CN110473197A - Material surface defect detection method, device, equipment and storage medium - Google Patents
Material surface defect detection method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN110473197A CN110473197A CN201910747956.0A CN201910747956A CN110473197A CN 110473197 A CN110473197 A CN 110473197A CN 201910747956 A CN201910747956 A CN 201910747956A CN 110473197 A CN110473197 A CN 110473197A
- Authority
- CN
- China
- Prior art keywords
- information
- image
- material surface
- detected
- characteristic pattern
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
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
-
- 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
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- 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
- G01N2021/8854—Grading and classifying of flaws
-
- 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
- G01N2021/8883—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 involving the calculation of gauges, generating models
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30136—Metal
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Theoretical Computer Science (AREA)
- Pathology (AREA)
- Data Mining & Analysis (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Immunology (AREA)
- Health & Medical Sciences (AREA)
- Analytical Chemistry (AREA)
- Signal Processing (AREA)
- Quality & Reliability (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The application provides a kind of material surface defect detection method, device, equipment and storage medium, is related to measuring for materials field.The image information that the application passes through acquisition material surface to be detected, it calculates separately image information and has marked the similarity information of each image in image collection, wherein, having marked image collection includes: multiple to have marked image, the defect information that image includes mark is each marked, if image information and the similarity information for having marked each image in image collection do not meet preset condition, then using feature identification model and default disaggregated model analysis processing image information, obtain the corresponding defect information of material surface to be detected, to realize the defect information progress secondary detection to material surface to be detected, detection efficiency can not only be promoted, and it can guarantee detection accuracy, reduce the generation of missing inspection event.
Description
Technical field
This application involves measuring for materials fields, in particular to a kind of material surface defect detection method, dress
It sets, equipment and storage medium.
Background technique
Metal material is widely used in the multiple fields such as industry, agricultural, traffic and daily life, so in social production
It usually needs to manufacture a large amount of metal material, to satisfy the use demand.And metal material actually manufacture after the completion of, surface is usually
It will appear wiping flower, dirty point, touch the defects of recessed, the subsequent use of metal material is influenced whether, for example, these defects may be accelerated
The corrosion of metal material reduces the service life of metal material.Therefore, after the completion of metal material manufactures, it is also necessary to metal
The surface defect of material is detected, to guarantee that metal material meets the requirements.
Currently, the major way detected for the surface defect of metal material are as follows: by manually to metal material
Surface is identified, the metal material of surface existing defects is sorted.
But when being detected using the detection mode of above-mentioned existing metal material surface defect to metal material, inspection
It surveys efficiency and accuracy rate is lower.
Summary of the invention
The purpose of the application is, provides a kind of material surface defect detection method, device, equipment and storage medium, uses
When the detection mode for solving existing metal material surface defect detects metal material, detection efficiency and accuracy rate compared with
Low problem.
To achieve the above object, technical solution used by the embodiment of the present application is as follows:
In a first aspect, the embodiment of the present application provides a kind of material surface defect detection method, this method comprises:
Obtain the image information of material surface to be detected；
It calculates separately image information and has marked the similarity information of each image in image collection, marked image collection packet
It includes: it is multiple to have marked image, each mark the defect information that image includes mark；
If image information and the similarity information for having marked each image in image collection do not meet preset condition, use
Feature identification model and default disaggregated model analysis processing image information, obtain the corresponding defect information of material surface to be detected；
Wherein, feature identification model is for identifying image information to obtain the attributive character of material surface to be detected；Default disaggregated model
Classify for the attributive character to material surface to be detected, and determines defect information according to classification results.
Optionally, feature identification model includes: fast convolution Area generation network and feature pyramid network；Above-mentioned use
Feature identification model and default disaggregated model analysis processing image information, obtain the corresponding defect information of material surface to be detected,
Include:
The characteristic information of image information is extracted using fast convolution Area generation network；
The corresponding N number of various sizes of characteristic pattern of characteristic information is generated using feature pyramid network, N is whole greater than 1
Number；
Using fast convolution Area generation network, N number of various sizes of characteristic pattern and N number of different default anchor frame ruler
It is very little, generate the characteristic pattern of Candidate Submission；
According to the characteristic pattern of Candidate Submission and default disaggregated model, the corresponding defect letter of material surface to be detected is obtained
Breath.
Optionally, above-mentioned using fast convolution Area generation network, N number of various sizes of characteristic pattern and N number of different
Default anchor frame size, generates the characteristic pattern of Candidate Submission, comprising:
It is N number of characteristic pattern according to sequence from small to large, successively assigns N number of default anchor frame size from big to small, obtain N
A characteristic pattern for assigning default anchor frame size；
Judge whether each characteristic pattern for assigning default anchor frame size meets proposal item by fast convolution Area generation network
Part obtains the characteristic pattern of Candidate Submission if satisfied, then generating and assessing.
