CN114088730B - Method and system for detecting aluminum-plastic bubble cap defects by using image processing - Google Patents
Method and system for detecting aluminum-plastic bubble cap defects by using image processing Download PDFInfo
- Publication number
- CN114088730B CN114088730B CN202210077042.XA CN202210077042A CN114088730B CN 114088730 B CN114088730 B CN 114088730B CN 202210077042 A CN202210077042 A CN 202210077042A CN 114088730 B CN114088730 B CN 114088730B
- Authority
- CN
- China
- Prior art keywords
- training
- aluminum
- ccd camera
- roi
- defect
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000007547 defect Effects 0.000 title claims abstract description 88
- 238000012545 processing Methods 0.000 title claims abstract description 41
- 238000000034 method Methods 0.000 title claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 111
- 238000001514 detection method Methods 0.000 claims abstract description 89
- 238000013528 artificial neural network Methods 0.000 claims abstract description 39
- 239000003814 drug Substances 0.000 claims abstract description 23
- 238000013136 deep learning model Methods 0.000 claims abstract description 21
- 229940079593 drug Drugs 0.000 claims abstract description 13
- 238000007789 sealing Methods 0.000 claims description 25
- 238000004806 packaging method and process Methods 0.000 claims description 23
- XAGFODPZIPBFFR-UHFFFAOYSA-N aluminium Chemical compound [Al] XAGFODPZIPBFFR-UHFFFAOYSA-N 0.000 claims description 14
- 229910052782 aluminium Inorganic materials 0.000 claims description 14
- 230000002452 interceptive effect Effects 0.000 claims description 12
- 238000005096 rolling process Methods 0.000 claims description 11
- 230000002950 deficient Effects 0.000 claims description 5
- 239000011888 foil Substances 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims 1
- 230000000694 effects Effects 0.000 abstract description 5
- 238000004519 manufacturing process Methods 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 239000000843 powder Substances 0.000 description 2
- 241001292396 Cirrhitidae Species 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Analytical Chemistry (AREA)
- Bioinformatics & Computational Biology (AREA)
- Quality & Reliability (AREA)
- Evolutionary Biology (AREA)
- Signal Processing (AREA)
- Chemical & Material Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Biochemistry (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Image Analysis (AREA)
- Containers And Plastic Fillers For Packaging (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention provides a method and a system for detecting defects of an aluminum-plastic blister by using image processing, wherein the system comprises the following steps: the device comprises a first CCD camera, a second CCD camera and a control device, wherein the control device is used for controlling a trigger sensor and sending a trigger detection signal to the first CCD camera or the second CCD camera; the processor is connected with a machine vision processing tool, and the machine vision processing tool is used for obtaining the ROI needing deep learning model detection; the device comprises a detection model, a target detector, a writing module and a classification module; and the training model is obtained by establishing a training deep neural network for iterative computation according to the defect state of the historical sample. In order to realize the detection of the defects of the aluminum-plastic blister packaged medicines and achieve the best detection effect, a large number of defect samples are collected, and accurate marked data are trained through a deep neural network so as to obtain a deep learning model.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a system for detecting defects of an aluminum-plastic bubble cap by using image processing.
Background
Along with the continuous upgrading of the management practice (GMP) of the production quality of medicines in recent years, the requirement of the pharmaceutical industry on the production quality of medicines is higher and higher, and pharmaceutical manufacturers have urgent needs for improving the production efficiency on the premise of ensuring the quality, so that the installation of a high-efficiency and high-precision detection system on an aluminum-plastic blister packaging machine is imperative.
The manual detection mode can not meet the requirements of production efficiency and detection precision, and the traditional machine vision detection scheme can only detect empty particles, large unfilled corners and large black spots on the surface of the aluminum foil by adopting a detection method of combining color difference with area, and can not meet the detection requirements of light-colored or same-colored flaws such as small black spots, powder leakage, surface damage, powder inclusion and the like.
