CN106846313A - Surface Flaw Detection method and apparatus - Google Patents
Surface Flaw Detection method and apparatus Download PDFInfo
- Publication number
- CN106846313A CN106846313A CN201710051079.4A CN201710051079A CN106846313A CN 106846313 A CN106846313 A CN 106846313A CN 201710051079 A CN201710051079 A CN 201710051079A CN 106846313 A CN106846313 A CN 106846313A
- Authority
- CN
- China
- Prior art keywords
- image
- workpiece
- garbor
- profile
- bar shaped
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20024—Filtering details
-
- 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
- G06T2207/20036—Morphological image processing
-
- 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
- G06T2207/30164—Workpiece; Machine component
Abstract
The invention discloses a kind of Surface Flaw Detection method,The image of the workpiece of input is converted to the profile for carrying out that the workpiece is extracted after binary conversion treatment after gray level image,And extract the gray level image of the workpiece,By being filtered treatment to gray level image,After on the vertical direction of the said surface for filtering out the workpiece and from the horizontal by 5 degree of bar shaped textures of angle,Carrying out edge sharpening treatment carries out obtaining contour images after self-adaption binaryzation treatment again,Contour detecting treatment is carried out using the patch of pre-set dimension size,So as to obtain corresponding characteristic vector,Bringing the good SVM models of training in advance into carries out differentiation computing,And export the result of the differentiation computing,Therefore the image of the corresponding scope of the workpiece can quickly and accurately be extracted,Greatly reduce the operand of subsequent processes,Can also avoid interference simultaneously,Robustness is also improved while improving the degree of accuracy for differentiating.
Description
Technical field
The present invention relates to workpiece quality detection field, more particularly to a kind of Surface Flaw Detection method and apparatus.
Background technology
The continuous progress of human society have with the use of new material to a certain extent it is close contact, for example bronze and
The introducing of steel uses the productivity and production efficiency that all once greatly improved human society, therefore with epoch-making meaning
Justice.Since 20th century, as the macromolecular material of the another innovation of Material Field, increasingly deep effect the mankind
The development process of society.Plastics are a family macromolecule materials known to us, with density is small, lightweight, good insulation preformance, Jie
The advantages of electrical loss is low, chemical stability is high and wearability is good, is widely used in industry, agricultural, building, national defence sophisticated industry
Deng and people live closely bound up every field, its growth rate leapt to four big industrial materials (plastics, steel, timber and
Cement) first of, " to mould Dai Gang ", " to mould Dai Mu " are into when the development trend of this life material circle.In field of plastics processing, most often
Processing method is that injection moulding its feature is that can process external form complexity, accurate size, the plastics system that quality is dense
Product., and processing of the Shooting Technique to various plastics has good adaptability, and production capacity is higher, and is easily achieved automatic
Change.In today that plastics industry is developed rapidly, no matter it is applied to the injection machine of injection molding process quantitatively, or in kind
On all occupy critical role, so as to increase most fast in turning into current plastics forming, the most processing method of production quantity it
One.However, in the production process of injection-molded item, being typically due to the bad of injecting condition and waiting the interference of various factors and cause workpiece
Surface can often produce the aesthetic appearances such as spot, pit, cut, aberration, defect, accuracy to size or functional defect, and
And these defects are likely to largely influence product further technique the processing even performance of product.Injection system
The existing defect inspection method of the product manual method that mainly manually range estimation and offline sampling analysis are combined is detected.But
This manual detection method can only be sorted to the product with obvious shape and surface defect, and substantially be increased in defect product
Necessary intervention is made when many to production equipment, is fixed a breakdown by means such as repair apparatus, adjusting parameters, resumed production；And
And the influence of the experience, psychology and physiologic factor due to examined personnel, i.e. manual detection are subject to testing staff's subjective factor
Influence is very big, thus manual detection method poor real, false drop rate is high, and cannot be quantitatively described, so as to have impact on
The accuracy and reliability of assessment.Additionally, under than relatively hazardous working environment, it is impossible to carry out manual detection.
The content of the invention
It is a primary object of the present invention to provide a kind of Surface Flaw Detection method and device, it is intended to solve artificial inspection
The product with obvious shape and surface defect can only be sorted existing for survey method, and due to examined personnel's
The influence of subjective factor is very big, and caused poor real, false drop rate is high, and cannot be quantitatively described, so as to have impact on
The accuracy of assessment and the technical problem of reliability.
To achieve the above object, a kind of Surface Flaw Detection processing method that the present invention is provided, including：
Bar shaped texture that the surface of input is had on vertical direction and with horizontal direction on into 5 degree of bar shaped textures of angle
Workpiece image IIBe converted to gray level image IGBinary conversion treatment is carried out afterwards, so as to obtain workpiece from the two of background separation
Value image IB, wherein, the surface of the workpiece has the bar shaped texture and described image I on vertical directionIIt is perpendicular
Nogata is to coincidence；
From the binary image IBAmong extract the profile F of the workpiecei, and according to the profile FiFrom the gray scale
Image IGAmong extract the gray level image I of the workpieceGi；
By the gray level image I to the workpieceGiTreatment is filtered, the said surface of the workpiece is filtered out
In bar shaped texture on the vertical direction and the horizontal direction into after being filtered after 5 degree of bar shaped textures of angle
Image BGi；
To described image BGiCarry out edge sharpening treatment and obtain edge sharpening image EGiCarry out self-adaption binaryzation again afterwards
Contour images D is obtained after treatmentGi；
Using the patch of pre-set dimension size to the contour images DGiContour detecting treatment is carried out, so as to obtain each
Characteristic vector V corresponding to the patchij, and by all workpiece features vector VijIt is normalized and composition characteristic
Vectorial set Vi, wherein, the subscript j is the characteristic vector in the set of eigenvectors and ViWithin sequence number；
By the characteristic vector set ViAmong each characteristic vector VijThe good SVM models of training in advance are brought into be sentenced
Other computing, and export the result of the differentiation computing, wherein, the differentiation result of the SVM models include normal and defect this two
Plant and differentiate result type.
Preferably, the bar shaped texture that the surface of input is had on vertical direction and with horizontal direction on into 5 degree folder
The image I of the workpiece of the bar shaped texture at angleIBe converted to gray level image IGCarry out binary conversion treatment afterwards, thus obtain by workpiece from
The binary image I of background separationBThe step of among, the binary conversion treatment be OSTU algorithms.
