Automatic feeding and removing system for detecting visible foreign matters in injection

Abstract

The invention relates to the technical field of injection foreign matter detection, in particular to an automatic feeding and removing system for injection visible foreign matter detection, which comprises an automatic feeding and bottle arranging module for injection medicine bottles, a rotation-visual detection module and a tracking, removing and sorting module, the automatic feeding and removing system for detecting the visible foreign matters in the injection comprises an automatic feeding and bottle arranging module, a rotation-visual detection module and a tracking, removing and sorting module, and the automatic feeding and removing system is provided with the automatic feeding and bottle arranging module, the rotation-visual detection module and the tracking, removing and sorting module; the system can effectively identify low-contrast foreign matters, especially a non-contact pneumatic removal mode, so that the possibility of direct or indirect contact between operators and products is isolated, the possibility of microbial and exogenous particle pollution is greatly reduced, the high requirements of sterile preparation production are met, data can be audited and tracked, and the production efficiency is improved. GMP (Good Manufacturing Practice) data integrity and drug traceability regulation requirements are met.

Description

Technical Field

[0001] This invention relates to the field of foreign matter detection technology for injectable drugs, specifically an automatic feeding and rejection system for detecting visible foreign matter in injectable drugs. Background Technology

[0002] The visible foreign matter detection and rejection system for injectable drugs is an intelligent equipment that integrates precision mechanics and machine vision. It uses a light inspection scheme to detect and process foreign matter inside injection bottles. It can be used on injection production lines to detect and classify visible foreign matter in products such as vials, ampoules, and pre-filled syringes to ensure the quality of injectable drugs.

[0003] The existing feeding and rejection system for visible foreign matter detection in injectables uses a light inspection method to detect and process foreign matter inside the injection vials. This has the following problems: (1) In the detection process, human operation is the main risk point for microbial and particulate contamination; (2) Light inspection relies on statistical records, making it difficult to bind and trace the detection data of a single product, and it is inconvenient to audit the detection data; (3) The efficiency of visible foreign matter detection in the detection line is low. Therefore, an automatic feeding and rejection system for visible foreign matter detection in injectables is proposed to address the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide an automatic feeding and rejection system for the detection of visible foreign matter in injectable drugs, which addresses the problems that human operation is the main risk point for microbial and particulate contamination in the detection process, that light inspection is difficult to achieve data binding and traceability for individual products and that it is inconvenient to audit the detection data, and that the efficiency of visible foreign matter detection in the detection line is low.