Disclosure of Invention
In view of the above, the present invention provides a method and a system for detecting defects of an aluminum-plastic blister by image processing.
The technical scheme adopted by the invention is as follows:
a method for detecting defects of an aluminum-plastic blister by using image processing comprises the following steps:
after the aluminum-plastic packaging machine feeds the pvc blister, triggering a first detection signal to be transmitted to a first CCD camera; the first CCD camera scans the pvc blister to acquire a first image of a first state;
triggering a second detection signal to be transmitted to a second CCD camera after the aluminum plastic heat-sealing rolling; the second CCD camera scans the near pvc surface to acquire a second image in a second state;
segmenting the first image and the second image, obtaining an ROI (region of interest) required to be detected by a deep learning model by using a machine vision processing tool, inputting the ROI into a detection model, and acquiring a corresponding detection result from a preset scheme obtained by training the detection model;
if the ROI does not obtain a corresponding detection result in the preset scheme, extracting the ROI, sending the ROI to a target detector to obtain an ROI attribute, writing identification features into the ROI attribute by using a writing module arranged in the target detector, and classifying the ROI into a preset sub-library with the same identification features in a training model according to the identification features;
according to the deep neural network branches used for training corresponding to the preset sub-library, the training model starts an avoidance mechanism, training resources of the training model are distributed to the deep neural network branches for performing iterative training on the ROI, and after training is completed, the training resources are distributed and stored to a subset corresponding to a preset scheme of the learning model according to the recognition features.
In the above, the training model is obtained according to the following method:
collecting a plurality of groups of defect samples, inputting the defect samples into a processing module to mark defect positions, recording defect states of the defect positions in the defect samples, simultaneously sending the defect samples to a target detector to obtain attributes of the defect samples, writing identification features in the attributes of the defect samples by using a writing module arranged in the target detector, and classifying the defect samples into preset sub-libraries with the same identification features according to the identification features;
correspondingly establishing a training deep neural network according to a preset sub-library, distributing training resources to the training deep neural network by a training model according to the size of a memory of the preset sub-library, performing iterative training on the defect samples in the corresponding preset sub-library, and distributing and storing the training deep neural network into a subset corresponding to a preset scheme of a learning model according to recognition characteristics after training.
In the above, the corresponding detection result is obtained in the preset scheme obtained by training the detection model, the detection result includes good products or defective products in different states, an ok or ng signal is sent to the removing device at the rear end according to the detection result, and the result is displayed on the interactive interface.
A system for detecting defects of an aluminum-plastic blister by image processing comprises
The first CCD camera is used for acquiring a first image of a first state after the aluminum-plastic packaging machine feeds the pvc blister;
a second CCD camera for a second image of a second state after the aluminum plastic heat sealing roll-in;
the control device is used for controlling the trigger sensor and sending a trigger detection signal to the first CCD camera or the second CCD camera;
a processor for performing segmentation processing on the first image and the second image;
the processor is connected with a machine vision processing tool, and the machine vision processing tool is used for obtaining an ROI needing deep learning model detection;
the detection model receives the ROI and acquires a corresponding detection result from a preset scheme obtained by training the detection model;
the target detector is used for extracting the ROI and acquiring the attribute of the ROI when the ROI does not acquire a corresponding detection result in a preset scheme;
a writing module, disposed in the target detector, for writing the identifying feature in the ROI attribute,
the classification module classifies the ROI into a preset sub-library with the same identification characteristics according to the identification characteristics;
and the training model is obtained by establishing a training deep neural network for iterative computation according to the defect state of the historical sample.
Preferably, the first CCD camera is arranged after the aluminum-plastic packaging machine feeds the pvc blister and before the aluminum-plastic heat-sealing roll-in, a first trigger sensor is arranged at the position where the aluminum-plastic packaging machine feeds the pvc blister, the first trigger sensor is used for triggering a first detection signal and sending the first detection signal to the control device, and the control device sends the first CCD camera for the first CCD camera to acquire a first image of a first state after the aluminum-plastic packaging machine feeds the pvc blister.
