CN111681231B - Industrial welding image defect detection method based on target detection - Google Patents
Industrial welding image defect detection method based on target detection Download PDFInfo
- Publication number
- CN111681231B CN111681231B CN202010525671.5A CN202010525671A CN111681231B CN 111681231 B CN111681231 B CN 111681231B CN 202010525671 A CN202010525671 A CN 202010525671A CN 111681231 B CN111681231 B CN 111681231B
- Authority
- CN
- China
- Prior art keywords
- welding
- defect
- industrial
- image
- welding 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
- 238000003466 welding Methods 0.000 title claims abstract description 282
- 230000007547 defect Effects 0.000 title claims abstract description 236
- 238000001514 detection method Methods 0.000 title claims abstract description 105
- 238000000034 method Methods 0.000 claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 34
- 238000013135 deep learning Methods 0.000 claims abstract description 21
- 239000013598 vector Substances 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 9
- 238000011478 gradient descent method Methods 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 abstract description 7
- 238000005516 engineering process Methods 0.000 description 7
- 238000007689 inspection Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000013507 mapping Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 239000002893 slag Substances 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000011179 visual inspection Methods 0.000 description 2
- 206010000372 Accident at work Diseases 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000002537 cosmetic Substances 0.000 description 1
- 238000005336 cracking Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000035515 penetration Effects 0.000 description 1
- 238000012800 visualization Methods 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- 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
- G06T2207/30152—Solder
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for detecting industrial welding image defects based on target detection, which comprises the steps of obtaining an industrial welding image; generating a training sample set according to welding defects in the industrial welding image; establishing a welding defect detection model; deep learning training is carried out on the welding defect detection model according to the training sample set; and carrying out welding defect detection on the industrial welding image by using the welding defect detection model after deep learning training. The method for detecting the industrial welding image defects based on the target detection aims to solve the problems of missed detection and error detection and complex process of manual welding defect detection work.
Description
Technical Field
The invention belongs to the technical field of industrial visual inspection and deep learning, and particularly relates to a method for detecting industrial welding image defects based on target detection.
Background
The welding defects refer to defects formed in the welding process of the welding joint part, and mainly comprise welding cracks, incomplete penetration, slag inclusion, air holes, appearance defects of welding seams and the like. These defects reduce the weld area, reduce the bearing capacity, produce stress concentrations, cause cracks; the fatigue strength is reduced, and the weldment is easy to crack to cause brittle failure. The welding production is influenced by various factors such as material, process, structure and the like, and the welding defect is easily generated due to improper operation in the welding operation. Weld defects are a major cause of boiler, pressure vessel failure and accidents. Therefore, in the industrial welding quality inspection, the welding defects of the welding points between the welding pieces are accurately detected, and the method is very important for avoiding serious industrial accidents. Weld cosmetic defects (surface defects) are defects that can be found from the surface of a workpiece without the aid of an instrument. Common welding appearance defects include undercuts, flashes, depressions, welding deformation and the like. And the condition of the welding point is easy to make mistakes by naked eyes in manual welding quality detection, so that the missed detection and the false detection of the welding defect are caused, and the subsequent engineering quality is influenced.
At present, no method for effectively and automatically detecting welding appearance defects of an industrial welding image exists, and only the problems of cloth flaw detection and other similar industrial detection can be used as comparison. A common method in the traditional industrial detection problem is to respectively model each category by using prior knowledge through a Gaussian mixture model and then detect by using the parameter distribution of the model. The model well describes texture characteristics of different classes, and has a good classification effect in a single scene. In practical application, industrial welding images are easily affected by changes of illumination, environment and the like, so that the method is large in calculation amount and poor in robustness. Deep semantic features of the image can be effectively extracted by training a target detection model through a deep learning technology, the position of the target is accurately positioned, and the target category is identified.
