CN105938620B - A kind of small-bore pipe inside weld Surface Defect Recognition device - Google Patents
A kind of small-bore pipe inside weld Surface Defect Recognition device Download PDFInfo
- Publication number
- CN105938620B CN105938620B CN201610232454.0A CN201610232454A CN105938620B CN 105938620 B CN105938620 B CN 105938620B CN 201610232454 A CN201610232454 A CN 201610232454A CN 105938620 B CN105938620 B CN 105938620B
- Authority
- CN
- China
- Prior art keywords
- weld
- endoscope
- image
- industrial
- training
- 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.)
- Expired - Fee Related
Links
- 230000007547 defect Effects 0.000 title claims abstract description 31
- 238000001514 detection method Methods 0.000 claims abstract description 14
- 238000012360 testing method Methods 0.000 claims abstract description 11
- 238000005286 illumination Methods 0.000 claims abstract description 4
- 238000003466 welding Methods 0.000 claims description 28
- 238000000034 method Methods 0.000 claims description 22
- 230000006870 function Effects 0.000 claims description 17
- 238000013528 artificial neural network Methods 0.000 claims description 16
- 238000012549 training Methods 0.000 claims description 16
- 230000005540 biological transmission Effects 0.000 claims description 10
- 238000000605 extraction Methods 0.000 claims description 7
- 210000002569 neuron Anatomy 0.000 claims description 6
- 230000007423 decrease Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 3
- 230000003044 adaptive effect Effects 0.000 claims description 2
- 239000007787 solid Substances 0.000 claims description 2
- 238000009434 installation Methods 0.000 claims 1
- 239000011159 matrix material Substances 0.000 description 7
- 238000010586 diagram Methods 0.000 description 4
- 239000011521 glass Substances 0.000 description 4
- 230000002463 transducing effect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000001788 irregular Effects 0.000 description 2
- 210000005036 nerve Anatomy 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 208000019155 Radiation injury Diseases 0.000 description 1
- 208000003464 asthenopia Diseases 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 238000003062 neural network model Methods 0.000 description 1
- 238000009659 non-destructive testing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000002945 steepest descent method Methods 0.000 description 1
- 230000031068 symbiosis, encompassing mutualism through parasitism Effects 0.000 description 1
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
- 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
- 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
-
- 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
- G06T2207/20084—Artificial neural networks [ANN]
-
- 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
- G06T2207/30136—Metal
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Length Measuring Devices By Optical Means (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
A kind of small-bore pipe inside weld Surface Defect Recognition device, the device include computer, industrial camera, adapter, industrial endoscope, Mobile portable formula light source, endoscope stationary fixture and pipe fitting to be measured.Industrial camera is connected with computer, and adapter one end is connected with industrial camera, and the other end is connected with industrial endoscope, can be used for the acquisition of weld image.Mobile portable formula light source is connected with industrial endoscope, for providing the illumination in pipe.Endoscope stationary fixture is connected with endoscope, positions for the centralized positioning of endoscope, up and down adjustment and 360 ° rotate.The recognition detection of test tube part internal weld seams surface defect is treated in completion.
Description
Technical field
The present invention relates to a kind of detection device of weld seam welding defect, especially a kind of small-bore pipe inside weld surface defect
Identification device.The device is suitable for the internal weld seams welding defect at the transducing devices osculum pipelines such as pipeline, boiler and header
Detection identification, belongs to field of non destructive testing.
Background technique
With the development of China's industrial sector, the transducing devices such as boiler, pipeline and header are as the pass for bearing high temperature and pressure
Key member, manufacturing quality are more and more paid close attention to by people.Since transducing devices should be born caused by being acted on as medium
Internal pressure, will also bear the stress as caused by internal-external temperature difference, and failure damage easily appears in interior welds region.Therefore urgent
It needs to develop a kind of effective internal weld seams weld defects detection method, is reliably transported for equipment such as pipeline, boiler and headers
Row provides technical guarantee.
Currently, lacking effective detection means to small-bore pipe internal weld seams surface welding quality, artificial detection is mostly used
Method.Due to the weld seam moved on testing staff's long focus screen, it is be easy to cause eye strain, causes missing inspection, and examine
The quality of personnel, the difference of skills and experience are tested, inevitably has deviation to the assurance of quality control standards (QCS), testing result is caused to be examined
The Subjective Factors of survey personnel are larger, are difficult to make defect the judgement of accurate quantitative analysis.In order to further increase osculum pipeline
The welding quality on inside weld surface improves the testing conditions of testing staff, and there is an urgent need to a kind of osculum pipeline inside weld surfaces to lack
Sunken automatic identifying method.