Preferably, the second CCD camera is arranged after the aluminum-plastic heat-sealing roll-in, a second trigger sensor is arranged at the tail end of the aluminum-plastic heat-sealing roll-in, the second trigger sensor is used for triggering a second detection signal and sending the second detection signal to the control device, and the control device sends the second CCD camera for the second CCD camera obtains a second image of a second state after the aluminum-plastic heat-sealing roll-in.
Preferably, a processing module, a recording module, a classification module, a plurality of preset sub-libraries and a plurality of training deep neural networks are arranged in the training model;
the processing module is used for marking the defect positions of a plurality of groups of collected defect samples;
the recording module is used for recording the defect state of the defect position in the defect sample;
the classification module is used for classifying the defect samples into preset sub-libraries with the same identification characteristics according to the identification characteristics written in the attributes of the defect samples;
the preset sub-library is used for storing corresponding defect samples according to the identification characteristics;
correspondingly establishing a training deep neural network according to a preset sub-library, distributing training resources to the training deep neural network by a training model according to the size of a memory of the preset sub-library, performing iterative training on the defect samples in the corresponding preset sub-library, and distributing and storing the training deep neural network into a subset corresponding to a preset scheme of a learning model according to recognition characteristics after training.
Preferably, an avoidance mechanism is further arranged in the training model, and the avoidance mechanism is used for selectively allocating training resources of the training model to a certain deep neural network branch.
Preferably, an avoidance mechanism is further arranged in the training model, and the avoidance mechanism is used for selectively allocating training resources of the training model to a certain training deep neural network.
Preferably, the system further comprises an interactive display control screen for displaying the result through the interactive interface.
After an aluminum-plastic packaging machine feeds a pvc blister, a first station is arranged before aluminum-plastic heat sealing, a first CCD camera is arranged at the first station and used for detecting a medicine flaw close to an aluminum foil surface; and a second station is arranged after the aluminum plastic heat sealing rolling, a second CCD camera is arranged at the second station and is used for detecting the medicine flaws on the near pvc surface, and the full coverage of the detection area of the two sides of the medicine is realized. The first CCD camera and the second CCD camera trigger the cameras to take pictures through in-place signals given by an aluminum-plastic packaging machine, an ROI (region of interest) needing deep learning model detection is segmented through image preprocessing, the ROI is transmitted into the model to be analyzed, whether the result is a good product or various defective products or not is predicted, an ok or ng signal is sent to the removing device according to the detection result, and the result is displayed on an interactive interface.
In order to realize the detection of the defects of the aluminum-plastic blister packaged medicines and achieve the best detection effect, a large number of defect samples are collected, accurate marked data are trained through a deep neural network, a deep learning model is further obtained, the effect of high-accuracy detection exceeding that of a human expert is achieved after multiple iterations, and finally the intelligent detection of the aluminum-plastic blister packaged medicines is achieved.
Drawings
The invention is illustrated and described only by way of example and not by way of limitation in the scope of the invention as set forth in the following drawings, in which:
FIG. 1 is a schematic block diagram of the system of the present invention;
FIG. 2 is a schematic diagram of a framework of a training model according to the present invention;
FIG. 3 is a flow chart of the detection according to the present invention;
FIG. 4 is a flow chart of training model establishment in the present invention.
Detailed Description
In order to make the objects, technical solutions, design methods, and advantages of the present invention more apparent, the present invention will be further described in detail by specific embodiments with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Explanation: the machine vision processing tool is HALCON of MVTec.
The ROI is a region of interest, which has a defect in the present invention.