In view of this, those skilled in the art need to provide a method for detecting defects of an industrial welding image based on target detection to solve the problems of missing detection, error detection and tedious process of manual welding defect detection.
Disclosure of Invention
Technical problem to be solved
The invention aims to solve the technical problems of missed detection and error detection and complex process of manual welding defect detection work.
(II) technical scheme
The invention provides a method for detecting industrial welding image defects based on target detection, which is characterized by comprising the following steps of:
s1, acquiring an industrial welding image;
s2, generating a training sample set according to welding defects in the industrial welding image;
s3, establishing a welding defect detection model;
s4, performing deep learning training on the welding defect detection model according to the training sample set;
and S5, carrying out welding defect detection on the industrial welding image by using the welding defect detection model after deep learning training.
Optionally, in step S1, the acquiring an industrial welding image specifically includes:
and shooting the industrial welding image through shooting equipment and storing the industrial welding image.
Optionally, step S2 specifically includes the following steps:
calibrating a welding defect area frame in the industrial welding image according to a welding appearance defect example;
judging and marking the type of the welding defect according to the appearance characteristic of the defect so as to obtain the training sample set containing the welding defect area frame and corresponding to the type of the welding defect.
Optionally, the calibrating the welding defect area frame in the industrial welding image with reference to the welding appearance defect example specifically includes:
the welding defect area frames are specific area ranges of the welding defects in the industrial welding image, and each area frame represents a welding defect area represented by t = [ x, y, w, h ];
wherein x is the horizontal coordinate of the upper left corner of the welding defect area frame, y is the vertical coordinate of the upper left corner of the welding defect area frame, w is the width of the welding defect area frame, and h is the height of the welding defect area frame.
Optionally, the determining and marking the type of the welding defect according to the defect appearance characteristics specifically includes:
judging the type of the welding defect according to the characteristic of the welding defect in each calibrated welding defect area frame, marking the type of the welding defect, and enabling the welding defect area frame contained in the training sample set to correspond to the type of the welding defect, wherein the type of the welding defect is represented as p.
Optionally, the establishing of the welding defect detection model specifically includes establishing a welding defect area frame presumption module, and a welding defect classification and regression module.
Optionally, the welding defect region frame inference module includes an image feature extractor and a region frame inference network;
inputting the industrial welding image to the image feature extractor, and extracting semantic features of the industrial welding image;
and determining a candidate region of the welding defect according to the semantic features of the industrial welding image, and generating a candidate region frame.
Optionally, the welding defect classification and regression module includes a welding defect type classifier and a welding defect region regression;
inputting the semantic features of the industrial welding image and the candidate region frame to the welding defect classification and regression module;
intercepting a corresponding region of interest from the semantic features of the industrial welding image according to the candidate region frame;
using a variable pooling layer to down-sample the region of interest into feature vectors, and using the welding defect type classifier to classify the feature vectors to obtain a welding defect type classification result, which is expressed as p * ；
Inputting the characteristic vector to the welding defect region regressor to obtain a regression result of a welding defect region frame, wherein the regression result is represented as t * ＝[x * ,y * ,w * ,h * ]；
Wherein x is * To the left of the prediction region boxUpper corner abscissa, y * Is the vertical coordinate of the upper left corner of the prediction region box, w * Is the width of the prediction region box, h * Is the height of the prediction region box.
Optionally, in step S4, the deep learning training method is a random gradient descent method, and the target parameters of the welding defect detection model are learned through a minimization loss function.
Optionally, the loss function is:
wherein p is i For the noted category of weld defects,
class of weld defect predicted for weld defect detection model, t i For marked weld defect area frames>
Welding defect zone box predicted for welding defect detection model, N cls Number of classes of welding defects, N reg And lambda is a balance factor and takes the value of 10.
