CN112634194A - Self-learning detection method for fabric defects in warp knitting process - Google Patents
Self-learning detection method for fabric defects in warp knitting process Download PDFInfo
- Publication number
- CN112634194A CN112634194A CN202011125372.9A CN202011125372A CN112634194A CN 112634194 A CN112634194 A CN 112634194A CN 202011125372 A CN202011125372 A CN 202011125372A CN 112634194 A CN112634194 A CN 112634194A
- Authority
- CN
- China
- Prior art keywords
- model
- training
- defects
- fabric
- computer
- 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.)
- Pending
Links
- 230000007547 defect Effects 0.000 title claims abstract description 54
- 238000001514 detection method Methods 0.000 title claims abstract description 38
- 239000004744 fabric Substances 0.000 title claims abstract description 35
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000008569 process Effects 0.000 title claims abstract description 25
- 238000009940 knitting Methods 0.000 title claims abstract description 16
- 238000012549 training Methods 0.000 claims abstract description 33
- 239000000203 mixture Substances 0.000 claims abstract description 7
- 230000008676 import Effects 0.000 claims abstract description 4
- 239000004753 textile Substances 0.000 description 6
- 238000009941 weaving Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000011895 specific detection Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 239000002759 woven fabric Substances 0.000 description 1
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
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/12—Edge-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/181—Segmentation; Edge detection involving edge growing; involving edge linking
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Treatment Of Fiber Materials (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a self-learning detection method for fabric defects in a warp knitting process, which is divided into two parts of model training and fabric defect detection, and meanwhile, a training model is repeatedly trained and updated in the defect detection process, and the method comprises the following steps: the camera starts to acquire and obtain image data, and at the moment, in a system learning stage, the system continuously acquires a series of pictures and imports a GMM (Gaussian mixture model) to perform iterative solution of model parameters; after the model is manufactured, a detection stage is started, at the moment, after a camera collects data, a computer firstly detects whether defects exist in a current picture or not based on a previous training model, if the defects exist, the computer alarms or stops the computer, and if the defects do not exist, the current picture is added into a training network, and the training model is updated.
Description
Technical Field
The invention relates to a self-learning detection algorithm for fabric defects in a warp knitting process.
Background
Warp knitting machines are a type of common textile equipment, and are widely applied to the textile industry due to the advantages of strong applicability, high yield and the like. However, during the weaving process of the warp knitting machine, yarn breakage is inevitably generated to form defects on the surface of the fabric, and finally the quality of the finished fabric is seriously influenced. The method has the advantages that the online detection of the fabric defects in the weaving process is an effective means for improving the yield of the textile products, on one hand, the defects can be immediately repaired when generated, meanwhile, the fabric defect positions can be pre-marked, the subsequent defect treatment is facilitated, and along with the development of the textile industry towards the direction of automation and intellectualization, higher requirements are put forward on the online detection of the automatic fabric defects.
The traditional online detection algorithm for the fabric defects is based on the shooting data of the woven fabric surface, and the detection is realized by analyzing the fabric texture through the computer vision technology. However, this method of fabric defect detection has the following problems:
(1) in terms of algorithm applicability, the method for realizing fabric defect detection based on fabric texture analysis needs to process fabric image data, so that a part of exposed areas of the fabric are required to be used for image acquisition after the fabric is woven and formed, and the existing warp knitting machines are shielded by mechanical structures after the fabric is woven and formed, so that the method is not strong in applicability.
(2) In terms of algorithm portability, as warp knitting machine textile technologies are various and fabric texture structures are different, the detection algorithm applicable to the same texture structure fabric is not necessarily applicable to other fabrics, namely, the portability of the method for realizing fabric defect detection based on fabric texture analysis is not strong.
(3) From the stability of the algorithm, in the weaving process of the warp knitting machine, the fabric with the same texture structure may have differences due to the influence of external illumination change, cloth tension and the like, so that the algorithm may fail or generate a large amount of false detections along with the lapse of time, and the stability is not strong.
Disclosure of Invention
Aiming at the problems, the invention aims to provide a self-learning detection algorithm for fabric defects in a warp knitting process, which realizes fabric defect detection in the weaving process, thereby improving the detection efficiency and reducing the defective rate and the production cost.
In order to achieve the purpose, the invention adopts the following technical scheme:
a self-learning detection method for fabric defects in a warp knitting process is divided into two parts of model training and fabric defect detection, and meanwhile, in the defect detection process, a training model is repeatedly trained and updated, and the method comprises the following steps: the camera starts to acquire and obtain image data, and at the moment, in a system learning stage, the system continuously acquires a series of pictures and imports a GMM (Gaussian mixture model) to perform iterative solution of model parameters; after the model is manufactured, a detection stage is started, at the moment, after a camera collects data, a computer firstly detects whether defects exist in a current picture or not based on a previous training model, if the defects exist, the computer alarms or stops the computer, and if the defects do not exist, the current picture is added into a training network, and the training model is updated.