Preferably, it is described from the binary image IBAmong extract the profile F of the workpiecei, and according to the profile Fi
From the gray level image IGAmong extract the gray level image I of the workpieceGiThe step of among, it is described from the binary image IB
Among extract the profile F of the workpieceiThe method for being used is erosion algorithm.
Preferably, the gray level image I by the workpieceGiTreatment is filtered, the workpiece is filtered out
Said surface the vertical direction on bar shaped texture and the horizontal direction on into 5 degree of bar shaped textures of angle it
Filtered image B is obtained afterwardsGiThe step of include：
To the gray level image I of the workpieceGiCarry out the first time Garbor filtering process acquisition first time filtered
Image AGi=IGi*g1；
To described image AGiCarry out second Garbor filtering process and obtain second filtered image BGi=
AGi*g2,
Wherein, the * is convolution algorithm, the Garbor kernel functions g of the first time Garbor filtering1With described
The Garbor kernel functions g of secondary Garbor filtering2Formula be defined as：
Long, the δ is the scale size of the Garbor kernel functions, and the θ is described to suppress angleIt is phase difference, (x, y)
It is described image IGiWith described image AGiAmong corresponding pixel points coordinate, carrying out the first time Garbor filtering process
When suppression angle, θ=90 °, suppression angle, θ=5 ° when second Garbor filtering process is carried out.
Preferably, the patch of the use pre-set dimension size is to the contour images DGiContour detecting treatment is carried out, from
And obtain the characteristic vector V corresponding to each described patchij, and by all workpiece features vector VijIt is normalized
And composition characteristic vector set ViThe step of among, the characteristic vector VijIncluding processing what is obtained by the contour detecting
5 features of gray average of the length of the profile, the position coordinates (x, y) of wide, the described profile of the profile and the profile
Value.
The present invention further provides a kind of Surface Flaw Detection device, including：
Image is input into pretreatment module, for the surface of input is had bar shaped texture on vertical direction and with level side
Upwards into 5 degree of image I of the workpiece of the bar shaped texture of angleIBe converted to gray level image IGBinary conversion treatment is carried out afterwards, so as to obtain
Must be by workpiece from the binary image I of background separationB, wherein, the surface of the workpiece has described on vertical direction
Bar shaped texture and described image IIVertical direction overlap；
Workpiece image acquisition module, for from the binary image IBAmong extract the profile F of the workpiecei, and according to
The profile FiFrom the gray level image IGAmong extract the gray level image I of the workpieceGi；
Filtering process module, for by the gray level image I to the workpieceGiTreatment is filtered, institute is filtered out
Into 5 degree of bar shapeds of angle in bar shaped texture on the vertical direction of the said surface for stating workpiece and the horizontal direction
Filtered image B is obtained after textureGi；
Profile extraction module, for described image BGiCarry out edge sharpening treatment and obtain edge sharpening image EGiAfterwards
Carry out obtaining contour images D after self-adaption binaryzation treatment againGi；
Characteristic extracting module, for the patch using pre-set dimension size to the contour images DGiCarry out contour detecting
Treatment, so as to obtain the characteristic vector V corresponding to each described patchij, and by all workpiece features vector VijEntered
Row normalization and composition characteristic vector set Vi, wherein, the subscript j is the characteristic vector in the set of eigenvectors and Vi
Within sequence number；
Output module is differentiated, for by the characteristic vector set ViAmong each characteristic vector VijBring training in advance into
Good SVM models carry out differentiation computing, and export the result for differentiating computing, wherein, the differentiation result bag of the SVM models
Both differentiate result type to include normal and defect.
Preferably, among described image input pretreatment module, the binary conversion treatment is OSTU algorithms.
Preferably, it is described from the binary image I among the workpiece image acquisition moduleBAmong extract the work
The profile F of partiThe method for being used is erosion algorithm.
Preferably, the filtering process module includes：
First time Garbor filter processing unit, for the gray level image I to the workpieceGiCarry out the filter of first time Garbor
Ripple treatment obtains the first time filtered image AGi=IGi*g1；
Second Garbor filter processing unit, for described image AGiSecond Garbor filtering process is carried out to obtain
Second filtered image BGi=AGi*g2,
Wherein, the * is convolution algorithm, the Garbor kernel functions g of the first time Garbor filtering1With described
The Garbor kernel functions g of secondary Garbor filtering2Formula be defined as：
Long, the δ is the scale size of the Garbor kernel functions, and the θ is described to suppress angleIt is phase difference, (x, y)
It is described image IGiWith described image AGiAmong corresponding pixel points coordinate, carrying out the first time Garbor filtering process
When suppression angle, θ=90 °, suppression angle, θ=5 ° when second Garbor filtering process is carried out.
Preferably, among the characteristic extracting module, the characteristic vector VijIncluding processing institute by the contour detecting
Length, the position coordinates (x, y) and the gray average 5 of the profile of wide, the described profile of the profile of the profile for obtaining
Individual characteristic value.
The present invention can quickly and accurately extract the image of the corresponding scope of the workpiece by above-mentioned pretreatment, so that
The operand of subsequent processes can be greatly reduced, while the picture material outside workpiece correspondence scope can also be avoided to sentencing
The interference of other places reason, robustness is also improved while improving the degree of accuracy for differentiating.And by above-mentioned filtering process, leach
Workpiece profile in itself and surface have bar shaped texture on vertical direction and with horizontal direction on into 5 degree of strip patterns of angle
Manage the interference and influence extracted on subsequent characteristics.Further, since to using the image after filtering process carry out edge sharpening it
Afterwards carrying out contours extract, therefore the reliability and robustness of contour feature extraction are substantially increased, at the same time using profile
The characteristics of feature has simple to operate, fast operation, good reliability as the characteristic parameter of SVM discrimination models.
Also, not in common binary processing method, the OSTU algorithms are called inter-class variance maximum algorithm, and its is main
Need to calculate first among processing procedure can by the separate optimal threshold of two classes, so that their variance within clusters are minimum, because
The characteristics of this algorithm tool can automatically obtain optimal classification threshold value according to the content of image.Also, due to the workpiece being input into
Picture material is simple, and color or intensity profile are more single, therefore the algorithm can be avoided using the algorithm to noise and mesh
Mark size is very sensitive, and cannot successfully manage the interference of the unfavorable factors such as the complex situation of picture material, with fortune
Calculate simple, speed is fast, it is possible to achieve the characteristics of real-time processing.Additionally, in some environment, if using the form of vectorization,
Can quickly cipher rounds, it is also possible to be readily employed the method for multi-threading parallel process.
