Vision-Based Smart Conveyor System for Food Filling

Abstract

The present invention relates to a vision-based smart conveyor system for food filling, which analyzes the condition of filling materials within a cup feeder transported along a conveyor based on vision recognition and artificial intelligence to determine quantitative accuracy, filling order, and the presence of foreign substances or inedible parts, and enables the automatic removal of abnormal filling materials based on the determination results. The present invention comprises a conveying unit that conveys a cup feeder filled with food at predetermined intervals; a vision recognition unit that is positioned along the conveying unit and acquires a real-time image of the filling state of the upper part of the cup feeder; an AI discrimination unit that includes a deep learning-based image analysis algorithm that analyzes an image input from the vision recognition unit to determine the quantitativeity, filling order, defects, foreign substances, or inedible parts of the filling; a control unit that controls the conveying operation of the cup feeder according to the analysis result of the AI ​​discrimination unit or commands the automatic discharge of the filling from the corresponding cup feeder if the filling does not satisfy standard conditions; and an ejection device that removes the abnormal filling according to the command of the control unit, so as to operate to immediately stop the conveying and automatically discharge the defective filling when an abnormality in the filling is detected.

Description

Technology Field

[0001] The present invention relates to automated equipment applied to a food filling process, and more specifically, to a vision-based smart conveyor system for food filling that analyzes the condition of filling materials within a cup feeder transported along a conveyor based on vision recognition and artificial intelligence to determine quantitative accuracy, filling order, and the presence of foreign substances or inedible parts, and automatically removes abnormal filling materials according to the determination results. Background Technology

[0002] Generally, in food manufacturing and packaging processes, various food ingredients are transported along a conveyor in the form of feeders filled with cups, containers, or pouches, while filling, sealing, and packaging are performed sequentially. In these processes, accurate quantitative measurement of the filling material, the proper filling sequence, and the presence of foreign substances or inedible parts are directly linked to product quality; therefore, accurately inspecting the filling condition is crucial.

[0003] In conventional food filling processes, the filling status has primarily been verified through weight sensors, weighing devices, or visual inspection by operators. While this method is partially effective in determining whether the total weight of the filling falls within a standard range, it has fundamental limitations in verifying the composition ratio of each food or the appropriateness of the filling order when the filling consists of various types of food.

[0004] In particular, conventional weight-based inspection methods had the problem of failing to detect defects even if specific food items were missing but other foods were overfilled, provided the total weight met the criteria. Furthermore, they had limitations in that they could not identify inedible parts or foreign substances if the weight change was minimal. Additionally, since information regarding the shape or arrangement of the food was not considered at all, it was difficult to apply these methods to determine whether there were visual abnormalities in the fillings.

[0005] Meanwhile, methods relying on visual inspection by workers have problems, such as judgment results varying depending on the inspector's skill level and the difficulty of simultaneously ensuring inspection speed and accuracy in mass production environments. Consequently, issues regarding the increased likelihood of inspection omissions or judgment errors, and the resulting decrease in production line efficiency, have been continuously raised.

[0006] Furthermore, conventional food filling equipment lacked the capability to automatically detect and immediately remove abnormalities in the filling process, often requiring operators to manually intervene or temporarily halt the line. This resulted in issues such as impaired process continuity and reduced productivity.

[0007] Furthermore, conventional technology lacked the ability to improve process conditions by accumulating or analyzing data on filling status determination results, which limited its ability to systematically analyze the causes of recurring filling defects or utilize them for long-term quality improvement. The problem to be solved

[0008] Accordingly, the present invention is proposed to resolve the aforementioned conventional problems, and the objective of the present invention is to provide a smart conveyor system for food filling that can accurately determine the quantitative accuracy of the filling material, the filling order, the presence of defects, and the presence of foreign substances or inedible parts by analyzing the condition of the food filled in the cup feeder in real time without relying on weight information or visual inspection by an operator.