(III) advantageous effects
The technical scheme of the invention has the following advantages:
the invention provides a method for detecting industrial welding image defects based on target detection, which comprises the steps of obtaining an industrial welding image; generating a training sample set according to welding defects in the industrial welding image; establishing a welding defect detection model; deep learning training is carried out on the welding defect detection model according to the training sample set; and carrying out welding defect detection on the industrial welding image by using the welding defect detection model after deep learning training. The detection method adopts a deep learning technology to realize automatic detection of the industrial welding image defects, solves the problems of missing detection and error detection and complex process of manual welding defect detection work, improves the working efficiency and reduces the industrial safety accident rate.
Drawings
FIG. 1 is a schematic flowchart of a method for detecting defects in an industrial welding image based on object detection according to an embodiment of the present invention;
FIG. 2 is a schematic diagram of a common welding defect in another industrial welding image defect detection method based on object detection according to an embodiment of the present invention;
FIG. 3 is a schematic view of a welding defect detection model in another industrial welding image defect detection method based on object detection according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a welding defect area frame estimation module in another industrial welding image defect detection method based on target detection according to an embodiment of the present invention;
FIG. 5 is a schematic diagram of a welding defect classification and regression module in another industrial welding image defect detection method based on object detection according to an embodiment of the present invention;
fig. 6 is a schematic flowchart of another method for detecting defects in an industrial welding image based on object detection according to an embodiment of the present invention.
In the figure: 2A, undercut; 2B, welding beading; 2C, air holes; 2D, slag inclusion; 2E, cracking; 100. an industrial welding image; 200. detecting a welding defect result; 300. an image feature extractor; 400. semantic features of the industrial welding image; 500. a zone box guess network; 600. a candidate region box; 700. mapping the region; 800. a region of interest; 900. a feature vector; 101. classifying the types of the welding defects; 102. regression results of welding defect areas; 103. detecting a visual result of the welding defect; F. inputting; G. outputting; H. visualization; I. mapping; J. intercepting; K. and (4) downsampling.
Detailed Description
The following detailed description of embodiments of the present invention is provided in connection with the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
With the rapid development of computer vision and deep learning technologies, the application of computer vision-based automatic detection technology in industrial production is becoming more and more extensive, and face recognition based on target detection is a typical application. However, the quality inspection of the industrial welding industry still mainly adopts manual visual inspection, and has the problems of high cost, low detection efficiency and the like. The automatic realization of welding defect detection and quality inspection by using computer vision detection, image processing and image recognition technology has become one of the problems to be solved urgently. The target detection is to find out all interested objects in the image, comprises two subtasks of object positioning and object classification, and determines the category and the position of the object at the same time. The target detection is a popular direction of computer vision and digital image processing, is widely applied to various fields of robot navigation, intelligent video monitoring, industrial detection, aerospace and the like, reduces the consumption of human capital through the computer vision, and has important practical significance.
As shown in fig. 1, according to an embodiment of the present invention, there is provided a method for detecting defects in an industrial welding image based on target detection, including the following steps:
s1, acquiring an industrial welding image;
s2, generating a training sample set according to welding defects in the industrial welding image;
s3, establishing a welding defect detection model;
s4, deep learning training is carried out on the welding defect detection model according to the training sample set;
and S5, detecting the welding defects of the industrial welding image by using the welding defect detection model after deep learning training.
In the embodiment, the detection method adopts the deep learning technology to realize the automatic detection of the industrial welding image defects, solves the problems of missing detection and error detection and complicated process of the manual welding defect detection work, improves the working efficiency and reduces the industrial safety accident rate.
In some optional embodiments, in step S1, the acquiring of the industrial welding image specifically includes:
and shooting an industrial welding image through shooting equipment and storing the industrial welding image.
In some optional embodiments, step S2 specifically includes the following steps:
calibrating a welding defect area frame in the industrial welding image according to the welding appearance defect example;
and judging and marking the welding defect type according to the defect appearance characteristics so as to obtain a training sample set containing a welding defect area frame and corresponding to the welding defect type.