Due to the adoption of the technical scheme, the invention has the following advantages:
(1) the invention carries out image processing based on the data of the knitting area of the warp knitting machine, reduces the requirements on the mechanical structure of the warp knitting machine and the difficulty of data acquisition, and is suitable for all warp knitting machines.
(2) The method is based on a self-learning mode, carries out defect detection based on a defect-free data training result, is suitable for all equipment of textile technology, and has strong transportability;
(3) the invention updates the training model in real time, avoids the influence of external environment transformation on the detection result, and has higher algorithm stability.
Drawings
FIG. 1 is a detailed image data acquired by the present invention.
FIG. 2 is a flow chart of the fabric defect on-line detection algorithm of the present invention.
FIG. 3 is a training model used in the present invention.
The reference numbers in the figures illustrate: 1, data of a defect-free template; 2, data to be detected with defects;
Detailed Description
The invention is described below with reference to the figures and examples.
Fig. 1 is image data specifically targeted by the present invention, where 1 is a defect-free template picture, 2 is a picture to be detected with defects, and the position of a rectangular area in the picture is the position of a defect. In a specific image processing process, defect detection is realized by comparing data in a rectangular area.
The specific detection steps of the algorithm are divided into two parts, namely model training and fabric defect detection, as shown in FIG. 2, and meanwhile, the training model is updated in real time in the defect detection process. The specific flow is that after the system is started, the camera starts to acquire image data, and at the moment, the system is in a system learning stage, and the system analyzes and processes the image data based on the convolutional neural network to obtain a training model. After the model is manufactured, a detection stage is started, at the moment, after a camera collects data, a computer firstly detects whether defects exist in a current picture or not based on a previous training model, if the defects exist, the computer alarms or stops, and if the defects do not exist, the current picture is added into a training network to realize the updating of the training model.
The training model used by the algorithm is shown in fig. 3. Based on the non-broken yarn picture sequence, the method firstly carries out convolution operation on the picture sequence so as to carry out local perception on the picture. Next, data compression is performed through a pooling operation to reduce over-fitting and improve picture fault tolerance. After convolution and pooling for many times, the picture characteristics are highlighted, and finally whether the picture has defects or not is judged through the full connection layer.
The invention is described below with reference to the figures and examples.
In fig. 1, 1 is a defect-free template picture, 2 is a picture to be detected with defects, a rectangular region in the picture is the position of the defect, in the learning process, a gaussian mixture model is specifically used for training, and the gaussian mixture model is specifically expressed as:
in the formula, x represents image data, p (x) represents the probability that the current image data is consistent with the defect-free data, K represents the number of Gaussian models, and pikThe weight of the kth Gaussian model is defined, p (x | k) is a Gaussian probability density value, and the specific calculation formula is as follows:
the result of the Gaussian probability density calculation depends on the input parameter x and the model internal parameter ukSum Σk. According to the Gaussian mixture model formula, the probability p (x) that the image data x is defect-free data can be output as long as the image data x is input, so that whether defects exist or not is judged. The system learning process is an optimization process of the Gaussian mixture model, namely pi is searched by iterationk、ukSum ΣkThe optimal solution of the parameters is solved by using an EM algorithm:
E step：
wi(k) is a sample xiProbability generated by the kth gaussian model.
M step：
The solving process firstly needs to set pik、ukSum ΣkThen repeatedly executing the E process and the M process until convergence, thereby obtaining pik、ukSum ΣkAnd (4) finishing training as a final result. At this time, when a test picture is input, the system can judge the test picture to obtain output P (x), so that whether defects exist in the current picture is judged, and detection is realized. When data needs to be expanded on the basis of original training data, the difference between the new expanded data and the original data is not large, so that the original pi is usedk、ukSum ΣkIteration is performed as an initial value, and the model can achieve rapid convergence.
The overall structural framework of the algorithm is divided into two parts, namely model training and fabric defect detection, as shown in FIG. 2, and meanwhile, in the defect detection process, the training model is repeatedly trained and updated. The specific flow is that after the system is started, the camera starts to acquire and obtain image data, and at the moment, in a system learning stage, the system continuously acquires a series of pictures and imports a GMM model to perform iterative solution of model parameters. After the model is manufactured, a detection stage is started, at the moment, after a camera collects data, a computer firstly detects whether defects exist in a current picture or not based on a previous training model, if the defects exist, the computer alarms or stops the computer, and if the defects do not exist, the current picture is added into a training network, and the training model is updated.