Referring to fig. 1 to 2, the present invention provides a system for detecting defects of an aluminum-plastic blister by image processing, comprising
The first CCD camera is used for acquiring a first image of a first state after the aluminum-plastic packaging machine feeds the pvc blister;
a second CCD camera for a second image of a second state after the aluminum plastic heat sealing roll-in;
the control device is used for controlling the trigger sensor and sending a trigger detection signal to the first CCD camera or the second CCD camera;
a processor for performing segmentation processing on the first image and the second image;
the processor is connected with a machine vision processing tool, and the machine vision processing tool is used for obtaining an ROI needing deep learning model detection;
the detection model receives the ROI and acquires a corresponding detection result from a preset scheme obtained by training the detection model;
the target detector is used for extracting the ROI and acquiring the attribute of the ROI when the ROI does not acquire a corresponding detection result in a preset scheme;
a writing module, disposed in the target detector, for writing the identifying feature in the ROI attribute,
the classification module classifies the ROI into a preset sub-library with the same identification characteristics according to the identification characteristics;
and the training model is obtained by establishing a training deep neural network for iterative computation according to the defect state of the historical sample.
Preferably, the first CCD camera is arranged after the aluminum-plastic packaging machine feeds the pvc blister and before the aluminum-plastic heat-sealing roll-in, a first trigger sensor is arranged at the position where the aluminum-plastic packaging machine feeds the pvc blister, the first trigger sensor is used for triggering a first detection signal and sending the first detection signal to the control device, and the control device sends the first CCD camera for the first CCD camera to acquire a first image of a first state after the aluminum-plastic packaging machine feeds the pvc blister.
Preferably, the second CCD camera is arranged after the aluminum-plastic heat-sealing roll-in, a second trigger sensor is arranged at the tail end of the aluminum-plastic heat-sealing roll-in, the second trigger sensor is used for triggering a second detection signal and sending the second detection signal to the control device, and the control device sends the second CCD camera for the second CCD camera obtains a second image of a second state after the aluminum-plastic heat-sealing roll-in.
Preferably, a processing module, a recording module, a classification module, a plurality of preset sub-libraries and a plurality of training deep neural networks are arranged in the training model;
the processing module is used for marking the defect positions of a plurality of groups of collected defect samples;
the recording module is used for recording the defect state of the defect position in the defect sample;
the classification module is used for classifying the defect samples into preset sub-libraries with the same identification characteristics according to the identification characteristics written in the attributes of the defect samples;
the preset sub-library is used for storing corresponding defect samples according to the identification characteristics;
correspondingly establishing a training deep neural network according to a preset sub-library, distributing training resources to the training deep neural network by a training model according to the size of a memory of the preset sub-library, performing iterative training on the defect samples in the corresponding preset sub-library, and distributing and storing the training deep neural network into a subset corresponding to a preset scheme of a learning model according to recognition characteristics after training.
Preferably, an avoidance mechanism is further arranged in the training model, and the avoidance mechanism is used for selectively allocating training resources of the training model to a certain deep neural network branch.
Preferably, an avoidance mechanism is further arranged in the training model, and the avoidance mechanism is used for selectively allocating training resources of the training model to a certain training deep neural network.
Preferably, the system further comprises an interactive display control screen for displaying the result through the interactive interface.
After an aluminum-plastic packaging machine feeds a pvc blister, a first station is arranged before aluminum-plastic heat sealing, a first CCD camera is arranged at the first station and used for detecting a medicine flaw close to an aluminum foil surface; and a second station is arranged after the aluminum plastic heat sealing rolling, a second CCD camera is arranged at the second station and is used for detecting the medicine flaws on the near pvc surface, and the full coverage of the detection area of the two sides of the medicine is realized. The first CCD camera and the second CCD camera trigger the cameras to take pictures through in-place signals given by an aluminum-plastic packaging machine, an ROI (region of interest) needing deep learning model detection is segmented through image preprocessing, the ROI is transmitted into the model to be analyzed, whether the result is a good product or various defective products or not is predicted, an ok or ng signal is sent to the removing device according to the detection result, and the result is displayed on an interactive interface.