Optionally, distinguishing and marking are performed according to texture features of common welding appearance defects, specifically, as shown in fig. 2, five common welding appearance defects are addressed in the embodiment of the present invention: and marking the undercut 2A, the welding beading 2B, the recess 2C, the air hole 2D and the crack 2E.
In some optional embodiments, the welding defect area frame in the industrial welding image is calibrated with reference to the welding appearance defect example, specifically:
the welding defect area frames are specific area ranges of the welding defects in the industrial welding image, each area frame represents a welding defect area, and the area is represented by t = [ x, y, w, h ];
wherein x is the horizontal coordinate of the upper left corner of the welding defect area frame, y is the vertical coordinate of the upper left corner of the welding defect area frame, w is the width of the welding defect area frame, and h is the height of the welding defect area frame.
In some optional embodiments, the type of the welding defect is judged and marked according to the appearance characteristics of the defect, specifically:
and judging the type of the welding defect according to the characteristics of the welding defect in each calibrated welding defect area frame, marking the type of the welding defect, wherein the type can be expressed as p, and the welding defect area frame contained in the training sample set corresponds to the type of the welding defect.
In some optional embodiments, as shown in fig. 3, the creating of the welding defect detection model specifically includes creating a welding defect region box inference module, a welding defect classification and regression module.
In some alternative embodiments, as shown in fig. 4, the welding defect region box guessing module includes an image feature extractor and a region box guessing network;
inputting an industrial welding image 100 to an image feature extractor 300, and extracting semantic features 400 of the industrial welding image;
and determining a candidate region of the welding defect according to the semantic features 400 of the industrial welding image, and generating a candidate region frame.
Specifically, the industrial welding image is input to the welding defect region frame estimation module, which is an image feature extractor to extract semantic features of the industrial welding image, and in this embodiment, resNet (Residual Neural Network) is used as the image feature extractor 300; next, the area frame estimation Network 500 estimates a candidate area having a welding defect in the image based on the semantic features 400 of the industrial welding image, and generates a candidate area frame 600, and in this embodiment, an RPN (Region pro-active Network) Network is used as the area frame estimation Network 500.
The RPN infers a candidate region with possible welding defects in the image through an anchor point mechanism, and particularly predicts a group of anchor points for each point on the semantic characteristics of the industrial welding image, wherein each group of anchor points consists of a series of candidate regions with fixed sizes and proportions.
Accurately, the RPN network outputs the probability that the candidate region belongs to the background while inferring the candidate region range of each anchor point, and if the probability that the candidate region belongs to the background is too high, the anchor point is filtered.
In some alternative embodiments, as shown in fig. 5, the weld defect classification and regression module includes a weld defect type classifier and a weld defect region regressor;
inputting the semantic features 400 and the candidate region frames 600 of the industrial welding image to a welding defect classification and regression module;
intercepting a corresponding region of interest 800 in the semantic features 400 of the industrial welding image according to the candidate region box 600;
the region of interest is down-sampled into feature vectors by using a variable pooling layer, and the feature vectors are classified by using a welding defect type classifier to obtain a welding defect type classification result expressed as p * ；
Inputting the characteristic vector to a welding defect region regressor to obtain a regression result of a welding defect region frame, wherein the regression result is represented as t * ＝[x * ,y * ,w * ,h * ]；
Wherein x is * To predict the upper left abscissa, y, of the region box * To predict the vertical coordinate, w, of the upper left corner of the region box * To predict the width of the region box, h * The height of the region box is predicted.
Specifically, image features and candidate region frames are input into a welding defect classification and regression module, corresponding regions of interest 800 are intercepted from the image features according to the candidate region frames 600, the regions of interest are down-sampled into feature vectors 900 by using a variable pooling layer, and the feature vectors 900 are classified by using a welding defect type classifier to obtain a welding defect type classification result 101;
finally, the feature vector 900 is input into a welding defect region regression device to obtain a welding defect region frame regression result 102.