[0009] In addition, another objective of the present invention is to provide a smart conveyor system for food filling that can effectively prevent the mixing of defective products while maintaining process continuity by immediately controlling the transfer of the corresponding cup feeder and automatically removing the defective filling material when an abnormality in the filling state is detected. means of solving the problem

[0010] To achieve the above objectives, a vision-based smart conveyor system for food filling according to the technical concept of the present invention comprises: a conveying unit that conveys a cup feeder filled with food at predetermined intervals; a vision recognition unit that is positioned along the conveying unit and acquires a vision of the filling state of the upper part of the cup feeder in real time; an AI discrimination unit that includes a deep learning-based image analysis algorithm that analyzes the image input from the vision recognition unit to determine the quantitative nature of the filling, the filling order, the defective area, foreign matter, or inedible area; a control unit that controls the conveying operation of the cup feeder according to the analysis result of the AI ​​discrimination unit or commands the automatic discharge of the filling from the corresponding cup feeder if the filling does not satisfy standard conditions; and an ejection device that removes the defective filling according to the command of the control unit, so that when an abnormality in the filling is detected, the conveying is immediately stopped and the defective filling is automatically discharged.

[0011] Here, the vision recognition unit may be characterized by including an RGB or RGB-D camera and further including an LED lighting unit and an illuminance correction module for stabilizing the shooting quality. Additionally, the vision recognition unit may be characterized by including a first camera and a second camera positioned at different angles to analyze the filling state inside the cup feeder from multiple angles, wherein the first camera photographs the top of the cup feeder and the second camera photographs the side or inclined portion of the cup feeder. Furthermore, the first camera and the second camera may be characterized by being installed in a fixed position or supported by a multi-axis adjustment bracket capable of adjusting the angle or height.

[0012] In addition, the AI ​​discrimination unit may be characterized by using a YOLO-based lightweight deep learning model to predict the quantitative accuracy, filling order, defect area, presence of foreign substances, and inedible area of ​​the filling material in real time. In addition, the AI ​​discrimination unit may be characterized by processing video frames in real time on a GPU module or an Edge-AI module and having a computational performance of 300 frames per second or more.

[0013] In addition, the control unit may be characterized by controlling the transfer of the cup feeder immediately to stop the transfer of the cup feeder and automatically discharge the filling from the cup feeder through an ejection device when the filling does not satisfy standard conditions, and by including a function to notify the operator of the abnormal condition through a warning light or HMI alarm.

[0014] In addition, the ejection device may be characterized by including a gripper for gripping the cup feeder and moving it to an inversion or discharge position.

[0015] In addition, the ejection device may be characterized by including a vacuum suction device for selectively removing only foreign substances or inedible parts from the filling material inside the cup feeder.

[0016] In addition, the ejection device may be characterized by having a gripper and a vacuum suction device alternately arranged along the cup feeder transfer direction to perform selective foreign matter removal or total reverse discharge depending on the state of the cup feeder.

[0017] In addition, the vacuum suction device may be characterized by further including an air blow or a small suction nozzle for selectively removing only the filling material in a designated area according to location information transmitted by the AI ​​identification unit.

[0018] In addition, the system may be characterized by including a data feedback module for storing data such as filler identification results, anomaly occurrence history, and removal frequency, and utilizing said data for model retraining of the AI ​​identification unit or process optimization.

[0019] In addition, the system may be characterized by including an industrial Ethernet or wireless communication module to link stored data with an MES, a manufacturing monitoring system, or a smart HACCP server.

[0020] In addition, the ejection device may be characterized by operating in at least one of inversion discharge, lower slide opening, and partial suction removal depending on the state of the cup feeder. Effects of the invention

[0021] The vision-based smart conveyor system for food filling according to the present invention analyzes the condition of food filled in a cup feeder in real time without relying on weight information or visual inspection by an operator, thereby having the effect of accurately determining the quantitative accuracy of the filling, the filling order, the presence of defects, and the presence of foreign substances or inedible parts.

