CN115100095A - PCB detection method based on unsupervised algorithm - Google Patents
PCB detection method based on unsupervised algorithm Download PDFInfo
- Publication number
- CN115100095A CN115100095A CN202111637789.8A CN202111637789A CN115100095A CN 115100095 A CN115100095 A CN 115100095A CN 202111637789 A CN202111637789 A CN 202111637789A CN 115100095 A CN115100095 A CN 115100095A
- Authority
- CN
- China
- Prior art keywords
- pcb
- information
- model
- rechecking
- unsupervised algorithm
- 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.)
- Granted
Links
Images
Classifications
-
- 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
- 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
- G06T7/001—Industrial image inspection using an image reference approach
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30141—Printed circuit board [PCB]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a PCB detection method based on an unsupervised algorithm, which is characterized in that an auxiliary AI detection device based on the unsupervised algorithm is connected in the original rechecked AI to realize the functions of quick learning and quick judgment, and the AI auxiliary detection is carried out on the point positions with high false alarm rate of the existing AOI device, so that the false alarm is reduced, and the first pass rate of the whole PCB is improved.
Description
Technical Field
The invention relates to the field of PCB detection, in particular to a PCB detection method based on an unsupervised algorithm.
Background
After the SMT (surface mount technology) process in the PCB (printed circuit board) manufacturing process, the quality of component mounting must be checked by AOI (automatic optical inspection) equipment. The conventional detection algorithm of the AOI device adopts a simple comparison mode, and manually sets a reference standard for component determination in advance, which is mostly low-dimensional quantization parameters, such as length, width and height of a component, offset from a central position, pixel area and the like. The low-dimensional comparison is not applicable to the judgment of some elements at some point positions on the PCB, for example, the characters on the elements become fuzzy after passing through a high-temperature chip mounting furnace, and the boundaries of 'OK' and 'NG' cannot be defined by using a plurality of parameters.
The conventional AI detection mode adopts a supervised learning algorithm, NG sample collection time is long, and a trained model needs to be obtained for a long time.
The existing mode of using AI to do detection is based on material (component) modeling, and the training process is to position each component on the PCB, then extract the image of the component by matting, resulting in: extra image processing steps are needed for positioning and matting, and the whole algorithm flow is complicated; what extract through the sectional drawing is every material picture, does not contain the whole image on this point position, is the model based on this material training when carrying out the deep learning model training, still will repeat the sectional drawing step during the detection, redetection, because the background board influence, the effect is not good enough.
Disclosure of Invention
The invention mainly aims to provide a PCB detection method based on an unsupervised algorithm, which carries out AI auxiliary detection on the point with high false alarm rate of the existing AOI equipment, reduces false alarms and improves the first pass rate of the whole PCB.
In order to achieve the purpose, the invention adopts the technical scheme that:
a PCB board detection method based on unsupervised algorithm is characterized in that: the method comprises the following steps:
s1, identifying the PCB image through the original AOI equipment, judging whether the PCB is qualified, and classifying the OK and NG detection information of the PCB into a special information base;
s2, rechecking the PCB information of the NG through rechecking AI equipment based on an unsupervised algorithm, uploading NG point location information on the rechecked PCB to an NG information base, putting the NG PCB into a maintenance station, and manually rechecking;
s3, manually rechecking the NG PCB to generate a log file, putting the qualified PCB after manual rechecking into an AI auxiliary detection device based on an unsupervised algorithm to perform repeated learning training, loading the information after learning training into a search library, and feeding back the information to a system for rechecking AOI equipment;
further, the failure information in S1 is a small graph of each tap point of the PCB that AOI outputs NG.
Further, the PCB which is manually determined to be NG in the S3 is subjected to rework processing, the PCB is subjected to detection again through original AOI equipment after rework, the division points which are manually determined to be qualified are placed into a cache folder for a training model of an AI auxiliary detection device, the AI auxiliary detection device is used for carrying out circular training of each point on the PCB, and after the training is finished, the model is placed into a corresponding point model folder to be used as a comparison model.