In recent years, for the identifying system for building weld defect, domestic and foreign scholars have done a large amount of research and development,
[the Assessmentofwelding defects for gas pipeline radiographs using such as Shafeek
Computer vision [J] .NDT&E International, 2004,37 (6): 301-307] development platform develop it is a
The welding defect identifying system of entitled AutoWDI.The system acquires weld seam egative film, recycles by building x-ray bombardment system
Obtained egative film is analyzed and processed by relevant image processing techniques, to obtain the relevant parameter of welding defect, and according to ginseng
Number information identifies defect.
For deficiency existing for current osculum pipeline inside weld surface defects detection, this patent proposes a kind of using in industry
The face of weld defect device of sight glass and industrial camera, method used by the present apparatus are based on gray level co-occurrence matrixes and BP nerve
The defects of network class identification judges weldering, overlap and burn-through type.Face of weld indent is realized based on technique of binocular stereoscopic vision
The identification of defect.
Summary of the invention
The purpose of the present invention is to provide a kind of small-bore pipe inside weld Surface Defect Recognition devices, are based particularly on gray scale
Co-occurrence matrix and BP neural network identification is broken, overlap and the defects of burn-through, identifies weld seam table based on technique of binocular stereoscopic vision
The identification device of concave defect in face.By the acquisition to small-bore pipe inside weld surface image, for being broken, overlap and burn-through etc.
Defect type obtains the characteristic parameter for containing different poor weld surfaces based on gray level co-occurrence matrixes, and classifier uses BP nerve net
Network butt-welding fitting workpiece face of weld defect is trained, classifies and identification.Binocular stereo vision is based on for concave surface defect
Technology obtains weld seam indent degree, realizes the identification of concave defect in face of weld.
It is proposed by the present invention based on gray level co-occurrence matrixes and BP neural network and based on the weldering of technique of binocular stereoscopic vision
The detection method of seam surface defect, basic principle are:
The weld seam different to surface welding quality carries out Image Acquisition, and the texture in weld seam usually embodies face of weld
Texture features information, the i.e. uneven characteristic of body surface.The good weld seam of welding quality is in scalelike mark shape, and texture is more
Uniformly.And welding quality it is poor weld seam its surface texture it is more mixed and disorderly, variation is irregular, and the textural characteristics of different defects
Also different.And gray level co-occurrence matrixes can be good at describing irregular texture, distinguish small morphological differences, it can be preferably
Extraction face of weld effective characteristic parameters.Its gray level co-occurrence matrixes is extracted from weld image, obtains face of weld
Characteristic parameter realizes the mathematical character of face of weld textural characteristics.
G (i, j) indicates gray level co-occurrence matrixes, and in order to simplify operation, matrix element is often indicated with probability value, i.e., by element G
(i, j) obtains the normalized value that each element is both less than 1 divided by the sum of each element S
(1) energy (ASM)
Energy reflects the uniformity coefficient and texture fineness degree of weld image intensity profile.If whole numbers in co-occurrence matrix
Value is equal, then energy value is smaller, conversely, energy value is larger if matrix numerical value differs.I.e. energy value can be shown that weld seam texture
The uniformity of variation and regular degree.
(2) contrast (Contrast)
The contrast of weld image can be understood as the clarity of weld image, i.e. weld seam clean mark degree.In weld seam
In image, the groove of texture is deeper, then its contrast C ON is bigger.
(3) correlation (Correlation)
Wherein:
Correlation has reacted the consistency of weld image texture.It is horizontal if there is horizontal direction texture in weld image
The COR of direction matrix is greater than the COR value of remaining direction matrix.Correlation metrizable spaces gray level co-occurrence matrixes element is in ranks side
Upward similarity degree, value size reflect the correlation of local gray level in weld image.When matrix element value homogeneous phase etc.
When, correlation is big, otherwise correlation is smaller.
(4) homogeney (Homogeneity)
The size of the mensurable weld image texture localized variation amount of homogeney.If in certain region the variable quantity of image texture compared with
Small, then the homogeney in the region is larger.