[0022] In addition, the present invention has the effect of improving the reliability and production efficiency of the entire food filling process by immediately controlling the transfer of the cup feeder and automatically removing the defective filling material when an abnormality in the filling state is detected, thereby effectively preventing the mixing of defective products while maintaining process continuity. Brief explanation of the drawing

[0023] FIG. 1 is a block diagram illustrating the overall functional flow of a vision-based smart conveyor system for food filling according to an embodiment of the present invention. FIG. 2 is a perspective view illustrating the overall configuration of a vision-based smart conveyor system for food filling according to an embodiment of the present invention. FIG. 3 is a plan view illustrating the arrangement of a conveying unit, an ejection device, and a downstream process device in a vision-based smart conveyor system for food filling according to an embodiment of the present invention. Specific details for implementing the invention

[0024] A vision-based smart conveyor system for filling products according to embodiments of the present invention will be described in detail with reference to the attached drawings. Since the present invention is susceptible to various modifications and may take various forms, specific embodiments are illustrated in the drawings and described in detail in the text. However, this is not intended to limit the present invention to the specific disclosed forms, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the present invention. Similar reference numerals have been used for similar components in the description of each drawing. In the attached drawings, the dimensions of the structures are shown enlarged or reduced to the actual size to ensure clarity of the present invention or to understand the schematic configuration.

[0025] Additionally, terms such as "first," "second," etc., may be used to describe various components, but said components should not be limited by said terms. These terms are used solely for the purpose of distinguishing one component from another. For example, without departing from the scope of the present invention, the first component may be named the second component, and similarly, the second component may be named the first component. Meanwhile, unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the present invention pertains. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0026] <Example>

[0027] FIG. 1 is a block diagram illustrating the overall functional flow of a vision-based smart conveyor system for food filling according to an embodiment of the present invention, FIG. 2 is a perspective view illustrating the overall configuration of a vision-based smart conveyor system for food filling according to an embodiment of the present invention, and FIG. 3 is a plan view illustrating the arrangement of a conveying unit, an ejection device, and a downstream process device in a vision-based smart conveyor system for food filling according to an embodiment of the present invention.

[0028] As described above, the vision-based smart conveyor system for food filling according to an embodiment of the present invention comprises a conveying unit (10), a vision recognition unit (20), an AI discrimination unit (30), a control unit (40), and an ejection device (50) as main components, and is configured to analyze the condition of food filled in a cup feeder in real time without relying on weight information or visual inspection by an operator, accurately determine the quantitative accuracy of the filling material, the filling order, whether there are defects, and whether foreign substances or inedible parts are mixed in, and to immediately control the conveyance of the corresponding cup feeder and automatically remove the abnormal filling material when an abnormality in the filling condition is detected, thereby effectively preventing the mixing of defective products while maintaining the continuity of the process.

[0029] Hereinafter, a vision-based smart conveyor system for food filling according to an embodiment of the present invention will be described in detail, focusing on each of the above components.

[0030] The above-mentioned transfer unit (10) serves to accurately supply cup feeders filled with food to the operating positions of the vision recognition unit (20), the ejection device (50), and the downstream process devices by transferring them at predetermined intervals. This transfer unit (10) is configured to include a conveyor (101), and the conveyor (101) is formed to continuously move a plurality of cup feeders (102) along the transfer direction while supporting them at regular intervals. The cup feeder (102) is a container into which food is filled by a food filling machine (110), and is positioned so that while moving along the conveyor (101), the filling state is captured by the vision recognition unit (20), and normal transfer or ejection operation is performed under the control of the control unit (40) according to the analysis result of the AI ​​determination unit (30).

[0031] The above transfer unit (10) may further include a rotary (103) as needed, and the rotary (103) supports the cup feeder (102) rotatably to adjust the direction or position of the cup feeder (102). Accordingly, the cup feeder (102) can be more accurately aligned with the shooting position of the vision recognition unit (20) or the operating position of the ejection device (50), and the accuracy of recognizing the filling state and the reliability of the ejection operation can be improved.

[0032] Additionally, the transfer unit (10) is configured to stop or resume the operation of the conveyor (101) according to a control signal from the control unit (40). That is, when an abnormal condition is detected by the AI ​​identification unit (30) regarding the quantitative accuracy of the filling material, the filling order, the presence of defects, or the presence of foreign substances or non-edible parts, the control unit (40) transmits a stop signal to the transfer unit (10) to stop the corresponding cup feeder (102) at the inspection or ejection position. Through this, the ejection device (50) can be selectively operated only for the cup feeder (102) where the abnormal filling material was detected, and the continuity of the process can be maintained without unnecessary line stoppage.