Optionally, above-mentioned characteristic pattern and default disaggregated model according to Candidate Submission, obtains material surface pair to be detected
The defect information answered, comprising:
The size of the characteristic pattern of Candidate Submission is changed into default fixed dimension；
According to the characteristic pattern and default disaggregated model for presetting fixed-size Candidate Submission, material surface to be detected is obtained
Corresponding defect information.
Optionally, above-mentioned using feature identification model and default disaggregated model analysis processing image information, it obtains to be detected
Before the corresponding defect information of material surface, this method further include:
Background separation processing before being carried out by exposure mask convolution Area generation network to image information, before obtaining image information
Scape image information；
Correspondingly, using feature identification model and default disaggregated model analysis processing image information, material to be detected is obtained
The corresponding defect information in surface, comprising:
Using feature identification model and default disaggregated model analysis processing foreground image information, material surface to be detected is obtained
Corresponding defect information.
It is optionally, above-mentioned to calculate separately image information and marked in image collection after the similarity information of each image,
This method further include:
If image information with marked in image collection the phase for existing in the similarity information of each image and meeting preset condition
Like degree information, then obtains and meet target corresponding to the similarity information of preset condition and marked image；
Determine that the corresponding defect information of material surface to be detected is identical as the defect information that target has marked image.
Second aspect, the embodiment of the present application provide a kind of System of Detecting Surface Defects For Material detection device, comprising:
Module is obtained, for obtaining the image information of material surface to be detected；
Computing module, the similarity information for calculating separately image information Yu having marked each image in image collection,
Mark image collection includes: multiple defect informations for having marked image, each having marked that image includes mark；
Processing module, if not meeting for image information with the similarity information for having marked each image in image collection pre-
If condition, then using feature identification model and default disaggregated model analysis processing image information, material surface pair to be detected is obtained
The defect information answered；Wherein, feature identification model is for identifying image information to obtain the attributive character of material surface to be detected；
Default disaggregated model determines that defect is believed for classifying to the attributive character of material surface to be detected, and according to classification results
Breath.
Optionally, feature identification model includes: fast convolution Area generation network and feature pyramid network；
Processing module, specifically for extracting the characteristic information of image information using fast convolution Area generation network；Using
Feature pyramid network generates the corresponding N number of various sizes of characteristic pattern of characteristic information, and N is the integer greater than 1；For using fast
Fast convolution Area generation network, N number of various sizes of characteristic pattern and N number of different default anchor frame size generate Candidate Submission
Characteristic pattern；According to the characteristic pattern of Candidate Submission and default disaggregated model, the corresponding defect letter of material surface to be detected is obtained
Breath.
Optionally, processing module successively assigns N number of from big specifically for being N number of characteristic pattern according to sequence from small to large
To small default anchor frame size, N number of characteristic pattern for assigning default anchor frame size is obtained；Sentenced by fast convolution Area generation network
Whether each characteristic pattern for assigning default anchor frame size that breaks meets proposal condition, if satisfied, then generation and assessment obtain Candidate Submission
Characteristic pattern.
Optionally, processing module, specifically for the size of the characteristic pattern of Candidate Submission is changed into default fixed dimension；Root
According to the characteristic pattern and default disaggregated model for presetting fixed-size Candidate Submission, the corresponding defect of material surface to be detected is obtained
Information.
Optionally, the device further include: preceding background separation module, in processing module using feature identification model and pre-
If disaggregated model analysis processing image information passes through exposure mask convolution before obtaining the corresponding defect information of material surface to be detected
Background separation processing, obtains the foreground image information of image information before Area generation network carries out image information；
Correspondingly, processing module is used for using feature identification model and default disaggregated model analysis processing foreground image letter
Breath, obtains the corresponding defect information of material surface to be detected.
Optionally, processing module, if being also used to image information and having marked the similarity information of each image in image collection
Middle to there is the similarity information for meeting preset condition, then acquisition meets target corresponding to the similarity information of preset condition and has marked
Infuse image；Determine that the corresponding defect information of material surface to be detected is identical as the defect information that target has marked image.
The third aspect, the embodiment of the present application provide a kind of detection device, including memory and processor, store in memory
There is the computer program that can be run in processor, processor realizes material table as described in relation to the first aspect when executing computer program
Planar defect detection method.
Fourth aspect, the embodiment of the present application provide a kind of computer readable storage medium, are stored thereon with computer program,
Material surface defect detection method as described in relation to the first aspect is realized when computer program is executed by processor.
The beneficial effect of the application is:
Material surface defect detection method, device provided by the embodiment of the present application, equipment and storage medium, can be by obtaining
The image information for taking material surface to be detected calculates separately image information and believes with the similarity for having marked each image in image collection
Breath, wherein having marked image collection includes: multiple defect informations for having marked image, each having marked that image includes mark, if
Image information and the similarity information for having marked each image in image collection do not meet preset condition, then identify mould using feature
Type and default disaggregated model analysis processing image information, obtain the corresponding defect information of material surface to be detected, to realize
Secondary detection is carried out to the defect information of material surface to be detected, detection efficiency can not only be promoted, and can guarantee to detect
Precision reduces the generation of missing inspection event.