Secondly as edge sensitive of the Gabor kernel functions for image, using the teaching of the invention it is possible to provide good set direction and yardstick are selected
Characteristic is selected, and it is insensitive for illumination variation, using the teaching of the invention it is possible to provide the adaptability good to illumination variation.Also, due to Gabor filters
Ripple mode is closely similar with the visual stimulus response of simple cell in human visual system, therefore is extracting the local space of target
There is good characteristic with frequency-domain information aspect.Additionally, by way of being filtered using Garbor, it is simple with computing, easily
In understanding, parameter is easy to adjust, and reduces the complexity of calculating, reduces amount of calculation, improves response speed.
Brief description of the drawings
Fig. 1 is the hardware architecture diagram of the image input unit part for realizing each embodiment of the invention；
Fig. 2 is the surface structure schematic diagram of the workpiece for realizing each embodiment of the invention；
Fig. 3 is the schematic flow sheet of Surface Flaw Detection method first embodiment of the present invention；
Fig. 4 is the schematic flow sheet of Surface Flaw Detection method second embodiment of the present invention；
Fig. 5 is the high-level schematic functional block diagram of Surface Flaw Detection device first embodiment of the present invention；
Fig. 6 is the high-level schematic functional block diagram of Surface Flaw Detection device second embodiment of the present invention.
The realization of the object of the invention, functional characteristics and advantage will be described further referring to the drawings in conjunction with the embodiments.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not intended to limit the present invention.
The apparatus of each embodiment of the invention is realized referring now to Description of Drawings.In follow-up description, use
For represent element such as " module ", " part " or " unit " suffix only for being conducive to explanation of the invention, itself
Not specific meaning.Therefore, " module " can be used mixedly with " part ".
Surface Flaw Detection method and device can be implemented in a variety of manners.For example, the place described in the present invention
Reason method and device can include mobile device device and fixing equipment device.Hereinafter it is assumed that terminal is fixing equipment device.
However, it will be understood by those skilled in the art that, in addition to the element for being used in particular for fixing purpose, implementation of the invention
The construction of mode can also apply to the apparatus of mobile type.
Fig. 1 is that the hardware configuration of the hardware architecture diagram of the image input unit part for realizing each embodiment of the invention shows
Meaning.As illustrated, being arranged on Z axis 2 for the camera 1 of image input, the height of the camera 1 can be along the Z axis 2
Vertical direction move up and down, such that it is able to adjust the focal length and field range size of the camera 1；One end of the Z axis 2 with
Y-axis 3 is connected, the two ends of the Y-axis 3 two X-axis 4 arranged in parallel with described image input block framework both sides are arranged on respectively
Connection, the Y-axis 3 can be moved along the direction of X-axis 4, such that it is able to adjust the field range of the camera 1.It is defeated in described image
The base for entering part sets light source 6, in the several ground glass roller bearings 5 of the upper parallel setting of the light source 6, so as to form one
The workpiece placement platform being made up of the ground glass roller bearing 5.The ground glass roller bearing 5 is driven so as to realize by a motor
Synchronous axial system.The angle of 45 degree of every turn of the motor one of the ground glass roller bearing 5, then the ground glass roller bearing 5
It is corresponding to rotate 45 degree, while the camera 1 will gather a width picture, and the picture is input into Surface Flaw inspection
Method and device is surveyed to be processed.Each workpiece has altogether to be needed to turn 8 45 degree, corresponds to the picture captured by each angle
Sequentially input to the Surface Flaw Detection method and device and processed, discriminate whether existing defects.
Fig. 2 is the surface structure schematic diagram of the workpiece for realizing each embodiment of the invention.As shown in the figure.The shape of the workpiece
Shape approximate cylinder type, can 360 degree of rollings；Printing opacity, and surface have bar shaped texture on vertical direction and with horizontal direction on
Into 5 degree of bar shaped textures of angle.
Reference picture 3, Fig. 3 is the schematic flow sheet of the first embodiment of Surface Flaw Detection method of the present invention.
Embodiment as shown in Figure 3, the Surface Flaw Detection method includes：
Step S10, image input pretreatment.
The surface that will be input into have bar shaped texture on vertical direction and with horizontal direction on into 5 degree of strip patterns of angle
The image I of the workpiece of reasonIBe converted to gray level image IGBinary conversion treatment is carried out afterwards, so as to obtain workpiece from background separation
Binary image IB, wherein, the surface of the workpiece has the bar shaped texture and described image I on vertical directionI's
Vertical direction overlaps.
Step S20, workpiece image are obtained.
I.e. from the binary image IBAmong extract the profile F of the workpiecei, and according to the profile FiFrom the ash
Degree image IGAmong extract the gray level image I of the workpieceGi。
Step S30, filtering process.
I.e. by the gray level image I to the workpieceGiTreatment is filtered, the said surface of the workpiece is filtered out
The vertical direction on bar shaped texture and the horizontal direction on into being filtered after 5 degree of bar shaped textures of angle
Image B afterwardsGi。
Step S40, contours extract.
I.e. to described image BGiCarry out edge sharpening treatment and obtain edge sharpening image EGiCarry out self adaptation two-value again afterwards
Contour images D is obtained after change treatmentGi。
Step S50, feature extraction.
I.e. using the patch of pre-set dimension size to the contour images DGiContour detecting treatment is carried out, so as to obtain every
Characteristic vector V corresponding to the individual patchij, and by all workpiece features vector VijIt is normalized and constitutes spy
Levy vectorial set Vi, wherein, the subscript j is the characteristic vector in the set of eigenvectors and ViWithin sequence number.
Step S60, differentiation output.
Will the characteristic vector set ViAmong each characteristic vector VijBringing the good SVM models of training in advance into is carried out
Differentiate computing, and export the result of the differentiation computing, wherein, the differentiation result of the SVM models include normal and defect this
Two kinds of differentiation result types.