Further, the unqualified information in S1 is transmitted to an original folder of the retest AI device, the unqualified information is packaged in the memory by the AI auxiliary detection device, and the comparison model information in the point model folder is compared with the unqualified information in the memory by the AI search engine.
Further, the AI retrieval engine does not find a comparison model corresponding to the point location, which indicates that training data is insufficient, and the side surface also indicates that the point location has less false alarm, and the AI retrieval engine does not have strong urgency for re-judging, and further performs NG processing on NG point locations for finding the comparison model, marks the NG point locations, and reconstructs the network folder.
Further, after finding the comparison model of the corresponding point location, the AI retrieval engine loads the model information into the memory of the rechecking AI device, and rapidly judges whether the model information is qualified or not in the memory, reconstructs the qualified pictures into OK network folders, and reconstructs the unqualified pictures into NG network folders.
Further, the pictures of each point location on one board in the network folder are used for reconstructing the PCB structure through AI, the board information with NG point locations is transmitted to a maintenance station for manual re-judgment, the PCB which is manually re-judged to be OK is put into an AI auxiliary detection device for learning and training, and the NG board is reworked and re-detected.
Compared with the prior art, the invention has the following beneficial effects:
an AI auxiliary detection mode is adopted to carry out AI recheck on the product detected by the original AOI, and the algorithm is used for preferentially selecting the point with high AOI omission factor, so that the overall efficiency and the accuracy are highest;
only positive samples (OK pictures) are collected, the model is trained, and the OK pictures can be collected quickly, so that the model can be trained in a short time to start judgment.
The AI model is built based on point locations rather than materials (elements). The advantages are that: the dot bitmap output by the AOI is directly utilized, so that the efficiency is high and the speed is high; the point bitmap comprises the material (element) and images and information around the element on the substrate, fully simulates the scene observed and checked by human eyes, and carries out multi-dimensional judgment, thereby being beneficial to reducing overkill.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The following detailed description of the preferred embodiments of the present invention, taken in conjunction with the accompanying drawings, will make the advantages and features of the present invention more comprehensible to those skilled in the art, and will thus provide a clear and concise definition of the scope of the present invention.
Referring to fig. 1, a PCB detection method based on an unsupervised algorithm includes the steps of:
s1, identifying the PCB image through the original AOI equipment, judging whether the PCB is qualified, and classifying the OK and NG detection information of the PCB into a special information base;
s2, rechecking the NG PCB information through the rechecking AI device based on the unsupervised algorithm, uploading the NG point location information on the rechecked PCB to an NG information base, putting the NG PCB into a maintenance station, and manually rechecking;
s3, manually rechecking the NG PCB to generate a log file, putting the qualified PCB after manual rechecking into an AI auxiliary detection device based on an unsupervised algorithm to perform repeated learning training, loading the information after learning training into a search library, and feeding back the information to a system for rechecking AOI equipment;
and the unqualified information in the S1 is a small graph of each point division point of the PCB output by AOI in NG.
And in the S3, the PCB which is manually re-judged to be NG is subjected to reworking processing, the PCB is subjected to detection again through the original AOI equipment after being reworked, the branch point positions which are manually re-judged to be qualified are placed into a cache folder for a training model of an AI auxiliary detection device, the AI auxiliary detection device is used for carrying out circular training of each point position on the PCB, and after the training is finished, the model is placed into a corresponding point position model folder to be used as a comparison model.
And transmitting the unqualified information in the S1 to an original folder of the retest AI device, packaging the unqualified information in a memory through an AI auxiliary detection device, and comparing the unqualified information with the comparison model information in the point model folder in the memory through an AI retrieval engine.
And if the AI retrieval engine does not find a comparison model of the corresponding point location, the training data is insufficient, the side surface also indicates that the point location has less false alarm, and the tightness is not strong by AI re-judgment, NG processing is carried out again on the NG point location for finding the comparison model, and the NG point location is marked to reconstruct the network folder.