[0005] To achieve the above objectives, the present invention provides the following technical solution: An automatic feeding and rejection system for detecting visible foreign matter in injectable drugs includes an automatic feeding and bottle handling module, a rotation-vision detection module, and a tracking rejection and sorting module. S1. Automatic feeding and bottle sorting module: a1. Feeding: The operator pours the whole tray or loose bottles into the hopper of the stainless steel lifting feeder with a capacity of 2000-5000 bottles; a2. Feeding and queuing: Medicine bottles are conveyed into a directional vibrating plate with a vibration frequency of 30-60 Hz, and pass through a spiral track and a screening device. A fiber optic sensor is installed at the outlet to confirm the bottle exit status. a3. Variable pitch conveying: The vibratory feeder outlet is connected to a bottle screw, and the screw pitch is matched with the bottle diameter; a4. Bottle Code Recognition: Before entering the inspection area, a QR code / barcode reader is set up. A fixed reader reads the product batch number, expiration date, and other information as the bottle passes through, and links this information to the subsequent inspection results. S2, Rotation-Vision Inspection Module: b1. Core Detection Area: The medicine bottles with equal spacing are conveyed into the dark chamber detection box, which is isolated from ambient light interference, by a conveyor belt or turntable. The dark chamber detection box adopts a V-block + synchronous belt clamping structure, and the clamping force is controlled at 8-15N to gently and firmly clamp the bottle body from both sides. b2. Defect Judgment: Rotation-Emergency Stop Processing: A servo motor drives a synchronous belt to rotate the medicine bottle at high speed. The inertia at the moment of emergency stop causes the foreign object to be suspended. Within 50-200 milliseconds after the bottle comes to a stop, the system sends a trigger signal to the camera and light source. Multi-camera simultaneous shooting and processing: Three high-resolution black and white industrial cameras with a resolution of 8 million to 12 million pixels and a frame rate sufficient to meet the production cycle are used. One of them serves as the main inspection camera: facing the bottle body, equipped with a telecentric lens, to obtain clear images of the main area of ​​the bottle body. The other two serve as auxiliary cameras: targeting the bottleneck and the rubber stopper field of view respectively. Image processing and judgment: The camera transmits the image to the image processing industrial control computer for image denoising, contrast enhancement and background correction, using multi-frame fusion and threshold segmentation analysis methods; S3, Tracking, Rejection, and Sorting Module: c1. Precise positioning and tracking: From the detection station to the rejection station, the medicine bottles move on the conveyor belt or turntable. The system tracks the position of each bottle in real time through a high-precision encoder. The detection results containing the bottle's unique ID or position information are stored in the information queue. c2. Pneumatic rejection: When a defective bottle arrives at the rejection station, the control system retrieves its information from the queue and triggers the pneumatic actuator to blow air to reject the defective bottle. c3. Diversion and collection: A waste chute is set up next to the main track. Waste bottles that are blown out enter the sealed waste collection box through the chute. Qualified products continue to move along the main track and enter the subsequent labeling and boxing process. The system automatically counts the waste and displays it on the HMI. An alarm is triggered when the waste box is full.

[0006] Preferably, in the automatic feeding and bottle sorting module of step S1, the sorting and queuing of bottles is carried out by means of a spiral track and a screening device to ensure that all bottles have their mouths facing upwards and are output in a single column.

[0007] Preferably, in the variable-pitch conveying of the automatic feeding and bottle sorting module in step S1, the dense bottle flow can be converted into a fixed gap, wherein the gap is set to 50-150mm according to the detection time.

[0008] Preferably, in the rotation-emergency stop processing of the rotation-visual detection module in step S2: the rotation speed is set to 800-1500 rpm, the rotation time is set to 1-2 seconds, and the motor itself brakes to bring the bottle to a complete stop within 100-300 milliseconds.

[0009] Preferably, in the multi-camera synchronous shooting process of the rotating-visual inspection module in step S2: the light source system of the above-mentioned industrial camera adopts a high-brightness, uniform LED surface light source as backlight to adapt to liquids of different colors and transparency, and the strobe illumination is strictly synchronized with the camera trigger to freeze the image.

[0010] Preferably, in the image processing and judgment of the rotation-visual detection module in step S2: a multi-frame fusion and threshold segmentation analysis method is adopted: for a single or multiple frames of still images, a grayscale threshold is set, suspicious points that are brighter than the background are identified, and the area, perimeter, aspect ratio and grayscale gradient of each suspicious point are calculated. The results are compared with a preset foreign object feature library to distinguish between real foreign objects such as glass shards, metal or fibers and false interference such as bubbles or inherent scratches on the bottle. Based on the size, quantity and type of the foreign object, combined with pharmacopoeia standards, the algorithm gives a "qualified" or "unqualified" signal within 10-30 milliseconds and attaches a defect code.

[0011] Preferably, in the pneumatic rejection process of the tracking rejection and sorting module in step S3: the rejection method uses 2-4 high-pressure pneumatic nozzles aligned with the lower part of the bottle body. Key parameters: the air source pressure is set to 0.4-0.7 MPa, the blowing time is set to a pulse drive of 10-50 milliseconds controlled by a high-speed solenoid valve, and the blowing timing is controlled by a closed loop of encoder position and bottle speed.

[0012] Compared with the prior art, the beneficial effects of the present invention are: 1. In this invention, through the automatic feeding and bottle handling module, the rotary-vision detection module, and the tracking rejection and sorting module set above, the system can effectively identify low-contrast foreign objects. It adopts a fully automated design, especially the non-contact pneumatic rejection method, which isolates the possibility of operators directly or indirectly contacting the product, greatly reducing the possibility of microbial and exogenous particulate contamination, meeting the high requirements of sterile preparation production, and the data is auditable and traceable, meeting the requirements of GMP data integrity and drug traceability regulations. 2. In this invention, through the automatic feeding and bottle unloading module, the rotary-vision inspection module, and the tracking rejection and sorting module set above, the system operates continuously at a stable and automated speed, increasing production efficiency by orders of magnitude. It can monitor equipment status in real time and continuously collect detection data. By replacing specific directional vibratory plates, bottle unloading bolts, and V-blocks + synchronous belts, and through digital formula management, the system can complete the switching production of different specifications and types of injections such as vials, ampoules, or pre-filled bottles in a short time, meeting the flexible production needs of small batches and multiple varieties. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the overall system module flow of the present invention; Figure 2 This is a schematic diagram illustrating the specific steps of the system module of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the invention.

[0015] Please see Figure 1-2 The present invention provides a technical solution: An automatic feeding and rejection system for detecting visible foreign matter in injectable drugs includes an automatic feeding and bottle handling module, a rotation-vision detection module, and a tracking rejection and sorting module. S1. Automatic feeding and bottle sorting module: a1. Feeding: The operator pours whole trays or loose bottles into the hopper of a stainless steel lifting feeder with a capacity of 2000-5000 bottles. The lifting feeder slowly rises, so that the bottle mouths gradually come into contact with the horizontal conveyor belt. The bottles are spread out and transported to the next stage by friction, avoiding collisions caused by waterfall-like falls. A photoelectric sensor is installed at the outlet of the hopper to monitor the bottle flow rate. When the material level is too low, an audible and visual alarm is triggered to prompt feeding. In the above process, a lifting feeder or a large hopper is used to automatically and smoothly transport a large number of bottles to the production line. a2. Feeding and Queuing: Medicine bottles are conveyed into a directional vibratory feeder with a spiral track and screening mechanism on the inner wall. The spiral track and screening device ensure that all bottles have their mouths facing upwards and are output in a single row. The vibration frequency and amplitude are controlled by a frequency converter. The vibration frequency is set to 30-60Hz and the amplitude is 0.5-2mm. The adjustment is made so that the bottles can "walk" smoothly without jumping. The spiral track is equipped with a "guide slot". Only bottles with their mouths facing upwards can pass through. Bottles with their mouths facing downwards will fall back to the bottom of the hopper for recirculation. The spiral track side wall is equipped with a baffle plate to ensure that the bottles move parallel to the track in the long axis direction. A fiber optic sensor is installed at the outlet to confirm the bottle output status and continuously monitor whether there are bottles output. If there are no bottles after a timeout, it is determined that the vibratory feeder is blocked and the system stops and alarms. a3. Variable pitch conveyor: The vibratory feeder outlet is connected to a bottle screw. The screw pitch matches the bottle diameter. The screw is driven by a servo motor and its speed is synchronized with the main production line. It can convert the dense bottle flow into a fixed gap. The gap is set to 50-150mm according to the detection time. A short section of unpowered free track is usually set behind the screw as a micro buffer to absorb the instantaneous speed difference between the preceding and following processes. a4. Bottle Code Recognition: Before entering the inspection area, a QR code / barcode reader is set up. When a bottle passes the photoelectric sensor, the reader takes a picture. The reader decodes the product batch number, serial number, and expiration date, and sends them to the host computer via Ethernet. The host computer generates a unique temporary ID for this bottle and binds the read code information with this ID, storing it in the database. This ID will follow the bottle until it is rejected or passes inspection and is linked to subsequent inspection results. S2, Rotation-Vision Inspection Module: b1. Core Inspection Area: The equally spaced medicine bottles are conveyed into a dark chamber inspection box that is isolated from ambient light interference via a conveyor belt or turntable. The interior of the dark chamber inspection box is coated with matte black paint to completely block external ambient light and is precisely guided to the rotating station. The interior of the dark chamber inspection box adopts a V-block + synchronous belt clamping structure with a clamping force controlled at 8-15N, gently and firmly clamping the bottle body from both sides. b2. Defect Judgment: Rotation-Emergency Stop Processing: A servo motor drives a synchronous belt to rotate the medicine bottle at high speed, with the rotation speed set to 800-1500 rpm and the rotation time set to 1-2 seconds. The motor itself brakes to bring the bottle to a complete stop within 100-300 milliseconds. The inertia at the moment of emergency stop causes the foreign object to suspend. After the emergency stop is completed, the industrial-grade PLC sets a programmable delay time Td between 50-200ms. The goal is to allow the foreign object suspended by inertia to move to the center of the liquid area, and the liquid surface eddies to basically subside, in the state of clearest imaging. After the delay ends, the PLC simultaneously sends a TTL high-level pulse signal with a pulse width of about 5ms to all cameras and light source controllers. Multi-camera synchronous shooting and processing: Three high-resolution monochrome industrial cameras with a resolution of 8-12 megapixels and a frame rate sufficient to meet the production cycle are used. One of them serves as the main inspection camera: facing the bottle body, equipped with a telecentric lens, to acquire clear images of the main body area of ​​the bottle. The other two serve as auxiliary cameras: targeting the bottleneck and the rubber stopper field of view respectively. Under the trigger of a pulse signal, the camera and the light source work together to capture images inside the bottle. The light source system of the above industrial cameras uses a high-brightness, uniform LED surface light source as backlight to adapt to liquids of different colors and transparency. The strobe illumination is strictly synchronized with the camera trigger to freeze the image. Image processing and judgment: The camera transmits the image to the image processing industrial control computer, which uses Halcon software, VisionPro software, OpenCV software or custom algorithm library to run the vision algorithm, analyze the image and make a pass / fail judgment: image denoising, contrast enhancement and background correction are performed, and multi-frame fusion and threshold segmentation analysis methods are used: for single or multiple frames of still images, a grayscale threshold is set, suspicious points that are brighter than the background are identified, the area, perimeter, aspect ratio and grayscale gradient of each suspicious point are calculated, and compared with the preset foreign object feature library to distinguish real foreign objects such as glass shards, metal or fiber from false interference such as bubbles or inherent scratches on the bottle. The algorithm gives a "pass" or "fail" signal within 10-30 milliseconds based on the size, quantity and type of foreign object, combined with pharmacopoeia standards, and attaches a defect code. Detailed algorithm flow: Image preprocessing: Flattening correction: Divide the image by a reference image of a blank bottle to eliminate the effects of uneven lighting and the inherent texture of the bottle.

[0016] Filtering and noise reduction: Apply medium-range filtering or Gaussian filtering to suppress random noise.

[0017] Region of Interest (ROI) definition: The precise outline of the liquid area in the image, excluding non-detection areas such as the bottle opening, bottle bottom, and label.

[0018] Defect extraction: Static threshold segmentation: For a single frame image after it has been statically rendered, a grayscale threshold is set, and all pixel regions with grayscale values ​​lower than this threshold are initially identified as "suspicious targets".

[0019] Feature analysis and classification: For each suspicious target, calculate a series of its characteristics: Area (pixels): converted to actual size (μm²).

[0020] Aspect ratio: distinguishes spherical microparticles (between 1 and 1.5) from fibers (>5).

[0021] Grayscale gradient / contour clarity: Real foreign objects have sharp edges, while bubbles have blurred edges.

[0022] Location: Whether it is near the bottle wall (possibly due to deposits) or in the center of the liquid.

[0023] The calculated feature vectors are compared with the "foreign object feature model" and "interference object feature model" pre-stored in the database based on rule judgment.

[0024] Example rule: IF (Area > 500 pixels AND Aspect Ratio < 3 AND Gray Scale Gradient > 50) THEN is determined as "glass shards".

[0025] Machine learning: Using SVM or CNN classifiers, directly output the probability of the foreign object type.

[0026] Overall judgment: Summarize the classification results of all suspicious targets.

[0027] The final judgment is made based on pharmacopoeia standards (e.g., particle number limits ≥ 10 μm) and the client's internal control standards.

[0028] For example: IF (any “glass shards” category foreign object present) OR (number of “fibers” > 2) OR (any foreign object size > 100μm) THEN is judged as “non-conforming”, defect code = 101.

[0029] Output: Within 30 milliseconds, the image processing software sends the judgment result (pass / fail + defect code) and the bottle's temporary ID to the industrial-grade PLC via TCP / IP or IO card; S3, Tracking, Rejection, and Sorting Module: c1. Precise positioning and tracking: From the detection station to the rejection station, the medicine bottles move on the conveyor belt or turntable. The system tracks the position of each bottle in real time through a high-precision encoder. The detection results containing the bottle's unique ID or position information are stored in the information queue. Specifically, the high-precision encoder resolution is 1000-5000 pulses per revolution, corresponding to a linear displacement accuracy of 0.1mm for the conveyor belt.

[0030] Establish a tracking queue: Industrial-grade PLCs maintain an internal first-in-first-out (FIFO) shift register as a tracking queue.

[0031] Enqueueing: Confirmed by photoelectric sensors, when a bottle leaves the detection station, the PLC stores its temporary ID and detection result in the first position of the queue.

[0032] Tracking: For every N pulses emitted by the encoder, the bottle moves a fixed distance. The distance the bottle moves in physical space corresponds strictly to the "distance steps" it moves in the queue. c2. Pneumatic Rejection: When a defective bottle arrives at the rejection station, the control system retrieves its information from the queue and triggers the pneumatic actuator to blow air to reject the defective bottle. The rejection method uses 2-4 high-pressure pneumatic nozzles aimed at the lower part of the bottle. Key parameters: air source pressure is set to 0.4-0.7 MPa, and the blowing time is set to a 10-50 millisecond pulse drive controlled by the PLC to the corresponding high-speed solenoid valve. Compressed air is sprayed out from the carefully designed nozzles through the instantaneously opened solenoid valve, forming a concentrated and short airflow. The airflow impacts the lower part of the bottle. The nozzle angle is set to 30-45 degrees with the horizontal, generating a torque that makes the bottle "fall" smoothly towards the waste chute instead of being blown away directly. The blowing timing is controlled by the encoder position and bottle speed in a closed loop. c3. Diversion and collection: A waste chute is set up next to the main track. Waste bottles that are blown out enter the sealed waste collection box through the chute. Qualified products continue to move along the main track and enter the subsequent labeling and boxing process. The system automatically counts the waste and displays it on the HMI. An alarm is triggered when the waste box is full.

[0033] This invention provides an automatic feeding and rejection system for detecting visible foreign matter in injectable drugs. It employs a central control system and data management, with an industrial-grade PLC as its control core. This PLC is responsible for the motion control of all electrical components, sensor signal processing, vision system processing, and HMI communication. The human-machine interface is a touchscreen HMI, providing the following functions: Parameter settings: bottle type, formula, size, rotation speed, light intensity, and detection sensitivity threshold.

[0034] Production monitoring: Real-time display of bottle / minute conveying speed, output, pass rate, and Pareto charts for various defects.

[0035] Image Review: It can store inspection images of all defective bottles for quality inspectors to review and for algorithm optimization.

[0036] Alarm Log: Records alarm information such as equipment malfunction, bottle blockage, and low gas pressure.

[0037] Data traceability: The system can generate a report on the coding information, visual images, and judgment results of each bottle, and upload it to the factory's MES system through interfaces such as OPC or UA to meet GMP data integrity requirements.

[0038] The system replaces the human eye with a high-resolution camera and a light source of a specific wavelength / angle. Combined with dynamic imaging excited by high-speed rotation, it can stably and repeatedly detect particles ≥50μm or even smaller, and effectively identify low-contrast foreign objects (such as transparent fibers). Its judgment is based on preset, quantifiable digital thresholds (such as area, aspect ratio, and grayscale gradient). It adopts a fully automated design, especially the non-contact pneumatic rejection method, which isolates the possibility of operators directly or indirectly contacting the product, greatly reducing the possibility of microbial and exogenous particle contamination. It meets the high requirements of sterile preparation production. The system starts with the identification (code reading) of a single vial and completely associates and stores the key parameters (rotation speed, emergency stop time, image, algorithm judgment result, defect classification, rejection action record) throughout the entire detection process. This forms an electronic quality file for each vial, and the data is auditable and traceable, fully meeting the requirements of GMP data integrity and drug traceability regulations.

[0039] The system operates continuously at a stable, automated speed, resulting in an order-of-magnitude increase in production efficiency. It integrates numerous photoelectric sensors, encoders, and self-diagnostic algorithms, enabling real-time monitoring of equipment status and providing early warnings or automatic compensation. By continuously collecting detection data, the system can leverage machine learning algorithms to continuously optimize its judgment model, reducing the false judgment rate, or adaptively adjust detection parameters according to the characteristics of different product types. Through the replacement of specific directional vibratory feeders, bottle-scraping bolts, and V-blocks with synchronous belts, as well as digital formula management, the system can quickly switch between different specifications and types of injections, such as vials, ampoules, or pre-filled vials, meeting the flexible production needs of small batches and multiple varieties.

[0040] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. An automatic feeding and rejection system for detecting visible foreign matter in injectable drugs, comprising an automatic feeding and sorting module for injectable drug vials, a rotation-vision detection module, and a tracking rejection and sorting module, characterized in that: The automatic feeding and rejection system for visible foreign matter detection in injectable drugs includes an automatic feeding and bottle handling module, a rotary vision inspection module, and a tracking rejection and sorting module. S1. Automatic feeding and bottle sorting module: a1. Feeding: The operator pours the whole tray or loose bottles into the hopper of the stainless steel lifting feeder with a capacity of 2000-5000 bottles; a2. Feeding and queuing: Medicine bottles are conveyed into a directional vibrating plate with a vibration frequency of 30-60 Hz, and pass through a spiral track and a screening device. A fiber optic sensor is installed at the outlet to confirm the bottle exit status. a3. Variable pitch conveying: The vibratory feeder outlet is connected to a bottle screw, and the screw pitch is matched with the bottle diameter; a4. Bottle Code Recognition: Before entering the inspection area, a QR code / barcode reader is set up. A fixed reader reads the product batch number, expiration date, and other information as the bottle passes through, and links this information to the subsequent inspection results. S2, Rotation-Vision Inspection Module: b1. Core Detection Area: The medicine bottles with equal spacing are conveyed into the dark chamber detection box, which is isolated from ambient light interference, by a conveyor belt or turntable. The dark chamber detection box adopts a V-block + synchronous belt clamping structure, and the clamping force is controlled at 8-15N to gently and firmly clamp the bottle body from both sides. b2. Defect Judgment: Rotation-Emergency Stop Processing: A servo motor drives a synchronous belt to rotate the medicine bottle at high speed. The inertia at the moment of emergency stop causes the foreign object to be suspended. Within 50-200 milliseconds after the bottle comes to a stop, the system sends a trigger signal to the camera and light source. Multi-camera simultaneous shooting and processing: Three high-resolution black and white industrial cameras with a resolution of 8 million to 12 million pixels and a frame rate sufficient to meet the production cycle are used. One of them serves as the main inspection camera: facing the bottle body, equipped with a telecentric lens, to obtain clear images of the main area of ​​the bottle body. The other two serve as auxiliary cameras: targeting the bottleneck and the rubber stopper field of view respectively. Image processing and judgment: The camera transmits the image to the image processing industrial control computer for image denoising, contrast enhancement and background correction, using multi-frame fusion and threshold segmentation analysis methods; S3, Tracking, Rejection, and Sorting Module: c1. Precise positioning and tracking: From the detection station to the rejection station, the medicine bottles move on the conveyor belt or turntable. The system tracks the position of each bottle in real time through a high-precision encoder. The detection results containing the bottle's unique ID or position information are stored in the information queue. c2. Pneumatic rejection: When a defective bottle arrives at the rejection station, the control system retrieves its information from the queue and triggers the pneumatic actuator to blow air to reject the defective bottle. c3. Diversion and collection: A waste chute is set up next to the main track. Waste bottles that are blown out enter the sealed waste collection box through the chute. Qualified products continue to move along the main track and enter the subsequent labeling and boxing process. The system automatically counts the waste and displays it on the HMI. An alarm is triggered when the waste box is full.