By above-mentioned steps S10 and step S20, the image of the corresponding scope of the workpiece can be quickly and accurately extracted,
Such that it is able to greatly reduce the operand of subsequent processes, while the picture material outside workpiece correspondence scope can also be avoided
Interference to differentiating treatment, robustness is also improved while improving the degree of accuracy for differentiating.And by the filter of above-mentioned steps S30
Ripple treatment, leach workpiece profile in itself and surface have bar shaped texture on vertical direction and with horizontal direction on into 5 degree
Interference and influence that the bar shaped texture of angle is extracted on subsequent characteristics.Further, since to being entered using the image after filtering process
Carrying out contours extract after row edge sharpening, therefore the reliability and robustness of contour feature extraction are substantially increased, with this
There is simple to operate, fast operation, the spy of good reliability as the characteristic parameter of SVM discrimination models using contour feature simultaneously
Point.
Further, the embodiment based on above-mentioned Fig. 3, in step S10, image input preprocessing process is that is, described to be input into
Surface have bar shaped texture on vertical direction and with horizontal direction on into 5 degree of image I of the workpiece of the bar shaped texture of angleI
Be converted to gray level image IGBinary conversion treatment is carried out afterwards, so as to obtain workpiece from the binary image I of background separationBStep
Among rapid, the binary conversion treatment is OSTU algorithms.
Maximum between-cluster variance
Common binary processing method is different from, the OSTU algorithms are called inter-class variance maximum algorithm, and it is mainly located
Need to calculate first among reason process can by the separate optimal threshold of two classes, so that their variance within clusters are minimum, therefore
The characteristics of algorithm tool can automatically obtain optimal classification threshold value according to the content of image.Also, due to the workpiece figure being input into
As content is simple, color or intensity profile are more single, therefore the algorithm can be avoided using the algorithm to noise and target
Size is very sensitive, and cannot successfully manage the interference of the unfavorable factors such as the complex situation of picture material, with computing
Simply, speed is fast, it is possible to achieve the characteristics of real-time processing.Additionally, in some environment, if using the form of vectorization, can
With quickly cipher rounds, it is also possible to be readily employed the method for multi-threading parallel process.
Further, the embodiment based on above-mentioned Fig. 3, obtains in step S20, workpiece image, i.e., described from the binaryzation
Image IBAmong extract the profile F of the workpiecei, and according to the profile FiFrom the gray level image IGAmong extract the work
The gray level image I of partGiThe step of among, it is described from the binary image IBAmong extract the profile F of the workpieceiUsed
Method be erosion algorithm.
The erosion algorithm is a kind of elimination boundary point, makes the processing procedure that border is internally shunk, and it can be used to disappear
Except small and insignificant object, such that it is able to be prevented effectively from noise jamming.
Reference picture 4, Fig. 4 is the stream of the step S30 of the second embodiment of Surface Flaw Detection method of the present invention
Journey schematic diagram.As shown in figure 4, the embodiment based on above-mentioned Fig. 3, the step S30, filtering process include：
Step S310, first time Garbor filtering process.
I.e. to the gray level image I of the workpieceGiAfter carrying out the first time Garbor filtering process acquisition first time filtering
Image AGi=IGi*g1。
Step S320, second Garbor filtering process.
I.e. to described image AGiCarry out second Garbor filtering process and obtain second filtered image BGi=
AGi*g2。
Wherein, the * is convolution algorithm, the Garbor kernel functions g of the first time Garbor filtering1With described
The Garbor kernel functions g of secondary Garbor filtering2Formula be defined as：
Long, the δ is the scale size of the Garbor kernel functions, and the θ is described to suppress angleIt is phase difference, (x, y)
It is described image IGiWith described image AGiAmong corresponding pixel points coordinate, carrying out the first time Garbor filtering process
When suppression angle, θ=90 °, suppression angle, θ=5 ° when second Garbor filtering process is carried out.
The Gabor kernel functions of edge sensitive due to to(for) image, using the teaching of the invention it is possible to provide good set direction and scale selection are special
Property, and it is insensitive for illumination variation, using the teaching of the invention it is possible to provide the adaptability good to illumination variation.Also, due to Gabor filtering sides
Formula is closely similar with the visual stimulus response of simple cell in human visual system, therefore is extracting the local space and frequency of target
Rate domain information aspect has good characteristic.Additionally, by way of being filtered using Garbor, it is simple with computing, it is easy to manage
Solution, parameter is easy to adjust, and reduces the complexity of calculating, reduces amount of calculation, improves response speed.
Further, the embodiment based on above-mentioned Fig. 4, the S50, characteristic extraction procedure, i.e., described use pre-set dimension is big
Small patch is to the contour images DGiCarry out contour detecting treatment, thus obtain feature corresponding to each described patch to
Amount Vij, and by all workpiece features vector VijIt is normalized and composition characteristic vector set ViThe step of among, institute
State characteristic vector VijThe length of the profile including being obtained by contour detecting treatment, wide, the described wheel of the profile
Wide position coordinates (x, y) and 5 characteristic values of gray average of the profile.
The length of the profile obtained as a result of contour detecting treatment, wide, the described profile of the profile
Position coordinates (x, y) and the gray average of the profile this 5 characteristic values, increased said features vector VijLatitude, increase
Strong its robustness, and SVM is differentiated more mature and reliable, fast response time, it is easy to Project Realization.
Surface Flaw Detection method in the first embodiment of the invention described above Surface Flaw Detection method can
The Surface Flaw Detection device that is there is provided with the first embodiment by Surface Flaw Detection device of the present invention is realized.