In order to realize the detection of the defects of the aluminum-plastic blister packaged medicines and achieve the best detection effect, a large number of defect samples are collected, accurate marked data are trained through a deep neural network, a deep learning model is further obtained, the effect of high-accuracy detection exceeding that of a human expert is achieved after multiple iterations, and finally the intelligent detection of the aluminum-plastic blister packaged medicines is achieved.
Referring to fig. 3, in addition, the present invention also provides a method for detecting defects of an aluminum-plastic blister by using image processing, comprising the following steps:
after the aluminum-plastic packaging machine feeds the pvc blister, triggering a first detection signal to be transmitted to a first CCD camera; the first CCD camera scans the pvc blister to acquire a first image of a first state;
triggering a second detection signal to be transmitted to a second CCD camera after the aluminum plastic heat-sealing rolling; the second CCD camera scans the near pvc surface to acquire a second image in a second state;
the method comprises the steps of segmenting a first image and a second image, obtaining an ROI (region of interest) needing deep learning model detection by using a machine vision processing tool, inputting the ROI into a detection model, obtaining a corresponding detection result from a preset scheme obtained by detection model training, obtaining a corresponding detection result from the preset scheme obtained by detection model training, sending ok or ng signals to a removing device at the rear end according to the detection result, and displaying the result on an interactive interface.
Referring to fig. 4, in the above, the training model is obtained according to the following method:
collecting a plurality of groups of defect samples, inputting the defect samples into a processing module to mark defect positions, recording defect states of the defect positions in the defect samples, simultaneously sending the defect samples to a target detector to obtain attributes of the defect samples, writing identification features in the attributes of the defect samples by using a writing module arranged in the target detector, and classifying the defect samples into preset sub-libraries with the same identification features according to the identification features;
correspondingly establishing a training deep neural network according to a preset sub-library, distributing training resources to the training deep neural network by a training model according to the size of a memory of the preset sub-library, performing iterative training on the defect samples in the corresponding preset sub-library, and distributing and storing the training deep neural network into a subset corresponding to a preset scheme of a learning model according to recognition characteristics after training.
When detecting, after the pvc blister is fed to the aluminum-plastic packaging machine, a first station is arranged before aluminum-plastic heat sealing, a first trigger sensor is arranged at the first station and used for triggering a first detection signal and sending the first detection signal to a control device, and the control device is sent to a first CCD camera and used for acquiring a first image of a first state after the pvc blister is fed to the aluminum-plastic packaging machine. And a second station is arranged after the aluminum-plastic heat-sealing rolling, a second trigger sensor is arranged at the second station and is used for triggering a second detection signal and sending the second detection signal to a control device, and the control device is sent to a second CCD camera and is used for acquiring a second image of a second state after the aluminum-plastic heat-sealing rolling by the second CCD camera. The full coverage of the detection areas on the two sides of the medicine is realized. The first CCD camera and the second CCD camera trigger the cameras to take pictures through in-place signals given by an aluminum-plastic packaging machine, an ROI (region of interest) needing deep learning model detection is segmented through image preprocessing, the ROI is transmitted into the model to be analyzed, whether the result is a good product or various defective products or not is predicted, an ok or ng signal is sent to the removing device according to the detection result, and the result is displayed on an interactive interface.
If the flaw does not obtain a corresponding detection result in the preset scheme, extracting the flaw, sending the flaw to a target detector to obtain a flaw attribute, writing identification features in the flaw attribute by using a writing module arranged in the target detector, and classifying the flaw into a preset sub-library with the same identification features in a training model according to the identification features;
according to the deep neural network branches used for training corresponding to the preset sub-library, the training model starts an avoidance mechanism, training resources of the training model are distributed to the deep neural network branches for iterative training of the flaws, and after training is completed, the training resources are distributed and stored to the subsets corresponding to the preset scheme of the learning model according to the recognition features.