Detailed description of the invention
Technical solution in ord to more clearly illustrate embodiments of the present application, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only some embodiments of the application, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 shows the flow diagram of material surface defect detection method provided by the embodiments of the present application；
Fig. 2 shows another flow diagrams of material surface defect detection method provided by the embodiments of the present application；
Fig. 3 shows the another flow diagram of material surface defect detection method provided by the embodiments of the present application；
Fig. 4 shows the another flow diagram of material surface defect detection method provided by the embodiments of the present application；
Fig. 5 shows the another flow diagram of material surface defect detection method provided by the embodiments of the present application；
Fig. 6 shows the another flow diagram of material surface defect detection method provided by the embodiments of the present application；
Fig. 7 shows the structural schematic diagram of System of Detecting Surface Defects For Material detection device provided by the embodiments of the present application；
Fig. 8 shows another structural schematic diagram of System of Detecting Surface Defects For Material detection device provided by the embodiments of the present application；
Fig. 9 shows the structural schematic diagram of detection device provided by the embodiments of the present application.
Specific embodiment
To keep the purposes, technical schemes and advantages of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application
In attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is
Some embodiments of the present application, instead of all the embodiments.The application being usually described and illustrated herein in the accompanying drawings is implemented
The component of example can be arranged and be designed with a variety of different configurations.
Therefore, the detailed description of the embodiments herein provided in the accompanying drawings is not intended to limit below claimed
Scope of the present application, but be merely representative of the selected embodiment of the application.Based on the embodiment in the application, this field is common
Technical staff's every other embodiment obtained without creative efforts belongs to the model of the application protection
It encloses.
It should also be noted that similar label and letter indicate similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined in a attached drawing, does not then need that it is further defined and explained in subsequent attached drawing.Furthermore, it is necessary to explanation
It is that term " first ", " second ", " third " etc. are only used for distinguishing description, is not understood to indicate or imply relative importance.
The embodiment of the present application provides a kind of material surface defect detection method, for detecting the defect information of material surface,
Realization identifies the material of existing defects.This method can be executed by detection device, the detection device can for server,
Any equipment or detecting instrument with processing function such as computer, the application to this with no restriction.
Fig. 1 shows the flow diagram of material surface defect detection method provided by the embodiments of the present application.Such as Fig. 1 institute
Show, the material surface defect detection method, comprising:
S101, the image information for obtaining material surface to be detected.
Wherein, material to be detected can be the metal materials such as aluminium, copper, be also possible to glass, paper or composite material, and or
Person may also mean that some electronic components with appearance requirement, timber products etc., for the concrete type of material to be detected,
The application is with no restriction.
Image information can be the picture shot by video camera to material surface to be detected, be also possible to by taking the photograph
After camera records a video to material to be detected, the image of the related material surface to be detected intercepted from video recording be can also be
Acquired image is scanned to material to be detected by other scanners, radar equipment etc..
For using metal product as material to be detected, metal product can pass through institute on production line after the completion of production
The camera of installation shoots metal product, obtains the photo of metal product surface, then the System of Detecting Surface Defects For Material detection side
Method for metal product when being detected, it can obtains the photo of metal product taken by camera as its surface
Image information.
S102, the similarity information for calculating separately image information and having marked each image in image collection, have marked image
Set includes: multiple defect informations for having marked image, each having marked that image includes mark.
Specifically, defect information may include: the information such as defect type and/or the defective locations of material surface to be detected.
Wherein, defect type may include at least one defect class such as spot, pit, scratch, color difference, defect of material surface to be detected
Type, defective locations can be every kind of defect type in the specific location of material surface to be detected, such as edge or center.
Optionally, before executing the material surface defect detection method, a large amount of known defect information can first be obtained
The image of material surface is detected, and these images are labeled according to known defect information, that is, mark should in each image
The defect information that material is detected corresponding to image, obtains by multiple mark image collections for having marked image and having been constituted.Example
Such as, the higher defect type of frequency that detection material occurs can be labeled in multiple images, to obtain being labeled with scarce
Multiple marked for falling into information has marked image collection composed by image.In the actual implementation process, this has marked image set
Conjunction, which can be, defaults in server or Computer Data library, can also acquire from other equipment or server, can be with
For the information for receiving user's input, the application is equally with no restriction.
Further, after the image information for getting above-mentioned material to be detected, can by acquired image information with
Each image of mark image collection compares, calculate separately image information to marked the similar of each image in image collection
Spend information.For example, image information can be calculated by modes such as Histogram Matching algorithm, matrix decomposition algorithm or comparative feature points
With marked the similarity between image, so as to obtain similarity information.