Reference picture 5, Fig. 5 lacks for the first embodiment of Surface Flaw Detection device of the present invention provides a kind of workpiece surface
Detection means 100 is fallen into, the Surface Flaw Detection device 100 includes：
Image is input into pretreatment module 10, for the surface of input is had bar shaped texture on vertical direction and with level
Into 5 degree of image I of the workpiece of the bar shaped texture of angle on directionIBe converted to gray level image IGBinary conversion treatment is carried out afterwards, so that
Obtain workpiece from the binary image I of background separationB, wherein, the surface of the workpiece has the institute on vertical direction
State bar shaped texture and described image IIVertical direction overlap.Workpiece image acquisition module 20, for from the binary image IB
Among extract the profile F of the workpiecei, and according to the profile FiFrom the gray level image IGAmong extract the ash of the workpiece
Degree image IGi；
Filtering process module, for by the gray level image I to the workpieceGiTreatment is filtered, institute is filtered out
Into 5 degree of bar shapeds of angle in bar shaped texture on the vertical direction of the said surface for stating workpiece and the horizontal direction
Filtered image B is obtained after textureGi。
Filtering process module 30, for for by the gray level image I to the workpieceGiIt is filtered treatment, mistake
Into 5 degree of institutes of angle in bar shaped texture on the vertical direction of the said surface for filtering the workpiece and the horizontal direction
State acquisition filtered image B after bar shaped textureGi。
Profile extraction module 40, for described image BGiCarry out edge sharpening treatment and obtain edge sharpening image EGiIt
Carry out obtaining contour images D after self-adaption binaryzation treatment again afterwardsGi。
Characteristic extracting module 50, for the patch using pre-set dimension size to the contour images DGiCarry out profile inspection
Survey is processed, so as to obtain the characteristic vector V corresponding to each described patchij, and by all workpiece features vector VijInstitute
It is normalized and composition characteristic vector set Vi, wherein, the subscript j be the characteristic vector in the set of eigenvectors and
ViWithin sequence number.
Differentiate output module 60, for by the characteristic vector set ViAmong each characteristic vector VijBring instruction in advance into
The SVM models perfected carry out differentiation computing, and export the result for differentiating computing, wherein, the differentiation result of the SVM models
Including normal and defect, both differentiate result type.
Pretreatment module 10 and workpiece image acquisition module 20 are input into by above-mentioned image, institute can be quickly and accurately extracted
The image of the corresponding scope of workpiece is stated, such that it is able to greatly reduce the operand of subsequent processes, while work can also be avoided
Picture material outside part correspondence scope also improves robust to differentiating the interference for processing while improving the degree of accuracy for differentiating
Property.And by the filtering process of above-mentioned filtering process module 30, leaching workpiece profile in itself and surface has side vertically
Upward bar shaped texture and with subsequent characteristics are extracted interference and influence into 5 degree of bar shaped textures of angle in horizontal direction.This
Outward, due to being carried out after edge sharpening carrying out contours extract using the image after filtering process, therefore substantially increase
Reliability and robustness that contour feature is extracted, are at the same time had using contour feature as the characteristic parameter of SVM discrimination models
Have simple to operate, fast operation, the characteristics of good reliability.
Further, the embodiment based on above-mentioned Fig. 5, described image is input into the table that will be input into described in pretreatment module 10
Face have bar shaped texture on vertical direction and with horizontal direction on into 5 degree of image I of the workpiece of the bar shaped texture of angleIConversion
It is gray level image IGBinary conversion treatment is carried out afterwards, so as to obtain workpiece from the binary image I of background separationBIt is treated
Among journey, the binary conversion treatment is OSTU algorithms.
Common binary processing method is different from, the OSTU algorithms are called inter-class variance maximum algorithm, and it is mainly located
Need to calculate first among reason process can by the separate optimal threshold of two classes, so that their variance within clusters are minimum, therefore
The characteristics of algorithm tool can automatically obtain optimal classification threshold value according to the content of image.Also, due to the workpiece figure being input into
As content is simple, color or intensity profile are more single, therefore the algorithm can be avoided using the algorithm to noise and target
Size is very sensitive, and cannot successfully manage the interference of the unfavorable factors such as the complex situation of picture material, with computing
Simply, speed is fast, it is possible to achieve the characteristics of real-time processing.Additionally, in some environment, if using the form of vectorization, can
With quickly cipher rounds, it is also possible to be readily employed the method for multi-threading parallel process.
Further, the embodiment based on above-mentioned Fig. 5, from the binaryzation described in the workpiece image acquisition module 20
Image IBAmong extract the profile F of the workpiecei, and according to the profile FiFrom the gray level image IGAmong extract the work
The gray level image I of partGiProcessing procedure among, it is described from the binary image IBAmong extract the profile F of the workpieceiInstitute
The method of use is erosion algorithm.
The erosion algorithm is a kind of elimination boundary point, makes the processing procedure that border is internally shunk, and it can be used to disappear
Except small and insignificant object, such that it is able to be prevented effectively from noise jamming.
Surface Flaw Detection method in the second embodiment of the invention described above Surface Flaw Detection method can
The Surface Flaw Detection device that is there is provided with the second embodiment by Surface Flaw Detection device of the present invention is realized.
Reference picture 6, the second embodiment of Surface Flaw Detection device of the present invention provides a kind of Surface Flaw inspection
Device is surveyed, based on the embodiment shown in above-mentioned Fig. 5, the filtering process module 30 includes：
First time Garbor filter processing unit 31, for the gray level image I to the workpieceGiCarry out first time Garbor
Filtering process obtains the first time filtered image AGi=IGi*g1；
Second Garbor filter processing unit 32, for described image AGiSecond Garbor filtering process is carried out to obtain
Obtain second filtered image BGi=AGi*g2,
Wherein, the * is convolution algorithm, the Garbor kernel functions g of the first time Garbor filtering1With described
The Garbor kernel functions g of secondary Garbor filtering2Formula be defined as：
Long, the δ is the scale size of the Garbor kernel functions, and the θ is described to suppress angleIt is phase difference, (x, y)
It is described image IGiWith described image AGiAmong corresponding pixel points coordinate, carrying out the first time Garbor filtering process
When suppression angle, θ=90 °, suppression angle, θ=5 ° when second Garbor filtering process is carried out.
The Gabor kernel functions of edge sensitive due to to(for) image, using the teaching of the invention it is possible to provide good set direction and scale selection are special
Property, and it is insensitive for illumination variation, using the teaching of the invention it is possible to provide the adaptability good to illumination variation.Also, due to Gabor filtering sides
Formula is closely similar with the visual stimulus response of simple cell in human visual system, therefore is extracting the local space and frequency of target
Rate domain information aspect has good characteristic.Additionally, by way of being filtered using Garbor, it is simple with computing, it is easy to manage
Solution, parameter is easy to adjust, and reduces the complexity of calculating, reduces amount of calculation, improves response speed.