(5) entropy (entropy)
Entropy can reflect information content possessed by weld image, indicate the non-uniform degree of texture or complicated journey in weld image
Degree.
(6) variance (variance)
Wherein u is the mean value of gray level co-occurrence matrixes each element.
This method carries out Classification and Identification to face of weld defect image using BP neural network, and BP network can learn and store
A large amount of input-output mode map relationship.Its learning rules are constantly to be adjusted using steepest descent method by backpropagation
The weight and threshold value of whole network keep the error sum of squares of network minimum.BP neural network model topology structure includes input layer, hidden
Containing layer and output layer.
BP algorithm is made of two processes of the forward calculation (forward-propagating) of data flow and the backpropagation of error signal.Just
To when propagating, the direction of propagation of data flow is again to output layer from input layer to hidden layer, and the state of every layer of neuron only influences
Next layer of neuron.If cannot get desired output, the backpropagation process of turning error signal, error signal in output layer
The direction of propagation be that output layer to hidden layer arrives input layer again.Alternately by the two processes, it is held in weight vector space
Row error function gradient decline strategy, one group of weight vector of dynamic iterative search make network error function reach minimum value, thus complete
At information extraction and Memory Process.
If the input layer of BP neural network has n node, hidden layer has q neuron, and output layer has m neuron, defeated
Entering the weight between layer and hidden layer is vki, the weight between hidden layer and output layer is wjk, the transmission function of hidden layer is f1
(), the transmission function of output layer are f2(·)。
The output of hidden layer node is (threshold value is written in sum term):
Export the output of node layer are as follows:
By formula (11) and formula (12), BP neural network just completes n-dimensional space vector (input layer) to m-dimensional space vector
The approximate mapping of (output layer).
Error function, that is, objective function sets P learning sample (x of input1x2…xp), p-th of sample obtains after inputting network
Output isThe error for defining p-th of sample is Ep, the mean square error of P sample is E.In order to avoid repeatedly
Generation number increases, and arithmetic speed slows down, and guarantees that the error of each sample is reduced, therefore choose mean square error as objective function.
Mean square error:
Network global error E is calculated according to formula (13), and compared with preset anticipation error, whether judges network error
It meets the requirements.When error reaches the precision of anticipation error or study number is greater than the number of iterations of setting, then e-learning knot
Beam;Otherwise, next learning sample and corresponding desired output are chosen, back to the net input value of neuronCarry out data forward-propagating, into next round learn, until meet error precision or
Learn to terminate learning process when number.Finally show the recognition result of weld seam welding quality.
The present invention completes the information extraction of weld seam hollow depth using binocular vision technology.Zhang Zhengyou calibration method pair is used first
The internal reference of camera is demarcated, and determines the focal length of camera;Stereo matching is carried out to image, disparity map is obtained, with curve matching
Method disparity map is refined;Finally use principle of triangulationWherein f is focal length, and b is parallax range, d
For parallax value, three-dimensional information building is completed, chooses at weld edge at 3 points and calculates three selected weld edge points extremely
The distance of camera is averaged the distance l for finding out weld edge to camera1, chosen a bit in Weld pipe mill, seek Weld pipe mill position
It sets to the distance l of camera2, l1And l2Difference be that can determine whether the hollow depth of weld seam.
Technical scheme is as follows:
Device of the present invention referring to Fig. 1, including computer 1, industrial camera 2, adapter 3, industrial endoscope 4,
Mobile portable formula light source 5, endoscope stationary fixture 6 and osculum pipeline 7 to be measured.Industrial camera 2 is connected with computer 1, adapter 3
One end is connected with industrial camera 2, and the other end is connected with industrial endoscope 4, computer 1, industrial camera 2, adapter 3, industry in
Sight glass 4 forms acquisition unit and is acquired to weld image.Mobile portable formula light source 5 is connected with industrial endoscope 4, for providing
7 intraoral illumination of osculum pipeline to be measured.Endoscope stationary fixture 6 is connected with industrial endoscope 4, and the center for industrial endoscope 4 is fixed
Position, up and down adjustment positioning and 360 ° of rotations.Complete the recognition detection to 7 internal weld seams surface defect of small orifice road to be measured.