And after finding the comparison model of the corresponding point position, the AI retrieval engine loads the model information into the memory of the rechecking AI equipment, quickly judges whether the model information is qualified or not in the memory, reconstructs the qualified pictures into the OK network folders and reconstructs the unqualified pictures into the NG network folders.
And reconstructing the PCB structure by the pictures of all point positions on one board in the network folder through AI, transmitting the board information with NG point positions to a maintenance station, carrying out manual re-judgment, putting the PCB which is manually re-judged to be OK into an AI auxiliary detection device for learning and training, and reworking and detecting the NG board again.
The above description is only a preferred embodiment of the present invention, and the scope of the present invention is not limited to the above embodiment, but equivalent modifications or changes made by those skilled in the art according to the present disclosure should be included in the scope of the present invention as set forth in the appended claims.
Claims (7)
1. A PCB board detection method based on unsupervised algorithm is characterized in that: the method comprises the following steps:
s1, identifying the PCB image through the original AOI equipment, judging whether the PCB is qualified, and classifying the OK and NG detection information of the PCB into a special information base;
s2, rechecking the PCB information of the NG through rechecking AI equipment based on an unsupervised algorithm, uploading NG point location information on the rechecked PCB to an NG information base, putting the NG PCB into a maintenance station, and manually rechecking;
and S3, manually rechecking the NG PCB to generate a log file, putting the qualified PCB after manual rechecking into an AI auxiliary detection device based on an unsupervised algorithm to perform repeated learning training, loading the information after learning training into a search library, and feeding the information back to a system for rechecking the AOI equipment.
2. The PCB detection method based on the unsupervised algorithm according to claim 1, wherein: the unqualified information in the S1 is a small graph of each branch point of the PCB of the AOI output NG.
3. The PCB detection method based on the unsupervised algorithm according to claim 1, wherein: and in the S3, the PCB which is manually re-judged to be NG is subjected to reworking processing, the PCB is subjected to detection again through the original AOI equipment after being reworked, the branch point positions which are manually re-judged to be qualified are placed into a cache folder for a training model of an AI auxiliary detection device, the AI auxiliary detection device is used for carrying out circular training of each point position on the PCB, and after the training is finished, the model is placed into a corresponding point position model folder to be used as a comparison model.
4. The PCB detection method based on the unsupervised algorithm according to claim 1, wherein: and transmitting the unqualified information in the S1 to an original folder of the retest AI device, packaging the unqualified information in a memory through an AI auxiliary detection device, and comparing the unqualified information with the comparison model information in the point model folder in the memory through an AI retrieval engine.
5. The PCB detection method based on the unsupervised algorithm according to claim 4, wherein: and the AI retrieval engine does not find a comparison model corresponding to the point location, which indicates that training data is insufficient, the side surface also indicates that the point location has less false alarm, and the AI retrieval engine does not have strong urgency of re-judgment, and carries out NG processing again on the NG point location for finding the comparison model, marks the NG point location and reconstructs the network folder.
6. The PCB detection method based on the unsupervised algorithm according to claim 4, wherein: and after finding the comparison model of the corresponding point position, the AI retrieval engine loads the model information into the memory of the rechecking AI equipment, quickly judges whether the model information is qualified or not in the memory, reconstructs the qualified pictures into the OK network folders and reconstructs the unqualified pictures into the NG network folders.