Further, the embodiment based on above-mentioned Fig. 6, pre-set dimension size is used described in the characteristic extracting module 50
Patch to the contour images DGiContour detecting treatment is carried out, so as to obtain the characteristic vector corresponding to each described patch
Vij, and by all workpiece features vector VijIt is normalized and composition characteristic vector set ViProcessing procedure among institute
State characteristic vector VijThe length of the profile including being obtained by contour detecting treatment, wide, the described wheel of the profile
Wide position coordinates (x, y) and 5 characteristic values of gray average of the profile.
The length of the profile obtained as a result of contour detecting treatment, wide, the described profile of the profile
Position coordinates (x, y) and the gray average of the profile this 5 characteristic values, increased said features vector VijLatitude, increase
Strong its robustness, and SVM is differentiated more mature and reliable, fast response time, it is easy to Project Realization.
It should be noted that herein, term " including ", "comprising" or its any other variant be intended to non-row
His property is included, so that process, method, article or device including a series of key elements not only include those key elements, and
And also include other key elements being not expressly set out, or also include for this process, method, article or device institute are intrinsic
Key element.In the absence of more restrictions, the key element limited by sentence "including a ...", it is not excluded that including this
Also there is other identical element in the process of key element, method, article or device.
The embodiments of the present invention are for illustration only, and the quality of embodiment is not represented.
Obviously, those skilled in the art should be understood that above-mentioned of the invention each modular unit or each step can be used
General computing device realized, alternatively, the program code that they can be can perform with computing device be realized, so that, can
To be stored in being performed by computing device in storage device, and in some cases, can be with different from herein
Order performs shown or described step, or they are fabricated to each integrated circuit modules respectively, or by them
Multiple modules or step single integrated circuit module is fabricated to realize.So, the present invention is not restricted to any specific hard
Part and software are combined.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but in many cases
The former is more preferably implementation method.Based on such understanding, technical scheme is substantially done to prior art in other words
The part for going out contribution can be embodied in the form of software product, and the computer software product is stored in a storage medium
In (such as ROM/RAM, magnetic disc, CD), including some instructions are used to so that a station terminal equipment (can be mobile phone, computer, clothes
Business device, air-conditioner, or network equipment etc.) perform method described in each embodiment of the invention.
The preferred embodiments of the present invention are these are only, the scope of the claims of the invention is not thereby limited, it is every to utilize this hair
Equivalent structure or equivalent flow conversion that bright specification and accompanying drawing content are made, or directly or indirectly it is used in other related skills
Art field, is included within the scope of the present invention.
Claims (10)
1. Surface Flaw Detection method, it is characterised in that including：
Bar shaped texture that the surface of input is had on vertical direction and with horizontal direction on into 5 degree of works of the bar shaped texture of angle
The image I of partIBe converted to gray level image IGBinary conversion treatment is carried out afterwards, so as to obtain workpiece from the binaryzation of background separation
Image IB, wherein, the surface of the workpiece has the bar shaped texture and described image I on vertical directionIVertical side
To coincidence；
From the binary image IBAmong extract the profile F of the workpiecei, and according to the profile FiFrom the gray level image IG
Among extract the gray level image I of the workpieceGi；
By the gray level image I to the workpieceGiBe filtered treatment, filter out the workpiece said surface it is described
Into obtaining filtered image after 5 degree of bar shaped textures of angle in bar shaped texture on vertical direction and the horizontal direction
BGi；
To described image BGiCarry out edge sharpening treatment and obtain edge sharpening image EGiCarry out self-adaption binaryzation treatment again afterwards
Contour images D is obtained afterwardsGi；
Using the patch of pre-set dimension size to the contour images DGiContour detecting treatment is carried out, so as to obtain described in each
Characteristic vector V corresponding to patchij, and by all workpiece features vector VijIt is normalized and composition characteristic vector
Set Vi, wherein, the subscript j is the characteristic vector in the set of eigenvectors and ViWithin sequence number；
By the characteristic vector set ViAmong each characteristic vector VijBringing the good SVM models of training in advance into carries out differentiation fortune
Calculate, and export the result of the differentiation computing, wherein, including normal and defect, both sentence the differentiation result of the SVM models
Other result type.
2. Surface Flaw Detection method as claimed in claim 1, it is characterised in that it is described the surface being input into is had it is perpendicular
The upward bar shaped texture of Nogata and with horizontal direction on into 5 degree of image I of the workpiece of the bar shaped texture of angleIBe converted to gray-scale map
As IGBinary conversion treatment is carried out afterwards, so as to obtain workpiece from the binary image I of background separationBThe step of among, described two
Value is processed as OSTU algorithms.
3. Surface Flaw Detection method as claimed in claim 1, it is characterised in that described from the binary image IB
Among extract the profile F of the workpiecei, and according to the profile FiFrom the gray level image IGAmong extract the ash of the workpiece
Degree image IGiThe step of among, it is described from the binary image IBAmong extract the profile F of the workpieceiThe method for being used
It is erosion algorithm.
4. Surface Flaw Detection method as claimed in claim 1, it is characterised in that the institute by the workpiece
State gray level image IGiBe filtered treatment, the bar shaped texture on the vertical direction of the said surface for filtering out the workpiece with
Into acquisition filtered image B after 5 degree of bar shaped textures of angle in the horizontal directionGiThe step of include：
To the gray level image I of the workpieceGiCarry out first time Garbor filtering process and obtain the first time filtered image
AGi=IGi*g1；
To described image AGiCarry out second Garbor filtering process and obtain second filtered image BGi=AGi*g2,
Wherein, the * is convolution algorithm, the Garbor kernel functions g of the first time Garbor filtering1With described second
The Garbor kernel functions g of Garbor filtering2Formula be defined as：
It is describedIt is describedThe λ is the wavelength of the Garbor kernel functions, described
δ is the scale size of the Garbor kernel functions, and the θ is described to suppress angleIt is phase difference, (x, y) is the figure
As IGiWith described image AGiAmong corresponding pixel points coordinate, described in when the first time Garbor filtering process is carried out
Suppress angle, θ=90 °, suppression angle, θ=5 ° when second Garbor filtering process is carried out.
5. Surface Flaw Detection method as claimed in claim 1, it is characterised in that the use pre-set dimension size
Patch is to the contour images DGiContour detecting treatment is carried out, so as to obtain the characteristic vector corresponding to each described patch
Vij, and by all workpiece features vector VijIt is normalized and composition characteristic vector set ViThe step of among, it is described
Characteristic vector VijThe length of the profile including being obtained by contour detecting treatment, wide, the described profile of the profile
Position coordinates (x, y) and the profile 5 characteristic values of gray average.