In some optional embodiments, in step S4, the deep learning training method is a stochastic gradient descent method, and the target parameters of the welding defect detection model are learned by a minimization loss function.
In some alternative embodiments, the loss function is:
wherein p is i For the noted category of weld defects,
class of weld defect predicted for weld defect detection model, t i For the marked welding-defective area frame->
Welding defect zone box predicted for welding defect detection model, N cls Number of classes of welding defects, N reg And lambda is a balance factor and takes the value of 10.
As shown in fig. 6, another method for detecting defects of an industrial welding image based on target detection is provided according to an embodiment of the present invention, which includes the following steps:
s201, shooting an industrial welding image through shooting equipment and storing the image;
s202, calibrating a welding defect area frame in the industrial welding image according to a welding appearance defect example;
s203, judging and marking the welding defect type according to the defect appearance characteristics to obtain a training sample set containing a welding defect area frame and corresponding to the welding defect type;
s204, establishing a welding defect detection model;
s205, learning a target parameter of a welding defect detection model by a minimized loss function, wherein the deep learning training method is a random gradient descent method;
and S206, detecting the welding defects of the industrial welding image by using the welding defect detection model after deep learning training.
Compared with other detection methods, the detection method of the industrial welding image defect based on the target detection provided by the embodiment of the invention has the advantages and effects that:
(1) The method comprises the steps of training a target detection model through deep learning, and automatically detecting welding defects in an industrial welding image;
(2) The welding defect detection model can effectively extract the shape, color and outline characteristics of the industrial welding image, and has higher robustness and characteristic description capability compared with the traditional method;
(3) Welding flaws of any scale can be detected, and an area of interest with an indefinite size is converted into a feature vector with a fixed scale, so that the method has good area positioning and type identification performance and high stability;
(4) The method realizes accurate detection of common welding defects based on a target detection technology, provides a welding defect detection flow with better expansibility, and is also suitable for industrial visual detection problems such as part flaw detection, product quality inspection and the like.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, several modifications and variations can be made without departing from the technical principle of the present invention, and these modifications and variations should also be regarded as the protection scope of the present invention.
Claims (5)
1. The method for detecting the defects of the industrial welding image based on the target detection is characterized by comprising the following steps of:
s1, acquiring an industrial welding image;
s2, generating a training sample set according to welding defects in the industrial welding image;
s3, establishing a welding defect detection model;
s4, performing deep learning training on the welding defect detection model according to the training sample set;
s5, carrying out welding defect detection on the industrial welding image by using a welding defect detection model after deep learning training;
the method specifically comprises the steps of establishing a welding defect area frame presumption module and a welding defect classification and regression module;
the welding defect region frame presumption module comprises an image feature extractor and a region frame presumption network;
inputting the industrial welding image to the image feature extractor, and extracting semantic features of the industrial welding image;
determining a candidate region of the welding defect according to the semantic features of the industrial welding image, and generating a candidate region frame;
the welding defect classification and regression module comprises a welding defect type classifier and a welding defect region regressor;
inputting the semantic features of the industrial welding image and the candidate region frame to the welding defect classification and regression module;
intercepting a corresponding region of interest from the semantic features of the industrial welding image according to the candidate region frame;
using a variable pooling layer to down-sample the region of interest into feature vectors, and using the welding defect type classifier to classify the feature vectors to obtain a welding defect type classification result, which is expressed as p * ；
Inputting the feature vector to the welding defect region regressor to obtain a regression result of a welding defect region frame, wherein the regression result is represented as t * ＝[x * ,y * ,w * ,h * ]；
Wherein x is * To predict the upper left abscissa, y, of the region box * Is the vertical coordinate, w, of the upper left corner of the prediction region box * For the width of the prediction region box, h * Is the height of the prediction region box;
in the step S4, the deep learning training method is a random gradient descent method, and target parameters of the welding defect detection model are learned through a minimized loss function;
the loss function is:
wherein p is i For the noted category of weld defects,
class of weld defect predicted for weld defect detection model, t i For marked weld defect area frames>
Welding defect zone box predicted for welding defect detection model, N cls Number of classes of welding defects, N reg The number of frames in the welding defect area is shown, and lambda is a balance factor and is 10; l is cls Classifying the loss function for the weld defect class, L reg For the weld defect region location regression loss function, { x, y, w, h } is the upper left-hand coordinate of the weld defect region location and the region width and height.