7. The PCB detection method based on the unsupervised algorithm according to the claim 5 or 6, characterized in that: the pictures of each point location on one board in the network folder are used for reconstructing the PCB structure through AI, the board information with NG point locations is transmitted to a maintenance station for manual re-judgment, the PCB which is manually re-judged as OK is put into an AI auxiliary detection device for learning and training, and the NG board is reworked and re-detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111637789.8A CN115100095B (en) | 2021-12-29 | 2021-12-29 | PCB detection method based on non-supervision algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111637789.8A CN115100095B (en) | 2021-12-29 | 2021-12-29 | PCB detection method based on non-supervision algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115100095A true CN115100095A (en) | 2022-09-23 |
| CN115100095B CN115100095B (en) | 2023-08-22 |
Family
ID=83287494
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111637789.8A Active CN115100095B (en) | 2021-12-29 | 2021-12-29 | PCB detection method based on non-supervision algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115100095B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117372368A (en) * | 2023-10-17 | 2024-01-09 | 苏州真目人工智能科技有限公司 | Appearance detection device and method based on cascade closed-loop deep learning algorithm |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109060817A (en) * | 2018-05-24 | 2018-12-21 | 牧德科技股份有限公司 | Artificial intelligence reinspection system and method thereof |
| CN109444172A (en) * | 2018-12-13 | 2019-03-08 | 苏州卓融新能源科技有限公司 | A kind of automatic optical detection device and its method suitable for artificial intelligence detection PCB |
| US20190130555A1 (en) * | 2017-10-27 | 2019-05-02 | Industrial Technology Research Institute | Automated optical inspection (aoi) image classification method, system and computer-readable media |
| CN109919925A (en) * | 2019-03-04 | 2019-06-21 | 联觉(深圳)科技有限公司 | Printed circuit board intelligent detecting method, system, electronic device and storage medium |
| US20190197679A1 (en) * | 2017-12-25 | 2019-06-27 | Utechzone Co., Ltd. | Automated optical inspection method using deep learning and apparatus, computer program for performing the method, computer-readable storage medium storing the computer program,and deep learning system thereof |
| CN110473170A (en) * | 2019-07-10 | 2019-11-19 | 苏州卓融新能源科技有限公司 | A kind of artificial intelligence detection method suitable for the true and false determining defects of pcb board |
| CN110659660A (en) * | 2018-06-29 | 2020-01-07 | 由田新技股份有限公司 | Automatic optical detection classification device using deep learning system and training device thereof |
| CN110705630A (en) * | 2019-09-27 | 2020-01-17 | 聚时科技(上海)有限公司 | Semi-supervised learning type target detection neural network training method, device and application |
| CN110992333A (en) * | 2019-11-29 | 2020-04-10 | 合肥奕斯伟材料技术有限公司 | COF Film defect AITraining database and application |
| CN111080633A (en) * | 2019-12-20 | 2020-04-28 | Oppo（重庆）智能科技有限公司 | Screen defect detection method and device, terminal equipment and storage medium |
| CN111474181A (en) * | 2020-03-31 | 2020-07-31 | 成都数之联科技有限公司 | Panel defect detecting and classifying system in box forming process stage |
| CN113012097A (en) * | 2021-01-19 | 2021-06-22 | 富泰华工业（深圳）有限公司 | Image rechecking method, computer device, and storage medium |
| CN113591965A (en) * | 2021-07-26 | 2021-11-02 | 格力电器(南京)有限公司 | AOI detection image processing method and device, storage medium and computer equipment |
-
2021
- 2021-12-29 CN CN202111637789.8A patent/CN115100095B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190130555A1 (en) * | 2017-10-27 | 2019-05-02 | Industrial Technology Research Institute | Automated optical inspection (aoi) image classification method, system and computer-readable media |
| US20190197679A1 (en) * | 2017-12-25 | 2019-06-27 | Utechzone Co., Ltd. | Automated optical inspection method using deep learning and apparatus, computer program for performing the method, computer-readable storage medium storing the computer program,and deep learning system thereof |
| CN109060817A (en) * | 2018-05-24 | 2018-12-21 | 牧德科技股份有限公司 | Artificial intelligence reinspection system and method thereof |