[0033] The transfer unit (10) configured in this way is not limited to simply moving the cup feeder (102), but is organically linked with the vision recognition unit (20), AI discrimination unit (30), control unit (40), and ejection device (50) so that the filling status determination and defect removal process can be performed at the correct location and timing, thereby serving as a foundational configuration that improves the reliability and production efficiency of the entire vision-based smart conveyor system for food filling.

[0034] The above vision recognition unit (20) plays the role of acquiring images of the filling status of the upper and side of the cup feeder (102) being transported along the transport unit (10) in real time and providing them to the AI ​​discrimination unit (30). This vision recognition unit (20) is configured to accurately observe the condition of the food filled in the cup feeder (102) in a non-contact manner, and is configured to stably secure basic image data for determining the quantitative accuracy of the filling, the filling order, whether there are defects, and whether foreign substances or inedible parts are mixed in.

[0035] The above vision recognition unit (20) is configured to include a first camera (120a) and a second camera (120b), and the first camera (120a) is positioned toward the top of the cup feeder (102) to capture the overall distribution state, area ratio, and exposure state of the food filled inside the cup feeder (102). Accordingly, the first camera (120a) can effectively capture the upper shape, filling range, and arrangement state between the food of the food filled in the cup feeder (102).

[0036] Additionally, the second camera (120b) is positioned toward the side or inclined direction of the cup feeder (102) and serves to photograph the height, stacking state, and depth shape of the food filled inside the cup feeder (102). Accordingly, the second camera (120b) can complementarily identify the stacking structure of the food, the filling state of hidden areas, and whether specific food is missing or overfilled, which are difficult to confirm with only top-up photography. In this way, by having the first camera (120a) and the second camera (120b) photograph the same cup feeder (102) from different angles, a foundation is established to analyze the filling state from multiple angles, which is difficult to judge with a single field of view.

[0037] The first camera (120a) and the second camera (120b) may be composed of RGB or RGB-D cameras, and may further include an LED lighting unit and an illuminance correction module to prevent degradation of image quality due to the shooting environment. By maintaining uniform lighting conditions inside the cup feeder (102) through these LED lighting unit and illuminance correction module, image distortion caused by the color of the food, surface reflection, or the influence of an external light source is minimized, and the image analysis accuracy of the AI ​​discrimination unit (30) can be improved.

[0038] Additionally, the vision recognition unit (20) may be configured to have the first camera (120a) and the second camera (120b) installed in a fixed position, or supported by a multi-axis adjustment bracket that can adjust the angle or height. Accordingly, the shooting angle and field of view can be flexibly adjusted according to the size, shape, or type of food being filled in the cup feeder (102), and can be applied to various food filling processes.

[0039] The vision recognition unit (20) configured in this way enables precise recognition of the filling state of the cup feeder (102) based on images without relying on weight information or visual inspection by an operator, thereby performing the role of providing key input information that allows the AI ​​discrimination unit (30) to accurately determine the state of the filling material in a subsequent stage.

[0040] The AI ​​discrimination unit (30) receives an image of the filling state of the cup feeder (102) obtained from the vision recognition unit (20), determines the state of the filling material through a deep learning-based image analysis algorithm, and provides the determination result to the control unit (40). This AI discrimination unit (30) performs a key function in automating and improving the reliability of the food filling process by precisely determining the filling state based only on image information without relying on weight information or visual inspection by a worker.

[0041] The AI ​​discrimination unit (30) is configured to include a deep learning-based image analysis algorithm, and the deep learning-based image analysis algorithm is configured to analyze an image input from the vision recognition unit (20) in real time using a YOLO-based lightweight deep learning model. Accordingly, the AI ​​discrimination unit (30) can recognize image information of food filled inside the cup feeder (102) on an object basis and comprehensively analyze the location, size, shape, and relative arrangement of each object.