Having described embodiments of the present invention, the foregoing description is intended to be exemplary, not exhaustive, and not limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen in order to best explain the principles of the embodiments, the practical application, or improvements made to the technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (7)
1. A method for detecting defects of an aluminum-plastic blister by using image processing is characterized by comprising the following steps:
after the aluminum-plastic packaging machine feeds the medicines to the pvc blister, triggering a first detection signal to be transmitted to a first CCD camera; the method comprises the following steps that a first CCD camera scans a pvc blister to obtain a first image of a first state, so that the first image is used for detecting a medicine flaw close to an aluminum foil surface;
triggering a second detection signal to be transmitted to a second CCD camera after the aluminum plastic heat-sealing rolling; the second CCD camera scans the near pvc surface to acquire a second image of a second state, so that the second image is used for detecting the medicine flaws of the near pvc surface;
the first image and the second image are segmented, a machine vision processing tool is used for obtaining an ROI (region of interest) required to be detected by a deep learning model, the ROI is input into the deep learning model, and a corresponding detection result is obtained through a preset scheme obtained by training the deep learning model;
if the ROI does not obtain a corresponding detection result in the preset scheme, extracting the ROI, sending the ROI to a target detector to obtain an ROI attribute, writing identification features into the ROI attribute by using a writing module arranged in the target detector, and classifying the ROI into a preset sub-library with the same identification features in a training model according to the identification features;
according to the deep neural network branches used for training corresponding to the preset sub-library, the training model starts an avoidance mechanism, training resources of the training model are distributed to the deep neural network branches for performing iterative training on the ROI, and after training is completed, the training resources are distributed and stored to a subset corresponding to a preset scheme of the deep learning model according to the recognition features;
the training model is obtained according to the following method:
collecting a plurality of groups of defect samples, inputting the defect samples into a processing module to mark defect positions, recording defect states of the defect positions in the defect samples, simultaneously sending the defect samples to a target detector to obtain attributes of the defect samples, writing identification features in the attributes of the defect samples by using a writing module arranged in the target detector, and classifying the defect samples into preset sub-libraries with the same identification features according to the identification features;
correspondingly establishing a training deep neural network according to a preset sub-library, distributing training resources to the training deep neural network branches by a training model according to the size of the memory of the preset sub-library, performing iterative training on the defect samples in the corresponding preset sub-library, and distributing and storing the training deep neural network branches into subsets corresponding to the preset scheme of the deep learning model according to the recognition characteristics after training.
2. The method for detecting the defects of the aluminum-plastic blister by using the image processing as claimed in claim 1, wherein the corresponding detection results are obtained by a preset scheme obtained by deep learning model training, the detection results comprise good products or defective products in different states, ok or ng signals are sent to a removing device at the rear end according to the detection results, and the results are displayed on an interactive interface.