| CN110659660A (en) * | 2018-06-29 | 2020-01-07 | 由田新技股份有限公司 | Automatic optical detection classification device using deep learning system and training device thereof |
| CN109444172A (en) * | 2018-12-13 | 2019-03-08 | 苏州卓融新能源科技有限公司 | A kind of automatic optical detection device and its method suitable for artificial intelligence detection PCB |
| CN109919925A (en) * | 2019-03-04 | 2019-06-21 | 联觉(深圳)科技有限公司 | Printed circuit board intelligent detecting method, system, electronic device and storage medium |
| CN110473170A (en) * | 2019-07-10 | 2019-11-19 | 苏州卓融新能源科技有限公司 | A kind of artificial intelligence detection method suitable for the true and false determining defects of pcb board |
| CN110705630A (en) * | 2019-09-27 | 2020-01-17 | 聚时科技(上海)有限公司 | Semi-supervised learning type target detection neural network training method, device and application |
| CN110992333A (en) * | 2019-11-29 | 2020-04-10 | 合肥奕斯伟材料技术有限公司 | COF Film defect AITraining database and application |
| CN111080633A (en) * | 2019-12-20 | 2020-04-28 | Oppo（重庆）智能科技有限公司 | Screen defect detection method and device, terminal equipment and storage medium |
| CN111474181A (en) * | 2020-03-31 | 2020-07-31 | 成都数之联科技有限公司 | Panel defect detecting and classifying system in box forming process stage |
| CN113012097A (en) * | 2021-01-19 | 2021-06-22 | 富泰华工业（深圳）有限公司 | Image rechecking method, computer device, and storage medium |
| CN113591965A (en) * | 2021-07-26 | 2021-11-02 | 格力电器(南京)有限公司 | AOI detection image processing method and device, storage medium and computer equipment |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117372368A (en) * | 2023-10-17 | 2024-01-09 | 苏州真目人工智能科技有限公司 | Appearance detection device and method based on cascade closed-loop deep learning algorithm |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115100095B (en) | 2023-08-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11132787B2 (en) | Method for monitoring manufacture of assembly units | |
| CN108686978B (en) | ARM-based fruit category and color sorting method and system | |
| CN110455822A (en) | A kind of detection method of pcb board defect | |
| CN107123111B (en) | Deep residual error network construction method for mobile phone screen defect detection | |
| CN109946319A (en) | Printed circuit board checking device, determine screen process press defect type method and computer can interpretation recording medium | |
| CN107749268A (en) | Screen detection method and equipment | |
| CN113538375A (en) | PCB defect detection method based on YOLOv5 | |
| US11657599B2 (en) | Method for detecting appearance of six sides of chip multi-layer ceramic capacitor based on artificial intelligence | |
| CN112651966A (en) | Printed circuit board micro-defect detection method based on ACYOLOV4_ CSP | |
| CN115015286B (en) | Chip detection method and system based on machine vision | |
| CN116718600B (en) | Detection device for intelligent detection of wire harness product | |
| CN115100095A (en) | PCB detection method based on unsupervised algorithm | |
| CN115775246A (en) | Method for detecting defects of PCB (printed circuit board) components | |
| CN116993725B (en) | Intelligent patch information processing system of flexible circuit board | |
| CN111307817A (en) | Online detection method and system for PCB production process of intelligent production line | |
| CN115941934A (en) | Multi-frame Fusion Video Noise Evaluation Method Based on Deep Convolutional Neural Network | |
| CN112750113B (en) | Glass bottle defect detection method and device based on deep learning and linear detection | |
| CN107123105A (en) | Images match defect inspection method based on FAST algorithms | |
| KR101993705B1 (en) | Apparatus and Method for Automatically Inputting Component Library Code for Automated Optical Inspection Machines | |
| Zhang et al. | A Cascaded Zoom-In Method for Defect Detection of Solder Joints | |
| CN113034432A (en) | Product defect detection method, system, device and storage medium | |
| CN117647531B (en) | Deep learning-based AOI detection method and system | |
| CN117115114B (en) | YOLO-based power electronic device production process target identification and defect detection method | |
| CN111539933B (en) | Direct-insert element detection method and system | |
| CN112085722B (en) | Training sample image acquisition method and device |
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 |