Endoscope stationary fixture 6 include base platform 8, clamp 9, lead screw 10, endoscope fixed card slot 11, ball bearing 12,
Movable bearing support 13, connecting rod 14;Clamp 9 is mounted on 8 surface of base platform, lead screw 10 is fixed among clamp 9, endoscope is solid
Determine card slot 11 to be mounted on lead screw 10;It is matched between the centre of base platform 8 and movable bearing support 13 by ball bearing 12, activity
The bottom of support 13 is connect with connecting rod 14.
The invention has the following advantages that (1), which substitutes Traditional x-ray with endoscope, acquires face of weld image, traditional X is avoided
Radiation injury of the ray to human body.(2) method based on gray level co-occurrence matrixes and BP neural network is used, can identify small orifice
Diameter inside weld welding quality situation avoids causing missing inspection because of artificial detection fatigue.
Detailed description of the invention
Fig. 1 detection device system diagram;
Fig. 2 endoscope stationary fixture schematic diagram;
Fig. 3 this inside weld Surface Defect Recognition flow chart
Fig. 4 (a)-Fig. 4 (e) training sample image schematic diagram;
Fig. 5 welding quality good sample characteristic parameter schematic table;
Fig. 6 is broken sample characteristics parameter schematic table;
Fig. 7 hole sample characteristics parameter schematic table;
Fig. 8 is partially welded sample characteristics parameter schematic table;
Fig. 9 overlap sample characteristics parameter schematic table;
Figure 10 BP neural network exports result table;
The weld image of the left and right position Figure 11 (a)-Figure 11 (b) acquisition;
Figure 12 disparity map;
In Fig. 1: 1, computer, 2, industrial camera, 3, adapter, 4, industrial endoscope, 5, Mobile portable formula light source, 6, interior
Sight glass stationary fixture, 7, osculum pipeline to be measured.
In Fig. 2: 8, base platform, 9, clamp, 10, lead screw, 11, endoscope fixed card slot, 12, ball bearing, 13, activity branch
Seat, 14, connecting rod.
Specific embodiment
Below with reference to specific experiment, the invention will be further described:
Step 1: experimental system is built: installing pilot system according to the detection device system diagram of Fig. 1, system includes calculating
Machine 1, volume are 29mm × 29mm × 57mm industrial camera 2, F35 optics adapter 3, diameter are 8mm industry rigid endoscope
4, Mobile portable formula light source 5 and endoscope stationary fixture 6.Industrial camera 2 is connected with computer 1,3 one end of adapter and industrial phase
Machine 2 is connected, and the other end is connected with industrial endoscope 4, can be used for the acquisition of weld image.In Mobile portable formula light source 5 and industry
Sight glass 4 is connected, for providing the illumination in pipe.Endoscope stationary fixture 6 is connected with endoscope 4, and the center for endoscope is fixed
Position and up and down adjustment positioning, detect the osculum pipeline that internal diameter is 30mm.
Step 2: capturing sample image: choose first 5 class different surfaces welding qualities image (it is good, hole, be partially welded,
Overlap and be broken), totally 5 × 50 width images are as training sample, (different welding quality face of weld images such as Fig. 4 of acquisition
(a) shown in-Fig. 4 (e)) gray level co-occurrence matrixes are based on, extract energy, the contrast, correlation of different welding quality faces of weld
The characteristic parameters such as property, homogeney, entropy and variance, and characteristic parameter is normalized in the characteristic parameter of extraction.(difference weldering
The characteristic parameter of quality face of weld is connect as shown in Fig. 5 to Fig. 9).
Step 3: the foundation of BP neural network: training sample is input in neural network, and the first of neural network is arranged
Initial value parameter, wherein input layer is set as 24, and output layer is set as 5, and hidden layer neuron is set as 20.Hidden layer transmission function
Logsig transmission function is selected, output layer transmission function selects logsig transmission function, and training algorithm selects gradient decline adaptive
Answer learning training function.Maximum frequency of training is set as 3000 times, training precision 0.01, learning rate 0.01 shows training
Between be divided into 50.