6. Surface Flaw Detection device, it is characterised in that including：
Image is input into pretreatment module, for the surface of input is had bar shaped texture on vertical direction and with horizontal direction
Into 5 degree of image I of the workpiece of the bar shaped texture of angleIBe converted to gray level image IGCarry out binary conversion treatment afterwards, thus obtain by
Binary image I of the workpiece from background separationB, wherein, the surface of the workpiece has the bar shaped on vertical direction
Texture and described image IIVertical direction overlap；
Workpiece image acquisition module, for from the binary image IBAmong extract the profile F of the workpiecei, and according to described
Profile FiFrom the gray level image IGAmong extract the gray level image I of the workpieceGi；
Filtering process module, for by the gray level image I to the workpieceGiTreatment is filtered, the work is filtered out
Into 5 degree of bar shaped textures of angle in bar shaped texture on the vertical direction of the said surface of part and the horizontal direction
Filtered image B is obtained afterwardsGi；
Filtering process module, for by the gray level image I to the workpieceGiTreatment is filtered, the work is filtered out
Into 5 degree of bar shaped textures of angle in bar shaped texture on the vertical direction of the said surface of part and the horizontal direction
Filtered image B is obtained afterwardsGi
Profile extraction module, for described image BGiCarry out edge sharpening treatment and obtain edge sharpening image EGiCarry out again afterwards
Contour images D is obtained after self-adaption binaryzation treatmentGi；
Characteristic extracting module, for the patch using pre-set dimension size to the contour images DGiContour detecting treatment is carried out,
So as to obtain the characteristic vector V corresponding to each described patchij, and by all workpiece features vector VijCarried out normalizing
Change and composition characteristic vector set Vi, wherein, the subscript j is the characteristic vector in the set of eigenvectors and ViWithin
Sequence number；
Output module is differentiated, for by the characteristic vector set ViAmong each characteristic vector VijBring training in advance into good
SVM models carry out differentiation computing, and export the result for differentiating computing, wherein, the differentiation result of the SVM models is included just
Often and defect both differentiation result types.
7. Surface Flaw Detection device as claimed in claim 6, it is characterised in that described image is input into pretreatment module
Among, the binary conversion treatment is OSTU algorithms.
8. Surface Flaw Detection device as claimed in claim 7, it is characterised in that the workpiece image acquisition module it
In, it is described from the binary image IBAmong extract the profile F of the workpieceiThe method for being used is erosion algorithm.
9. Surface Flaw Detection device as claimed in claim 6, it is characterised in that the filtering process module includes：
First time Garbor filter processing unit, for the gray level image I to the workpieceGiCarry out at the filtering of first time Garbor
Reason obtains the first time filtered image AGi=IGi*g1；
Second Garbor filter processing unit, for described image AGiCarry out second Garbor filtering process and obtain described
Second filtered image BGi=AGi*g2,
Wherein, the * is convolution algorithm, the Garbor kernel functions g of the first time Garbor filtering1With described second
The Garbor kernel functions g of Garbor filtering2Formula be defined as：
It is describedIt is describedThe λ is the wavelength of the Garbor kernel functions, described
δ is the scale size of the Garbor kernel functions, and the θ is described to suppress angleIt is phase difference, (x, y) is the figure
As IGiWith described image AGiAmong corresponding pixel points coordinate, described in when the first time Garbor filtering process is carried out
Suppress angle, θ=90 °, suppression angle, θ=5 ° when second Garbor filtering process is carried out.
10. Surface Flaw Detection device as claimed in claim 6, it is characterised in that among the characteristic extracting module,
The characteristic vector VijIncluding by the contour detecting process the length of the profile for being obtained, the profile it is wide, described
5 characteristic values of the position coordinates (x, y) of profile and the gray average of the profile.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710051079.4A CN106846313B (en) | 2017-01-23 | 2017-01-23 | Workpiece surface defect detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710051079.4A CN106846313B (en) | 2017-01-23 | 2017-01-23 | Workpiece surface defect detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106846313A true CN106846313A (en) | 2017-06-13 |
| CN106846313B CN106846313B (en) | 2020-03-10 |
Family
ID=59120802
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710051079.4A Active CN106846313B (en) | 2017-01-23 | 2017-01-23 | Workpiece surface defect detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106846313B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107680086A (en) * | 2017-09-27 | 2018-02-09 | 电子科技大学 | A kind of existing arc-shaped side has the material profile defect inspection method of straight line again |
| CN107909573A (en) * | 2017-12-04 | 2018-04-13 | 广东嘉铭智能科技有限公司 | Metal works annular surface knife mark detection method and device |
| CN108154510A (en) * | 2018-01-17 | 2018-06-12 | 深圳市亿图视觉自动化技术有限公司 | Method for detecting surface defects of products, device and computer readable storage medium |
| CN108782797A (en) * | 2018-06-15 | 2018-11-13 | 广东工业大学 | The control method and arm-type tea frying machine of arm-type tea frying machine stir-frying tealeaves |
| CN108960325A (en) * | 2018-07-03 | 2018-12-07 | 柳州市木子科技有限公司 | A kind of automobile metal plate work detection Identification of Cracks system based on SVM and Hog |
| CN109444169A (en) * | 2019-01-17 | 2019-03-08 | 南昌航空大学 | A kind of bearing defect detection method and system |
| CN110287990A (en) * | 2019-05-21 | 2019-09-27 | 山东大学 | Microalgae image classification method, system, equipment and storage medium |
| CN111390578A (en) * | 2020-06-08 | 2020-07-10 | 佛山市南海富大精密机械有限公司 | Five-axis linkage numerical control machine tool |
| CN111985497A (en) * | 2020-07-21 | 2020-11-24 | 国网山东省电力公司电力科学研究院 | Method and system for identifying crane operation under overhead transmission line |