2. The automatic feeding and rejection system for detecting visible foreign matter in injectable drugs according to claim 1, characterized in that: In step S1, during the automatic feeding and bottle sorting module's sorting and queuing process: a spiral track and a screening device ensure that all bottles have their mouths facing upwards and are output in a single column.

3. The automatic feeding and rejection system for detecting visible foreign matter in injectable drugs according to claim 1, characterized in that: In step S1, during the variable-pitch conveying of the automatic feeding and bottle sorting module, the dense bottle flow can be converted into a fixed gap, where the gap is set to 50-150mm according to the detection time.

4. The automatic feeding and rejection system for detecting visible foreign matter in injectable drugs according to claim 1, characterized in that: In step S2, during the rotation-emergency stop process of the vision detection module, the rotation speed is set to 800-1500 rpm, the rotation time is set to 1-2 seconds, and the motor itself brakes to bring the bottle to a complete stop within 100-300 milliseconds.

5. The automatic feeding and rejection system for detecting visible foreign matter in injectable drugs according to claim 1, characterized in that: In step S2, during the multi-camera synchronous shooting process of the rotating-visual inspection module: the light source system of the above-mentioned industrial camera uses a high-brightness, uniform LED surface light source as the backlight to adapt to liquids of different colors and transparency. The strobe illumination is strictly synchronized with the camera trigger to freeze the image.

6. The automatic feeding and rejection system for detecting visible foreign matter in injectable drugs according to claim 1, characterized in that: In step S2, the image processing and judgment of the rotation-visual inspection module adopts a multi-frame fusion and threshold segmentation analysis method: for a single or multiple frames of still images, a grayscale threshold is set to identify suspicious points that are brighter than the background. The area, perimeter, aspect ratio, and grayscale gradient of each suspicious point are calculated and compared with a preset foreign object feature library to distinguish between real foreign objects such as glass shards, metal, or fibers and false interference such as bubbles or inherent scratches on the bottle. Based on the size, quantity, and type of foreign objects, combined with pharmacopoeia standards, the algorithm gives a "qualified" or "unqualified" signal within 10-30 milliseconds and attaches a defect code.

7. The automatic feeding and rejection system for detecting visible foreign matter in injectable drugs according to claim 1, characterized in that: In step S3, during the pneumatic rejection process of the tracking rejection and sorting module, the rejection method uses 2-4 high-pressure pneumatic nozzles aligned with the lower part of the bottle. Key parameters include: air source pressure set to 0.4-0.7 MPa, blowing time set to a 10-50 millisecond pulse drive controlled by a high-speed solenoid valve, and blowing timing controlled by a closed loop of encoder position and bottle speed.

Images