Step 4: the identification of face of weld welding quality: 40 width images of acquisition carry out Classification and Identification, are based on gray scale symbiosis square
Battle array, extracts the characteristic parameters such as energy, contrast, correlation, homogeney, entropy and variance of acquisition image, and by the feature of extraction
Characteristic parameter is normalized in parameter.Acquired image pass through neural network recognization, calculate test sample with it is all kinds of
The matching degree of different surfaces welding quality training sample is chosen and 5 class different surfaces welding quality test sample matching degree highests
One kind, identify the surface welding quality of test sample.The general classification recognition correct rate of the method is 92.5% (BP nerve
The results are shown in Figure 10 for network output)
Step 5: weld seam hollow depth detection;Zhang Zhengyou calibration method is used to calibrate the focal length of camera as 30mm.To preliminary
It is judged as that the good pipeline of welding quality is detected;The reflective lesser station acquisition piece image of welded seam area is chosen (as schemed
Shown in 11 (a)), using motor drive, clamp 2 in mobile Fig. 2, moving distance 2mm, the second width image of acquisition is (as schemed
Shown in 11 (b)).Stereo matching is carried out to image and obtains initial parallax figure, then using curve matching refinement disparity map (such as Figure 12 institute
Show), utilize principle of triangulationComplete three-dimensional information building.3 weld edge points are chosen in central area to make even
The distance for obtaining weld edge to camera is 6.6758mm, and taking the distance of Weld pipe mill point is 6.0022mm, the two range difference
0.6736mm can be approximately that Weld pipe mill region is relative height differential with weld edge region.
It is a typical case of the invention above, it is of the invention using without being limited thereto.
Claims (1)
1. the inspection of the face of weld defect based on gray level co-occurrence matrixes and BP neural network and based on technique of binocular stereoscopic vision
Survey method, it is characterised in that:
Step 1: experimental system is built: installation experimental system, experimental system include computer, volume be 29mm × 29mm ×
Industrial camera, F35 optics adapter, the diameter of 57mm is that 8mm industry rigid endoscope, Mobile portable formula light source and endoscope are solid
Clamp tool;Industrial camera is connected with computer, and adapter one end is connected with industrial camera, and the other end is connected with industrial endoscope,
It can be used for the acquisition of weld image;Mobile portable formula light source is connected with industrial endoscope, for providing the illumination in pipe;Endoscope
Stationary fixture is connected with endoscope, centralized positioning and up and down adjustment positioning for endoscope, the osculum for being 30mm to internal diameter
Pipeline is detected;
Step 2: capturing sample image: choosing the image of 5 class different surfaces welding qualities first, and totally 5 × 50 width images are as instruction
Practice sample, is based on gray level co-occurrence matrixes, extracts energy, the contrast, correlation, homogeneity of different welding quality faces of weld
Property, entropy and Variance feature parameter, and characteristic parameter is normalized in the characteristic parameter of extraction;
Step 3: the foundation of BP neural network: training sample is input in neural network, and the initial value of neural network is arranged
Parameter, wherein input layer is set as 24, and output layer is set as 5, and hidden layer neuron is set as 20;The selection of hidden layer transmission function
Logsig transmission function, output layer transmission function select logsig transmission function, and training algorithm selects gradient decline is adaptive to learn
Practise training function;Maximum frequency of training is set as 3000 times, training precision 0.01, learning rate 0.01 shows training interval
It is 50;
Step 4: the identification of face of weld welding quality: 40 width images of acquisition carry out Classification and Identification, are based on gray level co-occurrence matrixes, mention
Energy, contrast, correlation, homogeney, entropy and the Variance feature parameter of acquisition image are taken, and the characteristic parameter of extraction is carried out
Normalized characteristic parameter;Acquired image passes through neural network recognization, calculates test sample and all kinds of different surfaces