2. The method for detecting defects in an industrial welding image based on target detection as claimed in claim 1, wherein in step S1, the acquiring of the industrial welding image specifically comprises:
and shooting the industrial welding image through shooting equipment and storing the industrial welding image.
3. The method for detecting the defects of the industrial welding image based on the target detection as claimed in claim 1, wherein the step S2 specifically comprises the following steps:
calibrating a welding defect area frame in the industrial welding image according to a welding appearance defect example;
judging and marking the welding defect type according to the defect appearance characteristics so as to obtain the training sample set containing the welding defect area frame and corresponding to the welding defect type.
4. The method for detecting defects in industrial welding images based on target detection as claimed in claim 3, wherein said calibrating the welding defect area frame in the industrial welding image with reference to the welding appearance defect example specifically comprises:
the welding defect area frames are specific area ranges of the welding defects in the industrial welding image, and each area frame represents a welding defect area represented by t = [ x, y, w, h ];
wherein x is the horizontal coordinate of the upper left corner of the welding defect area frame, y is the vertical coordinate of the upper left corner of the welding defect area frame, w is the width of the welding defect area frame, and h is the height of the welding defect area frame.
5. The method for detecting the defects of the industrial welding image based on the target detection as claimed in claim 3, wherein the judging and marking the types of the welding defects according to the appearance characteristics of the defects are specifically as follows:
judging the type of the welding defect according to the characteristic of the welding defect in each calibrated welding defect area frame, marking the type of the welding defect, and enabling the welding defect area frame contained in the training sample set to correspond to the type of the welding defect, wherein the type of the welding defect is represented as p.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010525671.5A CN111681231B (en) | 2020-06-10 | 2020-06-10 | Industrial welding image defect detection method based on target detection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010525671.5A CN111681231B (en) | 2020-06-10 | 2020-06-10 | Industrial welding image defect detection method based on target detection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111681231A CN111681231A (en) | 2020-09-18 |
| CN111681231B true CN111681231B (en) | 2023-04-18 |
Family
ID=72454415
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010525671.5A Active CN111681231B (en) | 2020-06-10 | 2020-06-10 | Industrial welding image defect detection method based on target detection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111681231B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI806004B (en) * | 2021-02-09 | 2023-06-21 | 麗臺科技股份有限公司 | AI process management system and method for automated visual inspection |
| CN112700446A (en) * | 2021-03-23 | 2021-04-23 | 常州微亿智造科技有限公司 | Algorithm model training method and device for industrial quality inspection |
| CN115564766B (en) * | 2022-11-09 | 2023-06-13 | 浙江振兴阿祥集团有限公司 | Preparation method and system of water turbine volute seat ring |
| CN118537340B (en) * | 2024-07-26 | 2024-10-25 | 江苏航运职业技术学院 | Cloud computing-based weld defect intelligent detection method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109285139A (en) * | 2018-07-23 | 2019-01-29 | 同济大学 | A kind of x-ray imaging weld inspection method based on deep learning |
| CN109472769A (en) * | 2018-09-26 | 2019-03-15 | 成都数之联科技有限公司 | A kind of bad image defect detection method and system |
| CN109859177A (en) * | 2019-01-17 | 2019-06-07 | 航天新长征大道科技有限公司 | Industrial x-ray image assessment method and device based on deep learning |
| CN109977948A (en) * | 2019-03-20 | 2019-07-05 | 哈尔滨工业大学 | A kind of stirring friction welding seam defect identification method based on convolutional neural networks |