[0042] Specifically, the AI ​​determination unit (30) determines whether the filling amount falls within a standard range by determining the quantitative nature of the filling material, and determines whether the filling order matches a pre-set standard by considering the characteristics of a process in which multiple types of food are filled sequentially. In addition, to determine whether a specific food is missing, it analyzes the distribution state of the filling material relative to the space inside the cup feeder (102) and determines whether a part of the filling material is missing. Furthermore, to determine whether foreign substances or inedible parts are mixed in, the AI ​​determination unit (30) can identify objects having a color, shape, or texture different from the food and classify them as abnormal.

[0043] The AI ​​discrimination unit (30) can improve discrimination accuracy by comprehensively utilizing images of different viewing angles obtained from the first camera (120a) and the second camera (120b). That is, by analyzing planar information based on an image taken from above and height and stacking information based on an image taken from the side or an inclined direction together, it is possible to more reliably identify even missing fillers, overfilling, or obscured foreign substances that are difficult to confirm with only a single viewing angle.

[0044] Additionally, the AI ​​discrimination unit (30) can be configured to process video frames in real time using a GPU module or an Edge-AI module, thereby securing computational performance of more than 300 frames per second, and thus can derive discrimination results without delay even for the cup feeder (102) that moves continuously along the transfer unit (10). Due to this real-time processing performance, the discrimination result can be transmitted to the control unit (40) before the cup feeder (102) reaches the ejection device (50), and the continuity of the process is maintained.

[0045] The AI ​​discrimination unit (30) configured in this manner analyzes the condition of the filling material from various angles based on image information provided by the vision recognition unit (20) to generate a reliable discrimination result, and provides the result to the control unit (40), thereby acting as a core component of the present invention that enables a rapid and accurate response to abnormal filling material.

[0046] The above control unit (40) plays the role of comprehensively controlling the operation of the transfer unit (10) and the ejection device (50) based on the filling status determination result transmitted from the AI ​​determination unit (30). This control unit (40) serves as a central component that manages the overall operation flow of the vision-based smart conveyor system for food filling, and controls the normal transfer or removal of abnormal filling materials to be performed at the correct time and location depending on the filling status.

[0047] The control unit (40) receives a determination result from the AI ​​determination unit (30) regarding the quantitative accuracy of the filling material, the filling order, whether there are defects, and whether foreign substances or non-edible parts are mixed in, and determines whether the determination result satisfies the standard conditions. If the determination result is determined to be normal, the control unit (40) drives the transfer unit (10) normally so that the cup feeder (102) is continuously transferred to the downstream process devices.

[0048] On the other hand, if an abnormality is detected in the filling state, the control unit (40) transmits a stop or deceleration signal to the transfer unit (10) to control the corresponding cup feeder (102) to be accurately aligned with the operating position of the ejection device (50). Subsequently, the control unit (40) can selectively drive the ejection device (50) according to the type and condition of the abnormal filling material to invert discharge the food filled in the cup feeder (102) or selectively remove a portion of the filling material. Through this, selective removal targeting only defective filling material becomes possible, and unnecessary line stoppages can be minimized.

[0049] In addition, the control unit (40) may include a function to notify the operator in real time of the occurrence of an abnormality in the filling status through a warning light or an HMI alarm. Accordingly, the operator can immediately recognize an abnormal situation that occurred during the process and, if necessary, quickly take follow-up measures.

[0050] In addition, the control unit (40) may be configured to manage and store filling status determination results, abnormal occurrence history, ejection execution history, etc., in the form of data, and such data may be utilized for process condition analysis, quality control, and future process optimization. Furthermore, the control unit (40) may be connected to an industrial Ethernet or wireless communication module to link the stored data with an external MES, manufacturing monitoring system, or smart HACCP server.

[0051] The control unit (40) configured in this manner organically controls the operation of the transfer unit (10) and the ejection device (50) based on the judgment result of the AI ​​determination unit (30), thereby enabling a rapid and accurate response to abnormalities in the filling state and performing a role in improving the continuity, reliability, and overall production efficiency of the food filling process.

[0052] In addition, the control unit (40) stores the filling status determination result, abnormal occurrence history, and ejection execution history as data, and can utilize the data for model retraining of the AI ​​determination unit (30) or optimization of process conditions, and furthermore, can be linked with an MES, manufacturing monitoring system, or smart HACCP server through an industrial Ethernet or wireless communication module.