Optionally, in the application some embodiments, image information can also be extracted respectively by gray level co-occurrence matrixes
Textural characteristics and the textural characteristics for having marked each image in image collection, and calculate separately the textural characteristics of image information with
Similarity between the textural characteristics of image is marked, respectively so as to by the textural characteristics of image information and mark image
Similarity between textural characteristics as image information and has marked the similarity information between image.Wherein, texture is become
Change slow image, the numerical value on gray level co-occurrence matrixes diagonal line is larger；And image faster for texture variations, gray scale
Numerical value on co-occurrence matrix diagonal line is smaller, and the value of diagonal line two sides is larger, therefore, gray level co-occurrence matrixes can accurately extract to
Detect the textural characteristics of material surface.
It should also be noted that, in the scheme of the application, it can also be to special by the extracted texture of gray level co-occurrence matrixes
Sign is calculated, its corresponding statistical characteristic value is obtained, and is then based on statistical characteristic value and is calculated image information and each mark figure
The big similarity as between, so that the calculation amount of similarity is reduced, to improve defects detection efficiency.
Above-mentioned image has been marked in addition, can also directly prestore in the application other embodiments, in aforementioned data library
The corresponding textural characteristics of each image in set, that is, can lead in advance before material surface defect detection method execution
It crosses gray level co-occurrence matrixes and extracts and respectively marked the textural characteristics of image, and its pre- is stored in database.To in the material table
In planar defect detection method implementation procedure, it is thus only necessary to the textural characteristics of image information are obtained, it can be by the line of the image information
The textural characteristics for respectively having marked image prestored in reason feature and database compare, and similarity are calculated, to improve material
The detection efficiency of surface defect.
If S103, image information and the similarity information for having marked each image in image collection do not meet preset condition,
Then using feature identification model and default disaggregated model analysis processing image information, the corresponding defect of material surface to be detected is obtained
Information.
Wherein, preset condition can be whether similarity is greater than or equal to preset threshold, for example, preset threshold can be
90%, 95%, 99% etc..For it is any marked image for, if this has marked the similarity between image and image information
Reach aforementioned preset threshold, then can determine that image information and the similarity information for having marked image meet preset condition, it is no
Then, it is determined that image information and the similarity information for having marked image do not meet preset condition.
Optionally, if image information does not meet default item with the similarity information for having marked each image in image collection
Part can then be determined and marked there is no that matched can mark image with the image information in image, that is, can not pass through
Image collection has been marked to identify to obtain the defect information of material to be detected.At this point it is possible to further use feature identification model
The image information is analyzed and processed with default disaggregated model, to obtain the corresponding defect information of material surface to be detected.Its
In, feature identification model can preset classification mould by identifying to obtain the attributive character of material surface to be detected to the image information
Type can classify to the attributive character of material surface to be detected, and determine its specific defect information according to classification results.
From the above mentioned, the embodiment of the present application calculates separately image letter by the image information of acquisition material surface to be detected
Breath and the similarity information for having marked each image in image collection, wherein having marked image collection includes: multiple mark figures
Picture has each marked the defect information that image includes mark, if image information to marked the similar of each image in image collection
Degree information does not meet preset condition, then using feature identification model and default disaggregated model analysis processing image information, obtains
The corresponding defect information of material surface to be detected carries out second level inspection to the defect information of material surface to be detected to realize
It surveys, detection efficiency can not only be promoted, and can guarantee detection accuracy, reduce the generation of missing inspection event.
Fig. 2 shows another flow diagrams of material surface defect detection method provided by the embodiments of the present application.
Optionally, in the embodiment of the present application, feature identification model may include: fast convolution Area generation network
(Faster Regional Convolutional Neural Networks, Faster RCNN) and feature pyramid network
(Feature Pyramid Networks, FPN).As shown in Fig. 2, above-mentioned using feature identification model and default disaggregated model point
Analysis processing image information, obtains the corresponding defect information of material surface to be detected, may include:
S201, the characteristic information that image information is extracted using fast convolution Area generation network.
Specifically, above-mentioned acquired image information is inputted into this feature identification model, feature identification model can be passed through
In Faster RCNN extract the characteristic information of the image information, for example, characteristic information can be used for characterizing the image information
Some Multidimensional numericals (or tensor) of attribute.
S202, the corresponding N number of various sizes of characteristic pattern of characteristic information is generated using feature pyramid network, N is greater than 1
Integer.
Further, the FPN structure in this feature identification model, can be by characteristic information extracted in above-mentioned steps S201
It is mapped as N number of characteristic pattern, and the size of N number of characteristic pattern is different, institute's table in preceding feature information is contained in N number of characteristic pattern
The attribute information of the image information of sign.
S203, using fast convolution Area generation network, N number of various sizes of characteristic pattern and N number of different default anchor
Frame size generates the characteristic pattern of Candidate Submission.
Wherein, the characteristic pattern of Candidate Submission can be used for characterizing the attributive character of material surface to be detected.
Optionally, before executing the material surface defect detection method, a large amount of known defect information can also first be obtained
Detection material surface image, then by K cluster cluster (K-means Clustering) method, in these images
Defect information carry out cluster calculation and to obtain multiple anchor frame sizes relevant to such detection material preset anchor as this
Frame size.Wherein, the quantity for presetting anchor frame size can be identical as features described above figure quantity, and N number of default anchor frame size is big
It is small to be all different.
