherein the foreign substance determination unit (3200) arranges the received plurality of captured frames in chronological order, creates a layer stack that considers each captured frame as a single layer and stacks them in a three-dimensional space, extracts the coordinates of a target determined to be a foreign substance or a bubble from each layer, and then calculates the amount of change in coordinates for each target by comparing the coordinates of the target extracted from an adjacent layer. An artificial intelligence-based low-density foreign object detection system comprising a scraper unit (3210) and a judgment unit (3220) that determines whether a target is a foreign object or a bubble based on information regarding the change in coordinates of a target over time received from the scraper unit (3210), wherein the judgment unit (3220) determines, based on the information regarding the change in coordinates of a target over time, that a target that continues to rise or disappears after rising is a bubble, and a target that continues to fall or gradually rises and then falls is a foreign object. Claim 2 In claim 1, the vision device (2000) generates 6 to 10 shooting frames per second and transmits them to the control unit (3000), an artificial intelligence-based low-density foreign object detection system. Claim 3 In claim 1, the scratch determination unit (3100) arranges a plurality of received shooting frames in chronological order and compares neighboring images to generate a binarization result value based on whether there is an image change regarding a specific target within the image, an artificial intelligence-based low-density foreign object detection system. Claim 4 delete Claim 5 delete Claim 6 In claim 1, the scratch detection unit (3100) and the foreign object detection unit (3200) include a machine learning learning model or a deep learning learning model that has been learned in advance, and input a plurality of captured frames into the machine learning learning model or the deep learning learning model in chronological order to obtain result information regarding bubbles and foreign objects, an artificial intelligence-based low-density foreign object detection system.

Images