[0053] The above-mentioned ejection device (50) serves to remove or discharge abnormal filling material from a cup feeder (102) in which an abnormality in the filling state is detected according to a control command of the control unit (40). This ejection device (50) is configured so that different removal methods can be selectively performed depending on the filling state and abnormality type of the cup feeder (102).

[0054] Specifically, the ejection device (50) may operate in at least one of the following methods depending on the condition of the cup feeder (102): inversion discharge, lower slide opening, and partial suction removal. In the event that a serious abnormality occurs throughout the entire filling material, an inversion discharge method may be applied in which the cup feeder (102) is inverted to discharge the entire filling material in order to remove the food filled in the cup feeder (102) in a batch. Additionally, if the filling material can be discharged downwards depending on the structure or installation conditions of the cup feeder (102), a lower slide opening method may be applied in which the lower part of the cup feeder (102) is opened to remove the filling material. Meanwhile, if foreign substances or inedible parts are mixed in only a part of the filling material, a partial suction removal method may be applied in which only the filling material in that area is removed.

[0055] In order to implement such various ejection operation methods, as shown in the drawing, a gripper (130a) and a vacuum suction device (130b) may be installed together in the ejection device (50). The gripper (130a) and the vacuum suction device (130b) are arranged along the slide rail (130) and alternately arranged along the transfer direction of the cup feeder (102), so that they can be selectively operated at appropriate positions under the control of the control unit (40).

[0056] The gripper (130a) performs the function of inverting the cup feeder (102) or moving it to a discharge position while gripping the cup feeder (102). Accordingly, when removal of the entire filling material is required, the control unit (40) can quickly remove the defective filling material by driving the gripper (130a) to invert the cup feeder (102) or transfer it to a discharge position outside the process.

[0057] The vacuum suction device (130b) is configured to operate when only a portion of the filling material inside the cup feeder (102) needs to be removed, such as foreign substances or inedible parts. The vacuum suction device (130b) can access a designated area based on location information transmitted from the AI ​​identification unit (30) and selectively suction and remove only the filling material in that area, and may additionally include an air blower or a small suction nozzle as needed. Accordingly, only the abnormal parts can be removed while maintaining the normal filling material, thereby minimizing the loss of raw materials.

[0058] In this way, by installing the gripper (130a) and the vacuum suction device (130b) together in a single ejection device (50) and working in cooperation, the ejection device (50) can selectively perform an appropriate method among full inversion discharge, lower slide opening, or partial suction removal depending on the filling state and abnormal type. Accordingly, unlike conventional technology that relies on a single removal method, it can flexibly respond to various filling abnormal situations and effectively prevent the mixing of defective products while maintaining process continuity.

[0059] Additionally, as illustrated in the drawing, the vision-based smart conveyor system for food filling according to an embodiment of the present invention may have downstream process devices arranged continuously for the cup feeder (102) that has passed through the ejection device (50). Specifically, the food filler (110) performs the role of filling food into the cup feeder (102), and the food filling hopper (160) performs the role of stably supplying the food to be filled. Subsequently, the pouch feeder (140a) and the pouch opener (140b) supply and open the pouch for the packaging process, and the liquid filler (150) can additionally fill the liquid components into the cup feeder (102) or the pouch as needed. In addition, the first sealing machine (170a) and the second sealing machine (170b) seal the pouch in stages, and the cooling discharger (180) cools the sealed product and discharges it to the outside. These downstream process devices are not core components of the present invention, but can be configured to assist in enabling food filling, inspection, removal, and packaging processes to be performed continuously on a single line.