| CN110570410A (en) * | 2019-09-05 | 2019-12-13 | 河北工业大学 | Detection method for automatically identifying and detecting weld defects |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11010888B2 (en) * | 2018-10-29 | 2021-05-18 | International Business Machines Corporation | Precision defect detection based on image difference with respect to templates |
-
2020
- 2020-06-10 CN CN202010525671.5A patent/CN111681231B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109285139A (en) * | 2018-07-23 | 2019-01-29 | 同济大学 | A kind of x-ray imaging weld inspection method based on deep learning |
| CN109472769A (en) * | 2018-09-26 | 2019-03-15 | 成都数之联科技有限公司 | A kind of bad image defect detection method and system |
| CN109859177A (en) * | 2019-01-17 | 2019-06-07 | 航天新长征大道科技有限公司 | Industrial x-ray image assessment method and device based on deep learning |
| CN109977948A (en) * | 2019-03-20 | 2019-07-05 | 哈尔滨工业大学 | A kind of stirring friction welding seam defect identification method based on convolutional neural networks |
| CN110570410A (en) * | 2019-09-05 | 2019-12-13 | 河北工业大学 | Detection method for automatically identifying and detecting weld defects |
Non-Patent Citations (2)
| Title |
|---|
| Welding defects detection based on deep learning with multiple optical sensors during disk laser welding of thick plates;Yanxi Zhang 等;《Journal of Manufacturing Systems》;20190428;全文 * |
| 计算机视觉在船舶焊缝缺陷识别的应用;王磊;《舰船科学技术》;20180423;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111681231A (en) | 2020-09-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111681231B (en) | Industrial welding image defect detection method based on target detection | |
| Chondronasios et al. | Feature selection for surface defect classification of extruded aluminum profiles | |
| CN109900711A (en) | Workpiece, defect detection method based on machine vision | |
| CN106651857B (en) | A kind of printed circuit board patch defect inspection method | |
| CN115797354B (en) | Method for detecting appearance defects of laser welding seam | |
| CN109615609A (en) | A kind of solder joint flaw detection method based on deep learning | |
| CN109859177A (en) | Industrial x-ray image assessment method and device based on deep learning | |
| CN111815555A (en) | Metal additive manufacturing image detection method and device combining anti-neural network with local binary | |
| CN110443791B (en) | Workpiece detection method and device based on deep learning network | |
| CN110596120A (en) | Glass boundary defect detection method, device, terminal and storage medium | |
| CN110264457A (en) | Weld seam autonomous classification method based on rotary area candidate network | |
| CN113034488B (en) | Visual inspection method for ink-jet printed matter | |
| CN109886960A (en) | The method of glass edge defects detection based on machine vision | |
| CN111539927B (en) | Detection method of automobile plastic assembly fastening buckle missing detection device | |
| CN114897908B (en) | Machine vision-based method and system for analyzing defects of selective laser powder spreading sintering surface | |
| CN110096980A (en) | Character machining identifying system | |
| CN114119591A (en) | Display screen picture quality detection method | |
| CN115018827A (en) | Automatic detection method for quality of building material weld joint | |
| CN115656182A (en) | Sheet material point cloud defect detection method based on tensor voting principal component analysis | |
| CN107121063A (en) | The method for detecting workpiece | |
| CN116205877A (en) | Intelligent detection method for weld joint surface defects based on deep reinforcement learning | |
| CN113793321B (en) | Casting surface defect dynamic detection method and device based on machine vision | |
| CN115953387A (en) | Radiographic image weld defect detection method based on deep learning | |
| KR20240039812A (en) | Defect detecting method and apparatus using deep learning | |
| Kiddee et al. | A geometry based feature detection method of V-groove weld seams for thick plate welding robots |
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 |