3. A system for detecting defects of an aluminum-plastic blister by using image processing is characterized by comprising
The first CCD camera is used for scanning the pvc blister after the aluminum plastic packaging machine feeds the medicines to the pvc blister to acquire a first image of a first state, so that the first CCD camera is used for detecting the defects of the medicines close to the aluminum foil surface;
the second CCD camera is used for scanning the near pvc surface after the aluminum plastic heat sealing rolling so as to obtain a second image in a second state, and therefore the second CCD camera is used for detecting the medicine flaws of the near pvc surface;
the control device is used for controlling the trigger sensor and sending a trigger detection signal to the first CCD camera or the second CCD camera;
a processor for performing segmentation processing on the first image and the second image;
the processor is connected with a machine vision processing tool, and the machine vision processing tool is used for obtaining an ROI needing deep learning model detection;
the deep learning model is used for receiving the ROI and acquiring a corresponding detection result through a preset scheme obtained by training the deep learning model;
the target detector is used for extracting the ROI and acquiring the attribute of the ROI when the ROI does not acquire a corresponding detection result in a preset scheme;
a writing module, disposed in the target detector, for writing the identifying feature in the ROI attribute,
the classification module classifies the ROI into a preset sub-library with the same identification characteristics according to the identification characteristics;
the training model is obtained by establishing a training deep neural network for iterative calculation according to the defect state of the historical sample;
the training model is internally provided with a processing module, a recording module, a classification module, a plurality of preset sub-libraries and a plurality of training deep neural networks;
the processing module is used for marking the defect positions of a plurality of groups of collected defect samples;
the recording module is used for recording the defect state of the defect position in the defect sample;
the classification module is used for classifying the defect samples into preset sub-libraries with the same identification characteristics according to the identification characteristics written in the attributes of the defect samples;
the preset sub-library is used for storing corresponding defect samples according to the identification characteristics;
correspondingly establishing a training deep neural network according to a preset sub-library, distributing training resources to the training deep neural network branches by a training model according to the size of the memory of the preset sub-library, performing iterative training on the defect samples in the corresponding preset sub-library, and distributing and storing the training deep neural network branches into subsets corresponding to the preset scheme of the deep learning model according to the recognition characteristics after training.
4. The system for detecting defects of aluminum-plastic blisters by using image processing as claimed in claim 3, wherein the first CCD camera is disposed at a first station, the first station is located after the aluminum-plastic packaging machine feeds the drugs to the pvc blisters and before the aluminum-plastic heat-sealing roll-pressing, a first trigger sensor is disposed at the position where the aluminum-plastic packaging machine feeds the drugs to the pvc blisters, the first trigger sensor is used for triggering the first detection signal and sending the first detection signal to the control device, and the control device is sent to the first CCD camera and used for the first CCD camera to obtain the first image of the first state after the aluminum-plastic packaging machine feeds the drugs to the pvc blisters.
5. The system for detecting defects of aluminum-plastic blister through image processing as claimed in claim 3, wherein the second CCD camera is disposed after the aluminum-plastic heat-sealing rolling, and a second trigger sensor is disposed at the end of the aluminum-plastic heat-sealing rolling for triggering a second detection signal and sending the second detection signal to the control device, and the control device sends the second detection signal to the second CCD camera for the second CCD camera to obtain a second image of a second state after the aluminum-plastic heat-sealing rolling.
6. The system for detecting defects of aluminum-plastic blisters by using image processing as claimed in claim 3, wherein an evasion mechanism is further arranged in the training model, and the evasion mechanism is used for selectively allocating training resources of the training model to a certain deep neural network branch.