[0060] Although preferred embodiments of the present invention have been described above, the present invention may use various variations, modifications, and equivalents. It is clear that the present invention can be applied in the same way by appropriately modifying the above embodiments. Therefore, the above description does not limit the scope of the present invention, which is defined by the limitations of the following claims. Explanation of the symbols

[0061] 10: Transfer unit 20: Vision recognition unit 30: AI Discrimination Unit 40: Control Unit 50: Ejection device 101: Conveyor 102: Cup feeder 103: Rotary 110: Food filling machine 120a: First camera 120b: Second camera 130: Slide rail 130a: Gripper 130b: Vacuum suction device 140a: Pouch feeder 140b: Pouch opener 150: Liquid charger 160: Food filling hopper 170a: Primary sealing machine 170b: Secondary sealing machine 180: Cooling exhaust

Claims

Claim 1 A vision-based smart conveyor system for food filling comprises: a conveying unit that conveys cup feeders filled with food at predetermined intervals; a vision recognition unit arranged along the conveying unit to acquire a real-time image of the filling state of the upper portion of the cup feeders; an AI discrimination unit including a deep learning-based image analysis algorithm that analyzes the image input from the vision recognition unit to determine the quantitative accuracy, filling order, defects, foreign substances, or inedible portions of the filling; a control unit that controls the conveying operation of the cup feeders according to the analysis result of the AI ​​discrimination unit or commands the automatic discharge of the filling from the corresponding cup feeders if the filling does not satisfy standard conditions; and an ejection device that removes the abnormal filling according to the command of the control unit, wherein the conveying is stopped immediately and the defective filling is automatically discharged when an abnormality in the filling is detected, and wherein the ejection device is characterized by having a gripper and a vacuum suction device alternately arranged along the conveying direction of the cup feeders to perform selective foreign substance removal or total reverse discharge depending on the condition of the cup feeders. Claim 2 A vision-based smart conveyor system for food filling according to claim 1, wherein the vision recognition unit includes an RGB or RGB-D camera, and includes an LED lighting unit and an illuminance correction module for stabilizing shooting quality. Claim 3 A vision-based smart conveyor system for food filling according to claim 2, wherein the vision recognition unit includes a first camera and a second camera positioned at different angles to analyze the filling state inside the cup feeder from various angles, and the first camera photographs the top of the cup feeder and the second camera photographs the side or inclined portion of the cup feeder. Claim 4 A vision-based smart conveyor system for food filling, characterized in that, in paragraph 3, the first camera and the second camera are installed in a fixed position or are supported by a multi-axis adjustment bracket capable of adjusting the angle or height. Claim 5 A vision-based smart conveyor system for food filling according to claim 1, characterized in that the AI ​​discrimination unit predicts the quantitative accuracy, filling order, defect area, presence of foreign substances, and inedible area of ​​the filling material in real time using a YOLO-based lightweight deep learning model. Claim 6 A vision-based smart conveyor system for food filling, characterized in that, in claim 5, the AI ​​discrimination unit processes video frames in real time in a GPU module or an Edge-AI module and has a computational performance of 300 frames or more per second. Claim 7 A vision-based smart conveyor system for food filling according to claim 1, characterized in that the control unit immediately stops the transfer of the cup feeder when the filling material does not satisfy standard conditions, controls the automatic discharge of the filling material from the cup feeder through an ejection device, and includes a function to notify the operator of an abnormal condition through a warning light or HMI alarm. Claim 8 A vision-based smart conveyor system for food filling according to claim 1, wherein the ejection device includes a gripper for gripping a cup feeder and moving it to an inversion or discharge position. Claim 9 A vision-based smart conveyor system for food filling according to claim 1, wherein the ejection device includes a vacuum suction device for selectively removing only foreign substances or inedible parts from the filling material inside the cup feeder. Claim 10 delete Claim 11 A vision-based smart conveyor system for food filling according to claim 9, wherein the vacuum suction device comprises an air blow or a small suction nozzle for selectively removing only the filling material in a designated area according to location information transmitted by the AI ​​discrimination unit. Claim 12 A vision-based smart conveyor system for food filling according to claim 1, characterized in that the system stores filling material identification results, anomaly occurrence history, and ejection execution history as data, and includes a data feedback module for utilizing said data for model retraining of an AI identification unit or process optimization. Claim 13 A vision-based smart conveyor system for food filling according to claim 12, characterized in that the system includes an industrial Ethernet or wireless communication module to link stored data with an MES, a manufacturing monitoring system, or a smart HACCP server. Claim 14 A vision-based smart conveyor system for food filling according to claim 1, wherein the ejection device operates in at least one of inversion discharge, lower slide opening, and partial suction removal depending on the state of the cup feeder.

Images