| TWI757590B (en) * | 2019-01-31 | 2022-03-11 | 中強光電股份有限公司 | Luminance uniform detection system and luminance uniform detection method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103175844A (en) * | 2012-03-16 | 2013-06-26 | 沈阳理工大学 | Detection method for scratches and defects on surfaces of metal components |
| CN203908956U (en) * | 2013-10-25 | 2014-10-29 | 重庆锦和鸿科技有限公司 | System for detecting surface defects of workpiece |
-
2017
- 2017-01-23 CN CN201710051079.4A patent/CN106846313B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103175844A (en) * | 2012-03-16 | 2013-06-26 | 沈阳理工大学 | Detection method for scratches and defects on surfaces of metal components |
| CN203908956U (en) * | 2013-10-25 | 2014-10-29 | 重庆锦和鸿科技有限公司 | System for detecting surface defects of workpiece |
Non-Patent Citations (4)
| Title |
|---|
| SHI-CHANG DU等: "An Adaptive Support Vector Machine-Based Workpiece Surface Classification System Using High-Definition Metrology", 《IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT》 * |
| XIANGHUA XIE: "A Review of Recent Advances in Surface Defect Detection using Texture analysis Techniques", 《ELECTRONIC LETTERS ON COMPUTER VISION AND IMAGE ANALYSIS》 * |
| 唐玮: "基于机器视觉的曲面工件表面缺陷检测方法研究", 《中国优秀硕士学位论文全文数据库-信息科技辑》 * |
| 陈裕潮等: "一种基于机器视觉的轮胎模具表面字符检测方法", 《锻压技术》 * |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107680086B (en) * | 2017-09-27 | 2020-10-23 | 电子科技大学 | Method for detecting material contour defects with arc-shaped edges and linear edges |
| CN107680086A (en) * | 2017-09-27 | 2018-02-09 | 电子科技大学 | A kind of existing arc-shaped side has the material profile defect inspection method of straight line again |
| CN107909573A (en) * | 2017-12-04 | 2018-04-13 | 广东嘉铭智能科技有限公司 | Metal works annular surface knife mark detection method and device |
| CN107909573B (en) * | 2017-12-04 | 2021-09-28 | 广东嘉铭智能科技有限公司 | Method and device for detecting knife lines on annular surface of metal workpiece |
| CN108154510A (en) * | 2018-01-17 | 2018-06-12 | 深圳市亿图视觉自动化技术有限公司 | Method for detecting surface defects of products, device and computer readable storage medium |
| CN108782797A (en) * | 2018-06-15 | 2018-11-13 | 广东工业大学 | The control method and arm-type tea frying machine of arm-type tea frying machine stir-frying tealeaves |
| CN108960325A (en) * | 2018-07-03 | 2018-12-07 | 柳州市木子科技有限公司 | A kind of automobile metal plate work detection Identification of Cracks system based on SVM and Hog |
| CN109444169A (en) * | 2019-01-17 | 2019-03-08 | 南昌航空大学 | A kind of bearing defect detection method and system |
| CN109444169B (en) * | 2019-01-17 | 2021-06-15 | 南昌航空大学 | Bearing defect detection method and system |
| TWI757590B (en) * | 2019-01-31 | 2022-03-11 | 中強光電股份有限公司 | Luminance uniform detection system and luminance uniform detection method |
| CN110287990A (en) * | 2019-05-21 | 2019-09-27 | 山东大学 | Microalgae image classification method, system, equipment and storage medium |
| CN111390578A (en) * | 2020-06-08 | 2020-07-10 | 佛山市南海富大精密机械有限公司 | Five-axis linkage numerical control machine tool |
| CN111390578B (en) * | 2020-06-08 | 2020-09-08 | 佛山市南海富大精密机械有限公司 | Five-axis linkage numerical control machine tool |
| CN111985497A (en) * | 2020-07-21 | 2020-11-24 | 国网山东省电力公司电力科学研究院 | Method and system for identifying crane operation under overhead transmission line |
| CN111985497B (en) * | 2020-07-21 | 2023-09-26 | 国网山东省电力公司电力科学研究院 | Crane operation identification method and system under overhead transmission line |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106846313B (en) | 2020-03-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106846313A (en) | Surface Flaw Detection method and apparatus | |
| CN107543828A (en) | A kind of Surface Flaw Detection method and system | |
| CN105447512B (en) | A kind of detection method and device for the beauty defect that essence slightly combines | |
| CN106610969A (en) | Multimodal information-based video content auditing system and method | |
| CN108765402B (en) | Non-woven fabric defect detection and classification method | |
| CN114549522A (en) | Textile quality detection method based on target detection | |
| CN108230315B (en) | Mask belt loss detection method based on machine vision | |
| CN113554080A (en) | Non-woven fabric defect detection and classification method and system based on machine vision | |
| CN116188475B (en) | Intelligent control method, system and medium for automatic optical detection of appearance defects | |
| CN109101976B (en) | Method for detecting surface defects of arc-extinguishing grid plate | |
| CN107093175B (en) | A kind of power battery welding quality detection method | |
| CN103034838A (en) | Special vehicle instrument type identification and calibration method based on image characteristics | |
| CN103175844A (en) | Detection method for scratches and defects on surfaces of metal components | |
| CN104198497A (en) | Surface defect detection method based on visual saliency map and support vector machine | |
| CN110349125A (en) | A kind of LED chip open defect detection method and system based on machine vision | |
| CN108647706A (en) | Article identification classification based on machine vision and flaw detection method | |
| CN113012098B (en) | Iron tower angle steel punching defect detection method based on BP neural network | |
| CN112488211A (en) | Fabric image flaw classification method | |
| CN109344768A (en) | Pointer breaker recognition methods based on crusing robot | |
| CN113221881B (en) | Multi-level smart phone screen defect detection method | |
| CN113706490B (en) | Wafer defect detection method | |
| CN109615604A (en) | Accessory appearance flaw detection method based on image reconstruction convolutional neural networks | |
| CN110096980A (en) | Character machining identifying system | |
| NU | Automatic detection of texture defects using texture-periodicity and Gabor wavelets | |
| CN108460344A (en) | Dynamic area intelligent identifying system in screen and intelligent identification Method |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20201203 Address after: Fonda YUNPU Industrial Zone Whampoa District Guangzhou road 510000, Guangdong province No. 6 Room 101 Patentee after: GUANGZHOU ZSROBOT INTELLIGENT EQUIPMENT Co.,Ltd. Address before: 510006 Panyu District, Guangzhou, Guangzhou University,, West Ring Road, No. 100 Patentee before: GUANGDONG University OF TECHNOLOGY |
|
| TR01 | Transfer of patent right |