To which feature identification model can use fast convolution Area generation network, according to above-mentioned N number of various sizes of spy
Sign figure and N number of different default anchor frame size, generate the characteristic pattern of Candidate Submission.
It is corresponding scarce to obtain material surface to be detected by S204, characteristic pattern and default disaggregated model according to Candidate Submission
Fall into information.
As described above, the characteristic pattern of Candidate Submission can be inputted default after the characteristic pattern for obtaining above-mentioned Candidate Submission
Disaggregated model, the attribute letter of default disaggregated model material surface to be detected according to included in the characteristic pattern of Candidate Submission
Breath, classifies to the attribute information of material surface to be detected, so as to determine material surface to be detected according to classification results
Defect information.
Fig. 3 shows the another flow diagram of material surface defect detection method provided by the embodiments of the present application.
Optionally, as shown in figure 3, it is above-mentioned use fast convolution Area generation network, N number of various sizes of characteristic pattern, with
And N number of different default anchor frame size, generate the characteristic pattern of Candidate Submission, comprising:
S301, according to sequence from small to large be N number of characteristic pattern, successively assign N number of default anchor frame size from big to small,
Obtain N number of characteristic pattern for assigning default anchor frame size.
As described above, N number of characteristic pattern has different size of size, the size of N number of default anchor frame size is also different.
It is alternatively possible to which N number of characteristic pattern is arranged according to the sequence of size from small to large, then, N number of default anchor frame size is pressed
It is respectively applied in the N number of characteristic pattern arranged from small to large according to size according to sequence from big to small, in order to be able to more accurate
The tiny flaw of material surface to be detected is detected on ground.
For example, characteristic pattern A, B, C if it exists, and its size relationship is A > B > C, there are 3 default anchor frame size c, d,
E, and its size relationship is c > d > e, then default anchor frame size c can be applied to characteristic pattern C, default anchor frame size d is answered
For characteristic pattern B, default anchor frame size e is applied to characteristic pattern A.
S302, judge whether each characteristic pattern for assigning default anchor frame size meets by fast convolution Area generation network and mention
View condition obtains the characteristic pattern of Candidate Submission if satisfied, then generating and assessing.
Specifically, Area generation network (the Region Proposal in fast convolution Area generation network can be passed through
Network, RPN) in N number of generation and assessment for assigning and being proposed on the characteristic pattern for presetting anchor frame size, according to default anchor frame
Size chooses characteristic pattern of the characteristic pattern as Candidate Submission for meeting proposal condition from N number of characteristic pattern.
Fig. 4 shows the another flow diagram of material surface defect detection method provided by the embodiments of the present application.
Optionally, as shown in figure 4, above-mentioned characteristic pattern and default disaggregated model according to Candidate Submission, obtains to be detected
The corresponding defect information of material surface, comprising:
S401, the size of the characteristic pattern of Candidate Submission is changed into default fixed dimension.
It optionally, can be first by the spy of Candidate Submission before the characteristic pattern of Candidate Submission is inputted default disaggregated model
Sign figure is all changed to default fixed dimension, that is, the characteristic pattern of the Candidate Submission inputted in default disaggregated model all has identical
The size of size.The default fixed dimension can be default disaggregated model in the training process used in characteristic pattern size,
So as to higher efficiency the characteristic pattern of Candidate Submission is identified.
S402, basis preset the characteristic pattern and default disaggregated model of fixed-size Candidate Submission, obtain to be checked measures and monitor the growth of standing timber
Expect the corresponding defect information in surface.
Optionally, above-mentioned using feature identification model and default disaggregated model analysis processing image information, it obtains to be detected
Before the corresponding defect information of material surface, this method further include:
Background separation processing before being carried out by exposure mask convolution Area generation network to image information, before obtaining image information
Scape image information.
Correspondingly, using feature identification model and default disaggregated model analysis processing image information, material to be detected is obtained
The corresponding defect information in surface, comprising:
Using feature identification model and default disaggregated model analysis processing foreground image information, material surface to be detected is obtained
Corresponding defect information.
It specifically, often include other nothings by the image information of material to be detected accessed by abovementioned steps S101
The background information of pass, for example, usually containing in the picture after shooting when camera shoots metal product from product line
Other background informations such as conveyer belt, the instrument and equipment on piece line are produced, and pass through exposure mask convolution Area generation network (Mask
Regional Convolutional Neural Networks, Mask RCNN) can to acquired image information carry out before
Background separation processing, obtains the foreground image information of image information namely the image information of material region to be detected.In the past
For stating metal product, then the extraneous backgrounds information such as conveyer belt, instrument and equipment can be filtered out, metal product location is only left
The foreground image information in domain, to further improve detection accuracy and detection efficiency.