The matching degree of welding quality training sample is chosen and 5 highest one kind of class different surfaces welding quality test sample matching degree, knowledge
Not Chu test sample surface welding quality;
Step 5: the detection of weld seam hollow depth: Zhang Zhengyou calibration method is used to calibrate the focal length of camera as 30mm;To preliminary judgement
It is detected for the good pipeline of welding quality;The reflective lesser station acquisition piece image of welded seam area is chosen, utilization is electronic
Machine driving, mobile clamp, moving distance 2mm acquire the second width image;Stereo matching is carried out to image and obtains initial parallax
Figure, then disparity map is refined using curve matching, utilize principle of triangulationThree-dimensional information building is completed, f is focal length,
B is parallax range, and d is parallax value;Central area choose 3 weld edge points take average acquisition weld edge to camera away from
It is 6.0022mm from the distance for for 6.6758mm, taking Weld pipe mill point, the two range difference 0.6736mm is approximately Weld pipe mill area
Domain is the relative height differential with weld edge region.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610232454.0A CN105938620B (en) | 2016-04-14 | 2016-04-14 | A kind of small-bore pipe inside weld Surface Defect Recognition device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610232454.0A CN105938620B (en) | 2016-04-14 | 2016-04-14 | A kind of small-bore pipe inside weld Surface Defect Recognition device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105938620A CN105938620A (en) | 2016-09-14 |
CN105938620B true CN105938620B (en) | 2018-12-25 |
Family
ID=57151425
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610232454.0A Expired - Fee Related CN105938620B (en) | 2016-04-14 | 2016-04-14 | A kind of small-bore pipe inside weld Surface Defect Recognition device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105938620B (en) |
Families Citing this family (27)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106338521B (en) * | 2016-09-22 | 2019-04-12 | 华中科技大学 | Increasing material manufacturing surface and internal flaw and pattern composite detection method and device |
CN106530291B (en) * | 2016-10-28 | 2019-01-22 | 重庆工商职业学院 | A kind of polished surface detection method and its system based on image procossing |
GB201711412D0 (en) * | 2016-12-30 | 2017-08-30 | Maxu Tech Inc | Early entry |
CN106910191B (en) * | 2017-03-06 | 2019-10-18 | 中国科学院合肥物质科学研究院 | A kind of bearing channel recognition methods based on depth image |
CN107421742B (en) * | 2017-08-30 | 2020-01-14 | 武汉理工大学 | Water lubrication tail bearing sticking-slipping phenomenon testing system |
CN107705312B (en) * | 2017-08-30 | 2021-02-26 | 合肥工业大学 | Method for extracting welded seam edge points based on line scanning data |
CN107941822A (en) * | 2017-11-28 | 2018-04-20 | 南通尚力机电工程设备有限公司 | A kind of fluting test method based on image procossing |
CN107971659A (en) * | 2017-11-28 | 2018-05-01 | 南通尚力机电工程设备有限公司 | A kind of on-line monitoring method of weld seam ending |
CN108846837B (en) * | 2018-06-01 | 2021-11-16 | 联想(北京)有限公司 | Object surface defect detection method and device |
CN109540903A (en) * | 2018-11-28 | 2019-03-29 | 合肥常青机械股份有限公司 | Automated production procedure quality intelligent diagnosing method based on pattern-recognition |
CN109829883B (en) * | 2018-12-19 | 2020-11-17 | 歌尔光学科技有限公司 | Product quality detection method and device |
CN109613012A (en) * | 2019-02-20 | 2019-04-12 | 南京工业大学 | Endoscope-based product defect detection system and detection method |
CN110161051B (en) * | 2019-03-07 | 2020-09-04 | 上海交通大学 | Rotary endoscope clamp suitable for tail end of manipulator and inner hole image acquisition method |
CN110412047A (en) * | 2019-07-18 | 2019-11-05 | 沈阳航空航天大学 | A kind of inner wall of the pipe slot measurement system and method |
CN112683787A (en) * | 2019-10-17 | 2021-04-20 | 神讯电脑(昆山)有限公司 | Object surface detection system and detection method based on artificial neural network |
CN112683921A (en) * | 2019-10-17 | 2021-04-20 | 神讯电脑(昆山)有限公司 | Image scanning method and image scanning system for metal surface |
CN110969611A (en) * | 2019-12-03 | 2020-04-07 | 广州特种承压设备检测研究院 | Pipeline weld defect detection method, device and system and storage medium |
CN111239160B (en) * | 2020-03-24 | 2020-12-25 | 航天云网数据研究院(广东)有限公司 | Distributed uniform grade measuring platform applying big data server |
CN111861974B (en) * | 2020-03-24 | 2021-04-13 | 江苏米多来数据科技有限公司 | Method for measuring distribution uniformity grade of application big data server |
CN111664807B (en) * | 2020-04-24 | 2022-08-12 | 河北津西钢铁集团股份有限公司 | Rotary structure optical method pipeline detection system |