Claims (1)
1. A self-learning detection method for fabric defects in a warp knitting process is divided into two parts of model training and fabric defect detection, and meanwhile, in the defect detection process, a training model is repeatedly trained and updated, and the method comprises the following steps: the camera starts to acquire and obtain image data, and at the moment, in a system learning stage, the system continuously acquires a series of pictures and imports a GMM (Gaussian mixture model) to perform iterative solution of model parameters; after the model is manufactured, a detection stage is started, at the moment, after a camera collects data, a computer firstly detects whether defects exist in a current picture or not based on a previous training model, if the defects exist, the computer alarms or stops the computer, and if the defects do not exist, the current picture is added into a training network, and the training model is updated.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011125372.9A CN112634194A (en) | 2020-10-20 | 2020-10-20 | Self-learning detection method for fabric defects in warp knitting process |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011125372.9A CN112634194A (en) | 2020-10-20 | 2020-10-20 | Self-learning detection method for fabric defects in warp knitting process |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112634194A true CN112634194A (en) | 2021-04-09 |
Family
ID=75303101
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011125372.9A Pending CN112634194A (en) | 2020-10-20 | 2020-10-20 | Self-learning detection method for fabric defects in warp knitting process |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112634194A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113567443A (en) * | 2021-06-22 | 2021-10-29 | 浙江大豪科技有限公司 | Control method and device of fabric flaw detection system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107169956A (en) * | 2017-04-28 | 2017-09-15 | 西安工程大学 | Yarn dyed fabric defect detection method based on convolutional neural networks |
| CN108288263A (en) * | 2017-12-21 | 2018-07-17 | 江南大学 | A kind of knitted fabric fault online test method based on Adaptive Neuro-fuzzy Inference |
| CN109509171A (en) * | 2018-09-20 | 2019-03-22 | 江苏理工学院 | A kind of Fabric Defects Inspection detection method based on GMM and image pyramid |
| CN109598712A (en) * | 2018-11-30 | 2019-04-09 | 北京百度网讯科技有限公司 | Quality determining method, device, server and the storage medium of plastic foam cutlery box |
| CN111062925A (en) * | 2019-12-18 | 2020-04-24 | 华南理工大学 | Intelligent cloth defect identification method based on deep learning |
-
2020
- 2020-10-20 CN CN202011125372.9A patent/CN112634194A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107169956A (en) * | 2017-04-28 | 2017-09-15 | 西安工程大学 | Yarn dyed fabric defect detection method based on convolutional neural networks |
| CN108288263A (en) * | 2017-12-21 | 2018-07-17 | 江南大学 | A kind of knitted fabric fault online test method based on Adaptive Neuro-fuzzy Inference |
| CN109509171A (en) * | 2018-09-20 | 2019-03-22 | 江苏理工学院 | A kind of Fabric Defects Inspection detection method based on GMM and image pyramid |
| CN109598712A (en) * | 2018-11-30 | 2019-04-09 | 北京百度网讯科技有限公司 | Quality determining method, device, server and the storage medium of plastic foam cutlery box |
| CN111062925A (en) * | 2019-12-18 | 2020-04-24 | 华南理工大学 | Intelligent cloth defect identification method based on deep learning |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113567443A (en) * | 2021-06-22 | 2021-10-29 | 浙江大豪科技有限公司 | Control method and device of fabric flaw detection system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Huang et al. | Fabric defect segmentation method based on deep learning | |
| CN111402197B (en) | Detection method for colored fabric cut-parts defect area | |
| CN103529051B (en) | A kind of Woven textiles flaw automatic on-line detection method | |
| CN109272500B (en) | Fabric classification method based on adaptive convolutional neural network | |
| CN113030108A (en) | Coating defect detection system and method based on machine vision | |
| CN106875373A (en) | Mobile phone screen MURA defect inspection methods based on convolutional neural networks pruning algorithms | |
| CN111260621B (en) | Positioning and identifying method for surface defects of printed circuit board | |
| CN111223093A (en) | AOI defect detection method | |
| CN109472769A (en) | A kind of bad image defect detection method and system | |
| CN109840900A (en) | A kind of line detection system for failure and detection method applied to intelligence manufacture workshop | |
| CN111861996A (en) | Printed fabric defect detection method | |
| CN112634194A (en) | Self-learning detection method for fabric defects in warp knitting process | |
| CN108802041A (en) | A kind of method that the small sample set of screen detection is quickly remodeled | |
| Zuo et al. | Application of YOLO object detection network in weld surface defect detection | |
| CN112907562A (en) | MobileNet-based SMT defect classification algorithm | |
| CN114972216A (en) | Construction method and application of texture surface defect detection model | |
| CN113112482A (en) | PCB defect detection method based on attention mechanism network | |
| CN111784691A (en) | Textile flaw detection method | |
| CN109871906B (en) | Cop appearance defect classification method based on deep convolutional neural network | |
| He et al. | Research on Fabric defect detection based on deep fusion DenseNet-SSD network | |
| Harreni et al. | Novel Multipath Convolutional Neural Network Based Fabric Defect Detection System | |
| CN108333183B (en) | Yarn-dyed shirt cut piece defect detection method based on DCGAN and DCNN | |
| Chen et al. | Adaptive gabor filtering for fabric defect inspection | |
| Zhang et al. | An improved DCGAN for fabric defect detection | |
| Xie et al. | Defect detection of printed circuit board based on small target recognition network |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210409 |