Fig. 5 shows the another flow diagram of material surface defect detection method provided by the embodiments of the present application.
Optionally, as shown in figure 5, the above-mentioned similarity for calculating separately image information and having marked each image in image collection
After information, this method further include:
If S501, image information and having marked in image collection to exist in the similarity information of each image and meeting preset condition
Similarity information, then obtain and meet target corresponding to the similarity information of preset condition and marked image.
As described above, preset condition can refer to whether similarity is greater than or equal to preset threshold, marked for any
For image, if this, which has marked the similarity between image and image information, reaches aforementioned preset threshold, image can be determined
Information and the similarity information for having marked image meet preset condition.Therefore, exist and image when having marked in image collection
When similarity between information is greater than the image of mark of preset threshold, the image conduct can be obtained in image collection from having marked
Target has marked image.
S502, determine that the corresponding defect information of material surface to be detected is identical as the defect information that target has marked image.
It due to having marked image collection is marked by the image of the detection material surface to a large amount of known defect information
Note gained, wherein it is known for each having marked the corresponding defect information for detecting material of image to be.So due to target
The similarity marked between image and the image information of material to be detected is greater than preset threshold, it is believed that material table to be detected
The corresponding defect information in face is identical or very close as the defect information that target has marked image, so target can have been marked
The defect information of image is determined as the defect information of material surface to be detected.
The embodiment of the present application also provides a kind of material surface defect detection method.Fig. 6 shows the embodiment of the present application offer
The another flow diagram of material surface defect detection method can be with as shown in fig. 6, the material surface defect detection method
Include:
S610, the image information for obtaining material surface to be detected.
S620, the similarity information for calculating separately image information and having marked each image in image collection.
Wherein, having marked image collection includes: multiple defect letters for having marked image, each having marked that image includes mark
Breath.
S630, judge image information and marked pre- with the presence or absence of meeting in the similarity information of each image in image collection
If the similarity information of condition.
The similarity information for meeting preset condition if it exists thens follow the steps S641 and step S642；Meet if it does not exist
The similarity information of preset condition then successively executes step S651 to step S657.
S641, acquisition meet target corresponding to the similarity information of preset condition and have marked image.
S642, determine that the corresponding defect information of material surface to be detected is identical as the defect information that target has marked image.
S651, preceding background separation processing is carried out to image information by exposure mask convolution Area generation network, obtains image letter
The foreground image information of breath.
S652, the characteristic information that foreground image information is extracted using fast convolution Area generation network.
S653, the corresponding N number of various sizes of characteristic pattern of characteristic information is generated using feature pyramid network, N is greater than 1
Integer.
S654, according to sequence from small to large be N number of characteristic pattern, successively assign N number of default anchor frame size from big to small,
Obtain N number of characteristic pattern for assigning default anchor frame size.
S655, judge whether each characteristic pattern for assigning default anchor frame size meets by fast convolution Area generation network and mention
View condition obtains the characteristic pattern of Candidate Submission if satisfied, then generating and assessing.
S656, the size of the characteristic pattern of Candidate Submission is changed into default fixed dimension；
S657, basis preset the characteristic pattern and default disaggregated model of fixed-size Candidate Submission, obtain to be checked measures and monitor the growth of standing timber
Expect the corresponding defect information in surface.
Method beneficial effect having the same described in the material surface defect detection method and previous embodiment, herein not
It repeats again.
The embodiment of the present application also provides a kind of System of Detecting Surface Defects For Material detection device, and the device is for executing in previous embodiment
The material surface defect detection method.Fig. 7 shows System of Detecting Surface Defects For Material detection device provided by the embodiments of the present application
Structural schematic diagram, as shown in fig. 7, the System of Detecting Surface Defects For Material detection device, comprising:
Obtain module 11, computing module 12 and processing module 13.Module 11 is obtained, for obtaining material surface to be detected
Image information；Computing module 12, the similarity information for calculating separately image information Yu having marked each image in image collection,
Having marked image collection includes: multiple defect informations for having marked image, each having marked that image includes mark；Processing module
13, if not meeting preset condition with the similarity information for having marked each image in image collection for image information, use
Feature identification model and default disaggregated model analysis processing image information, obtain the corresponding defect information of material surface to be detected；
Wherein, feature identification model is for identifying image information to obtain the attributive character of material surface to be detected；Default disaggregated model
Classify for the attributive character to material surface to be detected, and determines defect information according to classification results.
Optionally, feature identification model includes: fast convolution Area generation network and feature pyramid network；Processing module
13, specifically for extracting the characteristic information of image information using fast convolution Area generation network；Using feature pyramid network
The corresponding N number of various sizes of characteristic pattern of characteristic information is generated, N is the integer greater than 1；For using fast convolution Area generation
Network, N number of various sizes of characteristic pattern and N number of different default anchor frame size, generate the characteristic pattern of Candidate Submission, candidate
The characteristic pattern of proposal can be used for characterizing the attributive character of material surface to be detected；According to the characteristic pattern of Candidate Submission and default point
Class model obtains the corresponding defect information of material surface to be detected.
Optionally, processing module 13, specifically for according to sequence from small to large be N number of characteristic pattern, successively assign it is N number of from
Small default anchor frame size is arrived greatly, obtains N number of characteristic pattern for assigning default anchor frame size；Pass through fast convolution Area generation network
Judge whether each characteristic pattern for assigning default anchor frame size meets proposal condition, if satisfied, then generation and assessment obtain candidate and mention
The characteristic pattern of view.
Optionally, processing module 13, specifically for the size of the characteristic pattern of Candidate Submission is changed into default fixed dimension；
According to the characteristic pattern and default disaggregated model for presetting fixed-size Candidate Submission, it is corresponding scarce to obtain material surface to be detected
Fall into information.
Fig. 8 has gone out another schematic diagram of System of Detecting Surface Defects For Material detection device provided by the embodiments of the present application.Optionally, as schemed
Shown in 8, the System of Detecting Surface Defects For Material detection device further include: preceding background separation module 14, for using feature in processing module 13
Identification model and default disaggregated model analysis processing image information, before obtaining the corresponding defect information of material surface to be detected,
Background separation processing before being carried out by exposure mask convolution Area generation network to image information obtains the foreground image letter of image information
Breath.
Correspondingly, processing module 13 is used for using feature identification model and default disaggregated model analysis processing foreground image letter
Breath, obtains the corresponding defect information of material surface to be detected.
Optionally, processing module 13 are believed if being also used to image information with the similarity for having marked each image in image collection
There is the similarity information for meeting preset condition in breath, then obtains and met target corresponding to the similarity information of preset condition
Mark image；Determine that the corresponding defect information of material surface to be detected is identical as the defect information that target has marked image.
Due to System of Detecting Surface Defects For Material detection device provided by the embodiments of the present application, for executing institute in preceding method embodiment
Therefore the material surface defect detection method stated has preceding method whole beneficial effects as described in the examples, the application exists
This is repeated no more.
The embodiment of the present application also provides a kind of detection device, for executing preceding method material surface as described in the examples
Defect method.Fig. 9 shows the structural schematic diagram of detection device provided by the embodiments of the present application, as shown in figure 9, the detection device
Including memory 1002 and processor 1001, the computer program that can be run in processor 1001 is stored in memory 1002,
Processor 1001 realizes such as preceding method material surface defect detection method as described in the examples when executing computer program.Tool
Body implementation is similar with technical effect, and which is not described herein again.
Optionally, the embodiment of the present application also provides a kind of computer readable storage medium, is stored thereon with computer program,
Such as preceding method material surface defect detection method as described in the examples is realized when computer program is executed by processor.
The foregoing is merely preferred embodiment of the present application, are not intended to limit this application, for the skill of this field
For art personnel, various changes and changes are possible in this application.Within the spirit and principles of this application, made any to repair
Change, equivalent replacement, improvement etc., should be included within the scope of protection of this application.
Claims (14)
1. a kind of material surface defect detection method, which is characterized in that the described method includes:
Obtain the image information of material surface to be detected；
It calculates separately described image information and has marked the similarity information of each image in image collection, it is described to have marked image set
Conjunction include: it is multiple marked image, it is each described to have marked the defect information that image includes mark；
If described image information and the similarity information for having marked each image in image collection do not meet preset condition, use
Feature identification model and default disaggregated model analysis processing described image information, it is corresponding scarce to obtain the material surface to be detected
Fall into information；Wherein, the feature identification model is for identifying described image information to obtain the category of the material surface to be detected
Property feature；The default disaggregated model is used to classify to the attributive character of the material surface to be detected, and according to classification
As a result the defect information is determined.
2. the method according to claim 1, wherein the feature identification model includes: that fast convolution region is raw
At network and feature pyramid network；
It is described that described image information is handled using feature identification model and the analysis of default disaggregated model, obtain the material to be detected
The corresponding defect information in surface, comprising:
The characteristic information of described image information is extracted using the fast convolution Area generation network；
The corresponding N number of various sizes of characteristic pattern of the characteristic information is generated using the feature pyramid network, N is greater than 1
Integer；
Using the fast convolution Area generation network, N number of various sizes of characteristic pattern and N number of different default anchor
Frame size, generates the characteristic pattern of Candidate Submission, and the characteristic pattern of the Candidate Submission is used to characterize the attribute of material surface to be detected
Feature；
According to the characteristic pattern of the Candidate Submission and the default disaggregated model, it is corresponding to obtain the material surface to be detected
Defect information.
3. according to the method described in claim 2, it is characterized in that, described using the fast convolution Area generation network, N number of
The various sizes of characteristic pattern and N number of different default anchor frame size, generate the characteristic pattern of Candidate Submission, comprising:
It is N number of characteristic pattern according to sequence from small to large, successively assigns N number of default anchor frame size from big to small,
Obtain N number of characteristic pattern for assigning default anchor frame size；
Judge whether each characteristic pattern for assigning default anchor frame size meets by the fast convolution Area generation network to mention
View condition obtains the characteristic pattern of Candidate Submission if satisfied, then generating and assessing.