CN111340809A (en) * | 2020-05-21 | 2020-06-26 | 征图新视(江苏)科技股份有限公司 | Multi-station detection equipment based on flow association deep learning model |
CN112083017A (en) * | 2020-09-10 | 2020-12-15 | 上海航天精密机械研究所 | Intelligent detection device, method and medium for internal defects of welding seam |
CN112161579B (en) * | 2020-09-17 | 2022-06-17 | 武大巨成结构股份有限公司 | Anchor hole inner diameter measuring method |
CN112845159B (en) * | 2020-12-02 | 2023-01-24 | 河南航天液压气动技术有限公司 | Pipeline visual detection system and method based on machine learning |
CN112730428A (en) * | 2020-12-23 | 2021-04-30 | 联合汽车电子有限公司 | Extension rod weld joint detection system and method |
CN116416183A (en) * | 2021-12-29 | 2023-07-11 | 广东利元亨智能装备股份有限公司 | Weld quality detection area determination method, device, computer and storage medium |
CN117817223B (en) * | 2024-03-05 | 2024-07-05 | 宝鸡康盛精工精密制造有限公司 | Welding seam identification method for robot welding |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2010142731A1 (en) * | 2009-06-10 | 2010-12-16 | Siemens Aktiengesellschaft | Method and device for testing a weld joint for a shaft by means of a detection device introduced through a passage of the shaft; corresponding rotor shaft |
CN103439408A (en) * | 2013-08-20 | 2013-12-11 | 北京巴布科克·威尔科克斯有限公司 | Ultrasonic detection method for weld joint of small-caliber pipe shelf angle |
-
2016
- 2016-04-14 CN CN201610232454.0A patent/CN105938620B/en not_active Expired - Fee Related
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2010142731A1 (en) * | 2009-06-10 | 2010-12-16 | Siemens Aktiengesellschaft | Method and device for testing a weld joint for a shaft by means of a detection device introduced through a passage of the shaft; corresponding rotor shaft |
CN103439408A (en) * | 2013-08-20 | 2013-12-11 | 北京巴布科克·威尔科克斯有限公司 | Ultrasonic detection method for weld joint of small-caliber pipe shelf angle |
Non-Patent Citations (2)
Title |
---|
内窥镜检测方法技术研究;单黎波 等;《火箭推进》;20060831;第32卷(第4期);第54-62页 * |
多环片零件轴孔装配控制与检测技术研究;张涛;《中国优秀硕士学位论文全文数据库工程科技Ⅰ辑》;20110515(第05期);第B022-420页 * |
Also Published As
Publication number | Publication date |
---|---|
CN105938620A (en) | 2016-09-14 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105938620B (en) | A kind of small-bore pipe inside weld Surface Defect Recognition device | |
CN110047073B (en) | X-ray weld image defect grading method and system | |
CN104977313A (en) | Method and device for detecting and identifying X-ray image defects of welding seam | |
CN111091538B (en) | Automatic identification and defect detection method and device for pipeline welding seams | |
CN112766103B (en) | Machine room inspection method and device | |
CN108460760B (en) | Bridge crack image distinguishing and repairing method based on generation type countermeasure network | |
CN105938563A (en) | Weld surface defect identification method based on image texture | |
CN108090896B (en) | Wood board flatness detection and machine learning method and device and electronic equipment | |
CN106679567A (en) | Contact net and strut geometric parameter detecting measuring system based on binocular stereoscopic vision | |
EP3223249B1 (en) | Paper currency fold recognition apparatus and method | |
CN105787499B (en) | The camouflaged target recognition methods extracted based on K-means cluster and polarization information | |
CN114034704A (en) | Waste roller three-dimensional detection device and method based on multi-source data fusion | |
CN110232387A (en) | A kind of heterologous image matching method based on KAZE-HOG algorithm | |
CN114894642B (en) | Fatigue crack propagation rate testing method and device based on deep learning | |
CN104634248A (en) | Revolving shaft calibration method under binocular vision | |
CN109754406A (en) | Lithium battery pole slice burr detection device and method based on two-dimensional silhouette instrument | |
CN116105604A (en) | Steel pipe quality detection system and detection method | |
CN112881424A (en) | AI + fluorescence permeation small-sized pipe surface defect detection and quality grading method and system | |
TW202219494A (en) | A defect detection method and a defect detection device | |
CN116678826A (en) | Appearance defect detection system and method based on rapid three-dimensional reconstruction | |
CN105572133B (en) | Flaw detection method and device | |
CN113421236A (en) | Building wall surface water leakage apparent development condition prediction method based on deep learning | |
CN116930192A (en) | High-precision copper pipe defect detection method and system | |
CN108831844A (en) | Detect the method and system of wafer defect | |
CN116721080A (en) | Small-diameter pipe welding defect detection method based on shape and position sensing and balanced sampling |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20181225 |