7. The system for detecting defects of aluminum-plastic blisters by using image processing as claimed in claim 3, further comprising a control screen for interactive display, wherein the control screen is used for an interactive interface to display the results.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210077042.XA CN114088730B (en) | 2022-01-24 | 2022-01-24 | Method and system for detecting aluminum-plastic bubble cap defects by using image processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210077042.XA CN114088730B (en) | 2022-01-24 | 2022-01-24 | Method and system for detecting aluminum-plastic bubble cap defects by using image processing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114088730A CN114088730A (en) | 2022-02-25 |
| CN114088730B true CN114088730B (en) | 2022-04-12 |
Family
ID=80309193
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210077042.XA Active CN114088730B (en) | 2022-01-24 | 2022-01-24 | Method and system for detecting aluminum-plastic bubble cap defects by using image processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114088730B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114264668A (en) * | 2022-03-03 | 2022-04-01 | 心鉴智控(深圳)科技有限公司 | Method and system for detecting three-stage flaws of medical packaging box by image processing |
| CN114674831A (en) * | 2022-04-16 | 2022-06-28 | 心鉴智控(深圳)科技有限公司 | Method and system for imaging defects of blister pills |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107316295A (en) * | 2017-07-02 | 2017-11-03 | 苏州大学 | A kind of fabric defects detection method based on deep neural network |
| CN107657603B (en) * | 2017-08-21 | 2020-07-14 | 北京精密机电控制设备研究所 | Industrial appearance detection method based on intelligent vision |
| CN110298839A (en) * | 2019-07-10 | 2019-10-01 | 北京滴普科技有限公司 | A kind of open defect intelligent distinguishing system based on data-driven |
| CN110852395B (en) * | 2019-11-15 | 2023-11-14 | 鞍钢集团矿业有限公司 | Ore granularity detection method and system based on autonomous learning and deep learning |
| CN111242177B (en) * | 2020-01-02 | 2023-07-21 | 天津瑟威兰斯科技有限公司 | Method, system and equipment for detecting medicine package based on convolutional neural network |
| CN111340798A (en) * | 2020-03-16 | 2020-06-26 | 浙江一木智能科技有限公司 | Application of deep learning in product appearance flaw detection |
| CN111862064B (en) * | 2020-07-28 | 2022-08-09 | 桂林电子科技大学 | Silver wire surface flaw identification method based on deep learning |
| CN112784716A (en) * | 2021-01-12 | 2021-05-11 | 北京品恩科技股份有限公司 | Vehicle direction identification method based on deep learning |
| CN113682762A (en) * | 2021-08-27 | 2021-11-23 | 中国矿业大学 | Method and system for belt tear detection based on machine vision and deep learning |
| CN113838034B (en) * | 2021-09-27 | 2023-11-21 | 力度工业智能科技(苏州)有限公司 | Quick detection method for surface defects of candy package based on machine vision |
-
2022
- 2022-01-24 CN CN202210077042.XA patent/CN114088730B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN114088730A (en) | 2022-02-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114088730B (en) | Method and system for detecting aluminum-plastic bubble cap defects by using image processing | |
| US10733723B2 (en) | Methods and system for improved quality inspection | |
| CN108229561B (en) | Particle product defect detection method based on deep learning | |
| CN106548182B (en) | Pavement crack detection method and device based on deep learning and principal cause analysis | |
| Yazdi et al. | Feature extraction algorithm for fill level and cap inspection in bottling machine | |
| CN115156093A (en) | Battery shell defect detection method, system and device | |
| US12350888B2 (en) | AI-powered autonomous 3D printer | |
| CN111062285B (en) | Parking lot intelligent monitoring and recognizing system based on deep learning algorithm | |
| CN114596243A (en) | Defect detection method, apparatus, device, and computer-readable storage medium | |
| CN116245882A (en) | Circuit board electronic element detection method and device and computer equipment | |
| WO2023134286A1 (en) | Online automatic quality testing and classification method for cathode copper | |
| CN111461133A (en) | Express delivery surface single item name identification method, device, equipment and storage medium | |
| CN114331986A (en) | A method of dam crack identification and measurement based on unmanned aerial vehicle vision | |
| CN114723668A (en) | A method and system for quality detection of cathode copper based on area scan camera | |
| CN114324398A (en) | Intelligent visual detection system and detection method for medical and edible packaging label | |
| CN107403444B (en) | Identification system | |
| CN114674831A (en) | Method and system for imaging defects of blister pills | |
| CN114264668A (en) | Method and system for detecting three-stage flaws of medical packaging box by image processing | |
| CN119006406A (en) | Medicine packaging defect detection method and system based on artificial intelligence | |
| Park et al. | Patch-based crack detection in black box road images using deep learning | |
| CN118397248A (en) | Constant volume method and system based on machine vision | |
| KR102723094B1 (en) | Apparatus for inspecting products | |
| CN115793632A (en) | Unmanned vehicle control method and server | |
| CN117339901A (en) | Real-time online detection method, system and device for penicillin bottle label | |
| CN116152203A (en) | Small-batch-scene-oriented few-sample-instance segmented part detection method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |