CN108872243A - A kind of bearing roller detection method of surface flaw, system and device - Google Patents
A kind of bearing roller detection method of surface flaw, system and device Download PDFInfo
- Publication number
- CN108872243A CN108872243A CN201810396818.8A CN201810396818A CN108872243A CN 108872243 A CN108872243 A CN 108872243A CN 201810396818 A CN201810396818 A CN 201810396818A CN 108872243 A CN108872243 A CN 108872243A
- Authority
- CN
- China
- Prior art keywords
- image
- bearing roller
- light stream
- layered
- adjacent
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- 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
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
- G01N21/952—Inspecting the exterior surface of cylindrical bodies or wires
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Landscapes
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The present invention discloses a kind of bearing roller detection method of surface flaw, system and device, the method includes：The present invention first obtains m to adjacent bearing roller original image；Cutting processing is carried out to the adjacent bearing roller original image to m, obtains m to adjacent bearing roller image；Singular value decomposition reconstruction processing is carried out to the adjacent bearing roller image to m, obtains m to reconstructed image；Pyramid technology micronization processes are carried out to the reconstructed image to m, obtain m to layered image；Then light stream error model is determined；The light stream field matrix of the last layer in each layered image is determined according to the light stream error model；M light stream gray level images are determined according to the light stream field matrix of the last layer in each layered image；Threshold segmentation processing finally is carried out according to the m light stream gray level images, defect area segmentation is completed, improves the computational efficiency of optical flow field and the precision of defect Segmentation, reduce the complexity of optical flow computation.
Description
Technical field
The present invention relates to defect detecting technique fields, more particularly to a kind of bearing roller detection method of surface flaw, are
System and device.
Background technique
Key component of the bearing roller as rolling bearing, surface quality directly affect the performance of bearing, precision and
Service life.Therefore defects detection is carried out for bearing roller surface to have great importance.
Currently, Manual Visual Inspection, physical method detection and machine vision inspection can be divided into for bearing roller quality testing
It surveys.Wherein Manual Visual Inspection is main detection mode at this stage, but its influence vulnerable to subjective emotion, and detection efficiency
Low, great work intensity, easily causes missing inspection and erroneous detection, not can guarantee the quality of roller.Common physical detection methods mainly have
Sonic method, eddy current testing method, supercritical ultrasonics technology, magnaflux.Although above-mentioned lossless detection method also can be detected preferably
Bearing roller defect, but it generally requires artificially to determine in the detection process, and also has centainly to the detection environment of surrounding
Requirement.Machine Vision Detection is divided mainly at this stage has line-scan digital camera and area array cameras scheme；The detection of line-scan digital camera
Scheme, then to its relevant processing operation, can preferably handle circle by the way that acquired image is spliced into piece image
Column roller surface defect, but not applicable this kind of bearing roller of circular cone；Using the scheme of area array cameras, multiple cameras is needed to match
It closes, and the radiation modality of light source is had certain limitations.Based on the above issues, how to overcome the above problem, become this field
The problem of urgent need to resolve.
Summary of the invention
The object of the present invention is to provide a kind of bearing roller detection method of surface flaw, system and devices, to realize to axis
The acquisition of bearing roller image and detection to bearing roller surface defect, improve the accuracy of detection.
To achieve the above object, the present invention provides a kind of bearing roller detection method of surface flaw, the method includes：
M is obtained to adjacent bearing roller original image；
Cutting processing is carried out to the adjacent bearing roller original image to m, obtains m to adjacent bearing roller figure
Picture；
Singular value decomposition reconstruction processing is carried out to the adjacent bearing roller image to m, obtains m to reconstructed image；
Pyramid technology micronization processes are carried out to the reconstructed image to m, obtain m to layered image；
Determine light stream error model；
The light stream field matrix of the last layer in each layered image is determined according to the light stream error model；
M light stream gray level images are determined according to the light stream field matrix of the last layer in each layered image；
Threshold segmentation processing is carried out according to the m light stream gray level images, completes defect area segmentation.
Optionally, described that cutting processing is carried out to the adjacent bearing roller original image to m, m is obtained to adjacent
Bearing roller image, specifically includes：
Image grey level histogram is determined according to the bearing roller original image；
Using the lowest point gray value between two peak values in image grey level histogram as image segmentation threshold, bearing is rolled
It is bearing roller region that the gray value of sub- original image, which is greater than image segmentation threshold,；The gray value of bearing roller original image is small
It is background area in being equal to image segmentation threshold, multiple bearing roller binary pictures is determined according to described image grey level histogram
Picture；
It is maximum bearing roller region that the maximum image of area is chosen from multiple bearing roller binary images；
Minimum circumscribed rectangle is determined according to the maximum bearing roller region；
According to minimum circumscribed rectangle, cutting processing is carried out to adjacent bearing roller original image to m, obtains m to adjacent
Initial bearing roller image；
Position correction processing is carried out to the adjacent initial bearing roller image to m, obtains m to adjacent bearing roller
Image.
Optionally, described that singular value decomposition reconstruction processing is carried out to the adjacent bearing roller image to m, obtain m pairs
Reconstructed image specifically includes：
The adjacent bearing roller image is split, multiple singular values are obtained；
K maximum singular values are chosen from multiple singular values；
Reconstructed image is determined according to k maximum singular values.
Optionally, the determining light stream error model, specific formula are：
Wherein, Ω indicates k×The wicket of k, λ are data item regularization coefficients, and bfw indicates the weight coefficient of bilateral filtering,
U, v be respectively pixel light stream on the direction x and the direction y velocity vector, r (u, v) be velocity vector data item, ▽ u and
▽ v is respectively the gradient value of velocity vector u and the gradient value of velocity vector v, | | ▽ u | | it is the smooth item of velocity vector u, | | ▽
V | | it is the smooth item of velocity vector v, D (| | ▽ u | |) is the Diffusion Strategy of the anisotropic of the smooth item of velocity vector u, D (| |
▽ v | |) be velocity vector v smooth item anisotropic Diffusion Strategy, the constant that θ is positive,For the auxiliary of velocity vector u
Variable,The region formed for the auxiliary variable of velocity vector v, the wicket that region is k × k.
Optionally, the light stream field matrix that the last layer in each layered image is determined according to the light stream error model,
It specifically includes：
Determine the light stream of each pixel in each layered image respectively in the direction x and the direction y according to the light stream error model
Upper velocity vector；
According to the light stream of pixel each in each layered image, velocity vector determines each layering on the direction x and the direction y respectively
First layer light stream field matrix in image；
The optical flow field of the last layer in each layered image is determined according to first layer light stream field matrix in each layered image
Matrix.
Optionally, the light stream field matrix according to the last layer in each layered image determines m light stream gray level images, tool
Body includes：
According to the light stream field matrix of the last layer in each pair of layered image determine it is each to optical flow field in the layered image most
Big value and optical flow field minimum value；
According to it is each to optical flow field maximum value in the layered image and optical flow field minimum value to last in each layered image
The light stream field matrix of layer is normalized, and obtains m light stream gray level images.
The present invention also provides a kind of bearing roller surface defects detection system, the system comprises：
Module is obtained, for obtaining m to adjacent bearing roller original image；
Module is cut, for carrying out cutting processing to the adjacent bearing roller original image to m, obtains m to adjacent
Bearing roller image；
Reconstructed module obtains m for carrying out singular value decomposition reconstruction processing to the adjacent bearing roller image to m
To reconstructed image；
It is layered refinement module, for carrying out Pyramid technology micronization processes to the reconstructed image to m, obtains m to layering
Image；
Light stream error model determining module, for determining light stream error model；
Optical flow field matrix deciding module, for determining the last layer in each layered image according to the light stream error model
Light stream field matrix；
M light stream gray level image determining modules, for determining m according to the light stream field matrix of the last layer in each layered image
Zhang Guangliu gray level image；
Region segmentation module completes defect area for carrying out Threshold segmentation processing according to the m light stream gray level images
Segmentation.
Optionally, the cutting module, specifically includes：
Image grey level histogram determination unit, for determining image grayscale histogram according to the bearing roller original image
Figure；
Binary image determination unit, for being made with the lowest point gray value between two peak values in image grey level histogram
For image segmentation threshold, it is bearing roller region that the gray value of bearing roller original image, which is greater than image segmentation threshold,；By axis
It is background area that the gray value of bearing roller original image, which is less than or equal to image segmentation threshold, true according to described image grey level histogram
Fixed multiple bearing roller binary images；
Selection unit is maximum axis for choosing the maximum image of area from multiple bearing roller binary images
Bearing roller region；
Minimum circumscribed rectangle determination unit, for determining minimum circumscribed rectangle according to the maximum bearing roller region；
Unit is cut, for being cut to the adjacent bearing roller original image to m according to minimum circumscribed rectangle
Processing obtains m to adjacent initial bearing roller image；
Bearing roller image determination unit, for carrying out position correction to the adjacent initial bearing roller image to m
Processing obtains m to adjacent bearing roller image.
The present invention also provides a kind of bearing roller surface defect detection apparatus, the bearing roller surface defects detection dress
It sets including area array cameras, longitudinal screw mandrel slide unit, cross lead screw slide unit, workbench, Blue backlight, turntable, electromagnet, industrial personal computer
And turntable controller；
The cross lead screw slide unit is fixed on the workbench；The longitudinal screw mandrel slide unit is fixed on the cross lead screw
On slide unit；The area array cameras is fixed on longitudinal screw mandrel slide unit；The turntable is fixed on the workbench, is located at institute
State the front of area array cameras；The electromagnet position is mounted on the top of the turntable, for bearing roller to be sucked；The blue back
Light source is fixed on the workbench, positioned at the side of the turntable；The industrial personal computer is located at outside the workbench, passes through number
It is connected according to line with the area array cameras；The turntable controller is located at outside the workbench, is connected with the turntable.
Optionally, it is provided with guide rail on the longitudinal screw mandrel slide unit, for keeping area array cameras sliding in the longitudinal screw mandrel
It is moved on platform；It is provided with guide rail on the cross lead screw slide unit, is used for the longitudinal screw mandrel slide unit in the cross lead screw
It is moved on slide unit.
The specific embodiment provided according to the present invention, the invention discloses following technical effects：
The present invention first obtains m to adjacent bearing roller original image；To m to the adjacent bearing roller original image
Cutting processing is carried out, obtains m to adjacent bearing roller image；Singular value is carried out to the adjacent bearing roller image to m
Decomposed and reconstituted processing obtains m to reconstructed image；Pyramid technology micronization processes are carried out to the reconstructed image to m, obtain m pairs
Layered image；Then light stream error model is determined；The last layer in each layered image is determined according to the light stream error model
Light stream field matrix；M light stream gray level images are determined according to the light stream field matrix of the last layer in each layered image；Finally according to m
Zhang Suoshu light stream gray level image carries out Threshold segmentation processing, completes defect area segmentation, improves the computational efficiency of optical flow field and lacks
The precision for falling into segmentation, reduces the complexity of optical flow computation.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will be to institute in embodiment
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is bearing roller of embodiment of the present invention detection method of surface flaw flow chart；
Fig. 2 is bearing roller of embodiment of the present invention surface defects detection system construction drawing；
Fig. 3 is bearing roller of embodiment of the present invention surface defect detection apparatus structure chart；
Fig. 4 is the cutting schematic diagram of taper roller of embodiment of the present invention image；
Fig. 5 is that the embodiment of the present invention cuts to obtain the bearing roller image with scratch defects；
Fig. 6 is that the embodiment of the present invention cuts to obtain the bearing roller image with spot defects；
Fig. 7 is the bearing roller detection effect figure that the embodiment of the present invention has scratch defects；
Fig. 8 is the bearing roller detection effect figure that the embodiment of the present invention has spot defects.
Wherein, 1, area array cameras, 2, longitudinal screw mandrel slide unit, 3, cross lead screw slide unit, 4, workbench, 5, turntable, 6, electromagnetism
Iron, 7, bearing roller, 8, Blue backlight, 9, industrial personal computer, 10, turntable controller.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The object of the present invention is to provide a kind of bearing roller detection method of surface flaw, system and devices, to realize to axis
The acquisition of bearing roller image and detection to bearing roller surface defect, improve the accuracy of detection.
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
Fig. 1 is bearing roller of embodiment of the present invention detection method of surface flaw flow chart；As shown in Figure 1, the present invention provides
A kind of bearing roller detection method of surface flaw, the method includes：
Step S1：M is obtained to adjacent bearing roller original image；M is the positive integer more than or equal to 2.
Step S2：Cutting processing is carried out to the adjacent bearing roller original image to m, obtains m to adjacent bearing
Roller image；
Step S3：Singular value decomposition reconstruction processing is carried out to the adjacent bearing roller image to m, obtains m to reconstruct
Image；
Step S4：Pyramid technology micronization processes are carried out to the reconstructed image to m, obtain m to layered image；
Step S5：Determine light stream error model；
Step S6：The light stream field matrix of the last layer in each layered image is determined according to the light stream error model；
Step S7：M light stream gray level images are determined according to the light stream field matrix of the last layer in each layered image；
Step S8：Threshold segmentation processing is carried out according to the m light stream gray level images, completes defect area segmentation.
Detailed analysis is carried out to each step below：
Step S1：M is obtained to adjacent bearing roller original image；The bearing roller original image is using the present invention
Bearing roller surface defect detection apparatus obtained.
Step S2：It is described that cutting processing is carried out to adjacent bearing roller original image to m, m is obtained to adjacent bearing
Roller image, specifically includes：
Step S21：Image grey level histogram is determined according to the bearing roller original image；
Step S22：Using the lowest point gray value between two peak values in image grey level histogram as image segmentation threshold,
It is bearing roller region that the gray value of bearing roller original image, which is greater than image segmentation threshold,；By bearing roller original image
It is background area that gray value, which is less than or equal to image segmentation threshold, determines multiple bearing rollers two according to described image grey level histogram
Value image；
Step S23：It is maximum bearing rolling that the maximum image of area is chosen from multiple bearing roller binary images
Subregion；
Step S24：Minimum circumscribed rectangle is determined according to the maximum bearing roller region；
Step S25：According to minimum circumscribed rectangle, cutting processing is carried out to the adjacent bearing roller original image to m,
M is obtained to adjacent initial bearing roller image；
Step S26：Position correction processing is carried out to the adjacent initial bearing roller image to m, obtains m to adjacent
Bearing roller image；Specifically formula is：
Wherein, x1For the abscissa after correction, y1For the ordinate after correction, γ is the polar axis in polar coordinates system, and μ is
The angle that minimum circumscribed rectangle and horizontal direction are formed.
Step S3：Singular value decomposition reconstruction processing is carried out to the adjacent bearing roller image to m, obtains m to reconstruct
Image specifically includes：
Step S31：The adjacent bearing roller image is split, multiple singular values are obtained；Specifically formula is：
Wherein, A is bearing roller image, and r is the number of singular value, σiFor i-th of singular value, qiFor left singular vector, si
For right singular vector, i=1,2 ... r；
Step S32：K maximum singular values are chosen from multiple singular values；K is the positive integer more than or equal to 1.
Step S33：Reconstructed image is determined according to k maximum singular values；Specifically formula is：
Wherein, AkFor reconstructed image, k is the number of maximum singular value, σjFor j-th of maximum singular value, qjAfter reconstruct
Left singular vector, sjFor the right singular vector after reconstruct, j=1,2 ... k；
Using the above method, singular value decomposition reconstruction processing is carried out to the adjacent bearing roller image to m, obtains m
To reconstructed image.
Step S5：Determine light stream error model；Specifically formula is：
Wherein, Ω indicates k×The wicket of k, λ are data item regularization coefficients, for the ratio of condition data item and smooth item
Series of fortified passes system, bfw indicate the weight coefficient of bilateral filtering, and u (x, y), v (x, y) are respectively the light stream of pixel in the direction x and the side y
U (x, y) is abbreviated as u, v (x, y) is abbreviated as v, and r (u, v) is the number of velocity vector to write conveniently by upward velocity vector
According to item, ▽ u and ▽ v are respectively the gradient value of velocity vector u and the gradient value of velocity vector v, | | ▽ u | | for velocity vector u's
Smooth item, | | ▽ v | | it is the smooth item of velocity vector v, D (| | ▽ u | |) is the expansion of the anisotropic of the smooth item of velocity vector u
Strategy is dissipated, D (| | ▽ v | |) is the Diffusion Strategy of the anisotropic of the smooth item of velocity vector v, and the constant that θ is positive, value is very
It is small,For the auxiliary variable of velocity vector u,For the auxiliary variable of velocity vector v, the wicket that region is k × k is formed
Region.
Step S6：The light stream field matrix of the last layer in each layered image is determined according to the light stream error model, specifically
Including：
Step S61：Determine the light stream of each pixel in each layered image respectively in the direction x according to the light stream error model
With velocity vector on the direction y；The specific steps are：
Step S611：Fixed u, v, enabling the formula of the light stream error model is zero, right respectivelyDerivation is sweared
Measure equation：
Wherein, λ is data item regularization coefficient, and bfw indicates the weight coefficient of bilateral filtering, and u, v are respectively pixel
Light stream velocity vector on the direction x and the direction y, the constant that θ is positive, value very little,X is asked for the cell area image of k × k
It leads,For k × k cell area image to y derivation, r0ForWherein ItIt is I pairs of bearing roller original image
The derivation of time t；
It is determined according to above-mentioned vector equationValue；
Step S612：It is fixedAccording to formulaObtain corresponding Eulerian equationWherein, Ω indicates the wicket of k × k, and λ is data item regularization coefficient, is used to
The specific gravity relationship of condition data item and smooth item, u are the light stream velocity vector in the x direction of pixel, D (| | ▽ u | |) it is speed
The Diffusion Strategy of the anisotropic of the smooth item of vector u is spent, ▽ u is gradient value of the pixel on velocity vector u direction, | | ▽
U | | it is the smooth item of velocity vector u,For the auxiliary variable of velocity vector u,For the gradient value for standardizing vector u, θ
The constant being positive, value very little, div are divergence operator.
Specific formula be：Wherein, α, β are diffusion parameter, ▽u
For pixel velocity vector u gradient value, | | ▽ u | | be velocity vector u smooth item,For standardization vector u's
Gradient value.
Step S613：It is solved using fixed point iteration modeWherein,ForThe
N+1 iteration,For the auxiliary variable of velocity vector u, λ is data item regularization coefficient, for condition data item and smooth item
Specific gravity relationship,For dual variable,For every Anisotropic diffusion strategy of dual variable, τ is one between 0 to 1
A constant；
Step S614：According to formulaSolve u.
Step S615：The value of v is solved using above-mentioned same method for solving.
Step S62：According to the light stream of pixel each in each layered image, velocity vector is true on the direction x and the direction y respectively
First layer light stream field matrix in fixed each layered image；Specifically formula is：
w1=(du, dv)；
Wherein, w1For first layer light stream field matrix in layered image, du is the light stream of pixel in the direction x velocity vector
Differential, dv are the differential of the light stream velocity vector in y-direction of pixel；
Step S63：The last layer in each layered image is determined according to first layer light stream field matrix in each layered image
Light stream field matrix；Specifically formula is：
wk+1=wk+dwk；
Wherein, wk+1For+1 layer of light stream field matrix of kth, w in layered imagekFor kth layer light stream field matrix in layered image,
When k is 1, w1For first layer light stream field matrix in layered image.
Step S7：M light stream gray level images are determined according to the light stream field matrix of the last layer in each layered image, it is specific to wrap
It includes：
Step S71：It is determined according to the light stream field matrix of the last layer in each pair of layered image each in the layered image
Optical flow field maximum value and optical flow field minimum value；Specifically formula is：
Wherein, u ' (x, y) is the value of the light stream u of the last layer, and v ' (x, y) is the value of the light stream v of the last layer, and x, y divide
Not Wei the last layer cross, ordinate, c is natural number.
Step S72：According to it is each to optical flow field maximum value in the layered image and optical flow field minimum value to each layered image
The light stream field matrix of middle the last layer is normalized, and obtains m light stream gray level images；Specifically formula is：
Wherein, V (x, y) is the light stream scalar value of the last layer, VmaxFor optical flow field maximum value and VminFor optical flow field minimum
Value.
Step S8：Threshold segmentation processing is carried out according to the m light stream gray level images, completes defect area segmentation, specifically
Formula is：
Wherein I (x, y) is the pixel gray level threshold value after threshold process, and f (x, y) is m light stream gray level images, T (x, y)
For the threshold value of setting.
The present invention cut processing acquisition m to adjacent bearing rolling to the adjacent bearing roller original image to m
Subgraph, its purpose is to reduce calculation amount；M carries out at singular value decomposition reconstruct the adjacent bearing roller image
Reason obtains m to reconstructed image, its purpose is that the gray value of removal image respective pixel point is abnormal；To m to the reconstructed image
It carries out Pyramid technology micronization processes and obtains m to layered image, each layered image is then determined according to the light stream error model
The light stream field matrix of middle the last layer, its purpose is to the operational precisions when big displacement situation of calculating；According to each hierarchical diagram
The light stream field matrix of the last layer determines m light stream gray level images as in, carries out threshold value point according to the m light stream gray level images
Processing is cut, completes defect area segmentation, its purpose is to the precision convenient for dividing and improving defect Segmentation.
Fig. 2 is bearing roller of embodiment of the present invention surface defects detection system construction drawing；As shown in Fig. 2, the present invention also mentions
For a kind of bearing roller surface defects detection system, the system comprises：
Module 1 is obtained, for obtaining m to adjacent bearing roller original image；
Module 2 is cut, for carrying out cutting processing to the adjacent bearing roller original image to m, obtains m to adjacent
Bearing roller image；
Reconstructed module 3 obtains m for carrying out singular value decomposition reconstruction processing to the adjacent bearing roller image to m
To reconstructed image；
It is layered refinement module 4, for carrying out Pyramid technology micronization processes to the reconstructed image to m, obtains m to layering
Image；
Light stream error model determining module 5, for determining light stream error model；
Optical flow field matrix deciding module 6, for determining the last layer in each layered image according to the light stream error model
Light stream field matrix；
M light stream gray level image determining modules 7, for being determined according to the light stream field matrix of the last layer in each layered image
M light stream gray level images；
Region segmentation module 8 completes defect area for carrying out Threshold segmentation processing according to the m light stream gray level images
Regional partition.
Detailed analysis is carried out to modules below：
The cutting module 2, specifically includes：
Image grey level histogram determination unit, for determining image grayscale histogram according to the bearing roller original image
Figure；
Binary image determination unit, for being made with the lowest point gray value between two peak values in image grey level histogram
For image segmentation threshold, it is bearing roller region that the gray value of bearing roller original image, which is greater than image segmentation threshold,；By axis
It is background area that the gray value of bearing roller original image, which is less than or equal to image segmentation threshold, true according to described image grey level histogram
Fixed multiple bearing roller binary images；
Selection unit is maximum axis for choosing the maximum image of area from multiple bearing roller binary images
Bearing roller region；
Minimum circumscribed rectangle determination unit, for determining minimum circumscribed rectangle according to the maximum bearing roller region；
Unit is cut, for being carried out at cutting to adjacent bearing roller original image to m according to minimum circumscribed rectangle
Reason obtains m to adjacent initial bearing roller image；
Bearing roller image determination unit, for carrying out position correction to the adjacent initial bearing roller image to m
Processing obtains m to adjacent bearing roller image.
Fig. 3 is bearing roller of embodiment of the present invention surface defect detection apparatus structure chart.As shown in figure 3, the present invention provides
A kind of bearing roller surface defect detection apparatus, the bearing roller surface defect detection apparatus include area array cameras 1, Zong Xiangsi
Bar slide unit 2, cross lead screw slide unit 3, workbench 4, Blue backlight 8, turntable 5, electromagnet 6, industrial personal computer 9 and turntable controller
10；
The cross lead screw slide unit 3 is fixed on the workbench 4；The longitudinal screw mandrel slide unit 2 is fixed on the transverse direction
On screw rod slide unit 3；The area array cameras 1 is fixed on longitudinal screw mandrel slide unit 2；The turntable 5 is fixed on the workbench 4
On, positioned at the front of the area array cameras 1；The electromagnet 6 is mounted on the top of the turntable 3, for bearing rolling to be sucked
Son；The Blue backlight 8 is fixed on the workbench 4, positioned at the side of the turntable 5；The industrial personal computer 9 is located at described
Outside workbench 4, it is connected by data line with the area array cameras 1；The turntable controller 10 is located at outside the workbench 4
Portion is connected with the turntable 5.
Bearing roller surface defect detection apparatus of the present invention further includes hood.
The present invention is provided with guide rail on the longitudinal screw mandrel slide unit 2, for keeping area array cameras 1 sliding in the longitudinal screw mandrel
It is moved on platform 2；It is provided with guide rail on the cross lead screw slide unit 3, is used for the longitudinal screw mandrel slide unit 2 in the transverse wire
It is moved on bar slide unit 3.
The testing principle of bearing roller surface defect detection apparatus is：When bearing roller 7 reaches the center of turntable 5
When, bearing roller 7 is fixed in the unlatching of electromagnet 6 of turntable bottom, while turntable 5 starts turning, and area array cameras 1, which starts simultaneously at, to be adopted
Collect image, controls the speed of turntable 5 and the time for exposure ratio of area array cameras 1 by adjusting the numerical value on turntable controller 10
Relationship, image needed for obtaining detection.The speed of turntable 5 is empirically determined；Time for exposure will be set as Image Acquisition one
The one third of length in pixels time；Assuming that the speed of turntable is A, the size of each pixel of camera is B, then the time for exposure is B/
3*A。
The bearing roller surface defect detection apparatus that the present invention designs can be suitable for cylindrical roller and taper roller simultaneously
These two types of bearing rollers；Taper roller is irradiated using blue area array light source side simultaneously, using the irreflexive radiation modality of light source,
It can effectively reduce that metal surface is reflective, to improve the quality of collected image.
Concrete example：
Bearing roller surface defect detection apparatus designed by the invention obtains taper roller image, and Fig. 4 is circular cone rolling
The cutting schematic diagram of subgraph.As shown in figure 4, the radius of taper roller upper and lower end face is different, from roller small end face to large end face,
Its crosscutting radius surface constantly increases.When camera shoots taper roller side, camera shoots the range of small end face, as shown in Figure 4
It is 120 degree, wherein the radius of small end face is 5mm, and big end radius surface is 6mm；Since taper roller is a upper and lower end face radius
It differs, herein using the length L1 of small end face line segment AB as image interception width, intercepts side image, then section of its bottom large end face
Take length L2 equal with L1, can be calculated according to trigonometric function relationship is 92 degree, so not losing bearing roller table
Face information at least needs 4 pictures to obtain all information of roller surface；It is contemplated that the rotation precision of mechanical device and
The accuracy of dividing method has detected pit, scratch, rust staining, spot defects herein in order to prove the accuracy of this method respectively
Bearing roller original image.Every kind of defect is selected to 3 pairs of adjacent bearing roller original images of acquisition respectively, every image has
32 degree of surplus, to guarantee the accuracy of optical flow computation.
Bearing roller obtains bearing roller image by cutting after bearing roller image collecting device acquisition,
Fig. 5 is that the embodiment of the present invention cuts to obtain the bearing roller original image with scratch defects；Fig. 6 is cutting of the embodiment of the present invention
Obtain the bearing roller original image with spot defects；As seen in figs. 5-6, wherein (a), (b) are first pair of bearing roller original
Beginning image, (c), (d) be second pair of bearing roller original image, (e), (f) be third to bearing roller original image.The present invention
It is not limited to determine both the above defect, other defect can also be examined, for example, the defects of pit defect, rust staining.
Then it is detected using the above method, the result of detection is as Figure 7-8, and Fig. 7 has for the embodiment of the present invention
The bearing roller detection effect figure of scratch defects；Fig. 8 is the bearing roller detection effect that the embodiment of the present invention has spot defects
Figure, wherein (a) is the defect effect picture split for first pair of bearing roller original image, (b) for for second pair of axis
The defect effect picture that bearing roller original image is split (c) splits bearing roller original image for third
Defect effect picture can preferably divide defect area from can be seen that the present invention in Fig. 7-8.
The present invention overcomes the limitation of existing line-scan digital camera and area array cameras detection scheme, the bearing that the present invention designs is rolled
Sub- surface defect detection apparatus can be suitable for cylindrical roller and taper roller these two types bearing roller simultaneously；Pass through calculating simultaneously
Optical flow field out realizes the coarse positioning of defect area, and then carries out Threshold segmentation to the region after positioning, improves defect inspection
The precision of survey.
For the system disclosed in the embodiment, since it is corresponded to the methods disclosed in the examples, so the ratio of description
Relatively simple, reference may be made to the description of the method.
Used herein a specific example illustrates the principle and implementation of the invention, and above embodiments are said
It is bright to be merely used to help understand method and its core concept of the invention；At the same time, for those skilled in the art, foundation
Thought of the invention, there will be changes in the specific implementation manner and application range.In conclusion the content of the present specification is not
It is interpreted as limitation of the present invention.
Claims (10)
1. a kind of bearing roller detection method of surface flaw, which is characterized in that the method includes：
M is obtained to adjacent bearing roller original image；
Cutting processing is carried out to the adjacent bearing roller original image to m, obtains m to adjacent bearing roller image；
Singular value decomposition reconstruction processing is carried out to the adjacent bearing roller image to m, obtains m to reconstructed image；
Pyramid technology micronization processes are carried out to the reconstructed image to m, obtain m to layered image；
Determine light stream error model；
The light stream field matrix of the last layer in each layered image is determined according to the light stream error model；
M light stream gray level images are determined according to the light stream field matrix of the last layer in each layered image；
Threshold segmentation processing is carried out according to the m light stream gray level images, completes defect area segmentation.
2. bearing roller detection method of surface flaw according to claim 1, which is characterized in that it is described to m to the phase
Adjacent bearing roller original image carries out cutting processing, obtains m to adjacent bearing roller image, specifically includes：
Image grey level histogram is determined according to the bearing roller original image；
Using the lowest point gray value between two peak values in image grey level histogram as image segmentation threshold, by bearing roller original
It is bearing roller region that the gray value of beginning image, which is greater than image segmentation threshold,；The gray value of bearing roller original image is less than etc.
It is background area in image segmentation threshold, multiple bearing roller binary images is determined according to described image grey level histogram；
It is maximum bearing roller region that the maximum image of area is chosen from multiple bearing roller binary images；
Minimum circumscribed rectangle is determined according to the maximum bearing roller region；
According to minimum circumscribed rectangle, cutting processing is carried out to adjacent bearing roller original image to m, obtain m to it is adjacent just
Beginning bearing roller image；
Position correction processing is carried out to the adjacent initial bearing roller image to m, obtains m to adjacent bearing roller figure
Picture.
3. bearing roller detection method of surface flaw according to claim 1, which is characterized in that it is described to m to the phase
Adjacent bearing roller image carries out singular value decomposition reconstruction processing, obtains m to reconstructed image, specifically includes：
The adjacent bearing roller image is split, multiple singular values are obtained；
K maximum singular values are chosen from multiple singular values；
Reconstructed image is determined according to k maximum singular values.
4. bearing roller detection method of surface flaw according to claim 1, which is characterized in that the determining light stream error
Model, specific formula are：
Wherein, Ω indicates that the wicket of k × k, λ are data item regularization coefficients, and bfw indicates the weight coefficient of bilateral filtering, u, v
The respectively light stream of pixel velocity vector on the direction x and the direction y, r (u, v) are the data item of velocity vector, ▽ u and ▽ v
The respectively gradient value of the gradient value of velocity vector u and velocity vector v, | | ▽ u | | it is the smooth item of velocity vector u, | | ▽ v | |
For the smooth item of velocity vector v, the Diffusion Strategy of the anisotropic for the smooth item that D (| | ▽ u | |) is velocity vector u, D (| | ▽ v
| |) be velocity vector v smooth item anisotropic Diffusion Strategy, the constant that θ is positive,Become for the auxiliary of velocity vector u
Amount,The region formed for the auxiliary variable of velocity vector v, the wicket that region is k × k.
5. bearing roller detection method of surface flaw according to claim 1, which is characterized in that described according to the light stream
Error model determines the light stream field matrix of the last layer in each layered image, specifically includes：
According to the light stream error model determine the light stream of each pixel in each layered image respectively on the direction x and the direction y it is fast
Spend vector；
According to the light stream of pixel each in each layered image, velocity vector determines each layered image on the direction x and the direction y respectively
Middle first layer light stream field matrix；
The light stream field matrix of the last layer in each layered image is determined according to first layer light stream field matrix in each layered image.
6. bearing roller detection method of surface flaw according to claim 5, which is characterized in that described according to each hierarchical diagram
The light stream field matrix of the last layer determines m light stream gray level images as in, specifically includes：
It is determined according to the light stream field matrix of the last layer in each pair of layered image each to optical flow field maximum value in the layered image
With optical flow field minimum value；
According to it is each to optical flow field maximum value in the layered image and optical flow field minimum value to the last layer in each layered image
Light stream field matrix is normalized, and obtains m light stream gray level images.
7. a kind of bearing roller surface defects detection system, which is characterized in that the system comprises：
Module is obtained, for obtaining m to adjacent bearing roller original image；
Module is cut, for carrying out cutting processing to the adjacent bearing roller original image to m, obtains m to adjacent axis
Bearing roller image；
Reconstructed module obtains m counterweight for carrying out singular value decomposition reconstruction processing to the adjacent bearing roller image to m
Composition picture；
It is layered refinement module, for carrying out Pyramid technology micronization processes to the reconstructed image to m, obtains m to layered image；
Light stream error model determining module, for determining light stream error model；
Optical flow field matrix deciding module, for determining the light stream of the last layer in each layered image according to the light stream error model
Field matrix；
M light stream gray level image determining modules, for determining m light according to the light stream field matrix of the last layer in each layered image
Flow gray level image；
Region segmentation module completes defect area point for carrying out Threshold segmentation processing according to the m light stream gray level images
It cuts.
8. bearing roller surface defects detection system according to claim 7, which is characterized in that the cutting module, tool
Body includes：
Image grey level histogram determination unit, for determining image grey level histogram according to the bearing roller original image；
Binary image determination unit, for using the lowest point gray value between two peak values in image grey level histogram as figure
As segmentation threshold, it is bearing roller region that the gray value of bearing roller original image, which is greater than image segmentation threshold,；Bearing is rolled
It is background area that the gray value of sub- original image, which is less than or equal to image segmentation threshold, more according to the determination of described image grey level histogram
A bearing roller binary image；
Selection unit is maximum bearing rolling for choosing the maximum image of area from multiple bearing roller binary images
Subregion；
Minimum circumscribed rectangle determination unit, for determining minimum circumscribed rectangle according to the maximum bearing roller region；
Unit is cut, for being carried out at cutting to the adjacent bearing roller original image to m according to minimum circumscribed rectangle
Reason obtains m to adjacent initial bearing roller image；
Bearing roller image determination unit, for carrying out position correction processing to the adjacent initial bearing roller image to m,
M is obtained to adjacent bearing roller image.
9. a kind of bearing roller surface defect detection apparatus, which is characterized in that the bearing roller surface defect detection apparatus packet
It includes area array cameras, longitudinal screw mandrel slide unit, cross lead screw slide unit, workbench, Blue backlight, turntable, electromagnet, industrial personal computer and turns
Platform controller；
The cross lead screw slide unit is fixed on the workbench；The longitudinal screw mandrel slide unit is fixed on the cross lead screw slide unit
On；The area array cameras is fixed on longitudinal screw mandrel slide unit；The turntable is fixed on the workbench, is located at the face
The front of array camera；The electromagnet position is mounted on the top of the turntable, for bearing roller to be sucked；The Blue backlight
It fixes on the workbench, positioned at the side of the turntable；The industrial personal computer is located at outside the workbench, passes through data line
It is connected with the area array cameras；The turntable controller is located at outside the workbench, is connected with the turntable.
10. bearing roller surface defect detection apparatus according to claim 9, which is characterized in that in the longitudinal screw mandrel
Guide rail is provided on slide unit, for moving area array cameras on the longitudinal screw mandrel slide unit；On the cross lead screw slide unit
It is provided with guide rail, for moving the longitudinal screw mandrel slide unit on the cross lead screw slide unit.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810396818.8A CN108872243B (en) | 2018-04-28 | 2018-04-28 | Bearing roller surface defect detection method, system and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810396818.8A CN108872243B (en) | 2018-04-28 | 2018-04-28 | Bearing roller surface defect detection method, system and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108872243A true CN108872243A (en) | 2018-11-23 |
| CN108872243B CN108872243B (en) | 2020-11-13 |
Family
ID=64327214
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810396818.8A Active CN108872243B (en) | 2018-04-28 | 2018-04-28 | Bearing roller surface defect detection method, system and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108872243B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503342A (en) * | 2019-08-27 | 2019-11-26 | 深圳供电局有限公司 | The retired system of comprehensive equipment based on big data analysis technology |
| CN113222895A (en) * | 2021-04-10 | 2021-08-06 | 河南巨捷电子科技有限公司 | Electrode defect detection method and system based on artificial intelligence |
| CN113222932A (en) * | 2021-05-12 | 2021-08-06 | 上海理工大学 | Small intestine endoscope image feature extraction method based on multi-convolution neural network integrated learning |
| CN113592831A (en) * | 2021-08-05 | 2021-11-02 | 北京方正印捷数码技术有限公司 | Method and device for detecting printing error and storage medium |
| CN116900412A (en) * | 2023-09-14 | 2023-10-20 | 山东昌啸商贸有限公司 | Steel plate equidistant automatic cutting regulation and control method based on visual processing |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104657948A (en) * | 2015-02-09 | 2015-05-27 | 重庆大学 | Laser underwater imaged image denoising and enhancing method for ocean exploration |
| CN105261042A (en) * | 2015-10-19 | 2016-01-20 | 华为技术有限公司 | Optical flow estimation method and apparatus |
| CN106803265A (en) * | 2017-01-06 | 2017-06-06 | 重庆邮电大学 | Multi-object tracking method based on optical flow method and Kalman filtering |
| CN107132238A (en) * | 2017-06-28 | 2017-09-05 | 南昌航空大学 | A kind of image collecting device for bearing roller periphery |
-
2018
- 2018-04-28 CN CN201810396818.8A patent/CN108872243B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104657948A (en) * | 2015-02-09 | 2015-05-27 | 重庆大学 | Laser underwater imaged image denoising and enhancing method for ocean exploration |
| CN105261042A (en) * | 2015-10-19 | 2016-01-20 | 华为技术有限公司 | Optical flow estimation method and apparatus |
| CN106803265A (en) * | 2017-01-06 | 2017-06-06 | 重庆邮电大学 | Multi-object tracking method based on optical flow method and Kalman filtering |
| CN107132238A (en) * | 2017-06-28 | 2017-09-05 | 南昌航空大学 | A kind of image collecting device for bearing roller periphery |
Non-Patent Citations (1)
| Title |
|---|
| DU-MING TSAI等: "Low-contrast surface inspection of mura defects in liquid crystal displays using optical flow-based motion analysis", 《MACHINE VISION AND APPLICATIONS》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503342A (en) * | 2019-08-27 | 2019-11-26 | 深圳供电局有限公司 | The retired system of comprehensive equipment based on big data analysis technology |
| CN113222895A (en) * | 2021-04-10 | 2021-08-06 | 河南巨捷电子科技有限公司 | Electrode defect detection method and system based on artificial intelligence |
| CN113222932A (en) * | 2021-05-12 | 2021-08-06 | 上海理工大学 | Small intestine endoscope image feature extraction method based on multi-convolution neural network integrated learning |
| CN113592831A (en) * | 2021-08-05 | 2021-11-02 | 北京方正印捷数码技术有限公司 | Method and device for detecting printing error and storage medium |
| CN113592831B (en) * | 2021-08-05 | 2024-03-19 | 北京方正印捷数码技术有限公司 | Printing error detection method, device and storage medium |
| CN116900412A (en) * | 2023-09-14 | 2023-10-20 | 山东昌啸商贸有限公司 | Steel plate equidistant automatic cutting regulation and control method based on visual processing |
| CN116900412B (en) * | 2023-09-14 | 2024-01-09 | 山东昌啸商贸有限公司 | Steel plate equidistant automatic cutting regulation and control method based on visual processing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108872243B (en) | 2020-11-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108872243A (en) | A kind of bearing roller detection method of surface flaw, system and device | |
| CN104730091B (en) | The extraction of gas turbine blades defect and analysis method based on region segmentation detection | |
| US9311720B2 (en) | Automated saw cut correction for 3D core digital modeling from computerized tomography scanner (CTS) images | |
| CN104101600B (en) | Cross Section of CC Billet testing of small cracks method | |
| CN106383130A (en) | Short steel tube surface defect detection device based on machine vision | |
| EP0489295B1 (en) | Liquid crystal panel inspection method | |
| CN106996748A (en) | A kind of wheel footpath measuring method based on binocular vision | |
| CN108489394A (en) | A kind of large-scale sheet metal works almost T-stable automatic detection device and method | |
| CN113063795B (en) | Expressway tunnel defect position determining method and system | |
| CN105327950A (en) | Hot rolling intermediate billet camber detecting device and detecting method considering thickness change | |
| US20220076021A1 (en) | System and method for automatic visual inspection with deep learning | |
| CN113554587A (en) | Molten pool image geometric feature extraction method and system based on deep learning | |
| CN113781458A (en) | Artificial intelligence based identification method | |
| CN113935971A (en) | Method and device for detecting surface defects of composite material | |
| CN106770333B (en) | BGA defect automatic detection method | |
| Han et al. | SSGD: A smartphone screen glass dataset for defect detection | |
| CN115112682A (en) | Visual mobile phone curved glass detection device and method | |
| CN202947696U (en) | Large-diameter fiber preformed rod eccentricity testing device | |
| CN105445285A (en) | Visual detection device and method for wire rods without tension constraint | |
| CN107862679A (en) | The determination method and device of image detection region | |
| JP2940405B2 (en) | Microscope equipment | |
| CN115502940A (en) | Instrument suitable for three-dimensional detection of surface defects of precise optical element | |
| Qi et al. | Quantitative detection of minor defects in metal materials based on variation coefficient of CT image | |
| CN115047015A (en) | Online X-ray nondestructive flaw detection method and system | |
| CN113111828A (en) | Three-dimensional defect detection method and system for bearing |
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 |