4. according to the method described in claim 3, it is characterized in that, according to the characteristic pattern of the Candidate Submission and described default
Disaggregated model obtains the corresponding defect information of the material surface to be detected, comprising:
The size of the characteristic pattern of the Candidate Submission is changed into default fixed dimension；
According to the characteristic pattern and the default disaggregated model for presetting the fixed-size Candidate Submission, obtain described to be detected
The corresponding defect information of material surface.
5. method according to claim 1-4, which is characterized in that described using feature identification model and default point
Class model analysis processing described image information, before obtaining the corresponding defect information of the material surface to be detected, the method
Further include:
Background separation processing, obtains described image information before being carried out by exposure mask convolution Area generation network to described image information
Foreground image information；
Correspondingly, described using feature identification model and default disaggregated model analysis processing described image information, obtain it is described to
Detect the corresponding defect information of material surface, comprising:
The foreground image information is handled using feature identification model and the analysis of default disaggregated model, obtains the material to be detected
The corresponding defect information in surface.
6. method according to claim 1-4, which is characterized in that it is described calculate separately described image information with
It marks in image collection after the similarity information of each image, further includes:
If described image information with marked in image collection the phase for existing in the similarity information of each image and meeting preset condition
Like degree information, then obtains target corresponding to the similarity information for meeting preset condition and marked image；
Determine that the corresponding defect information of the material surface to be detected is identical as the defect information that the target has marked image.
7. a kind of System of Detecting Surface Defects For Material detection device, which is characterized in that described device includes:
Module is obtained, for obtaining the image information of material surface to be detected；
Computing module, the similarity information for calculating separately described image information Yu having marked each image in image collection, institute
State marked image collection include: it is multiple marked image, it is each described to have marked the defect information that image includes mark；
Processing module, if not meeting for described image information with the similarity information for having marked each image in image collection pre-
If condition, then using feature identification model and default disaggregated model analysis processing described image information, described to be checked measure and monitor the growth of standing timber is obtained
Expect the corresponding defect information in surface；Wherein, the feature identification model is described to be checked for identifying to obtain to described image information
Survey the attributive character of material surface；The default disaggregated model is for dividing the attributive character of the material surface to be detected
Class, and the defect information is determined according to classification results.
8. device according to claim 7, which is characterized in that the feature identification model includes: that fast convolution region is raw
At network and feature pyramid network；
The processing module, the feature specifically for being extracted described image information using the fast convolution Area generation network are believed
Breath；The corresponding N number of various sizes of characteristic pattern of the characteristic information is generated using the feature pyramid network, N is greater than 1
Integer；Using the fast convolution Area generation network, N number of various sizes of characteristic pattern and N number of different default anchor
Frame size, generates the characteristic pattern of Candidate Submission, and the characteristic pattern of the Candidate Submission is used to characterize the attribute of material surface to be detected
Feature；According to the characteristic pattern of the Candidate Submission and the default disaggregated model, it is corresponding to obtain the material surface to be detected
Defect information.
9. device according to claim 8, which is characterized in that the processing module, specifically for according to from small to large
Sequence is N number of characteristic pattern, successively assigns N number of default anchor frame size from big to small, obtains N number of assign and presets anchor frame
The characteristic pattern of size；Judge that each characteristic pattern for assigning default anchor frame size is by the fast convolution Area generation network
It is no to meet proposal condition, the characteristic pattern of Candidate Submission is obtained if satisfied, then generating and assessing.
10. device according to claim 9, which is characterized in that the processing module is specifically used for the Candidate Submission
The size of characteristic pattern change into default fixed dimension；According to the characteristic pattern and institute for presetting the fixed-size Candidate Submission
Default disaggregated model is stated, the corresponding defect information of the material surface to be detected is obtained.
11. according to the described in any item devices of claim 7-10, which is characterized in that described device further include: preceding background separation
Module；
The preceding background separation module, at the processing module is using feature identification model and default disaggregated model analysis
Described image information is managed, before obtaining the corresponding defect information of the material surface to be detected, passes through exposure mask convolution Area generation
Background separation processing, obtains the foreground image information of described image information before network carries out described image information；
Correspondingly, the processing module is used to handle the foreground image using feature identification model and the analysis of default disaggregated model
Information obtains the corresponding defect information of the material surface to be detected.
12. according to the described in any item devices of claim 7-10, which is characterized in that the processing module, if being also used to described
Image information with marked in image collection the similarity information for existing in the similarity information of each image and meeting preset condition, then
It obtains target corresponding to the similarity information for meeting preset condition and has marked image；Determine the material surface to be detected
Corresponding defect information is identical as the defect information that the target has marked image.
13. a kind of detection device, which is characterized in that including memory and processor, being stored in memory can transport in processor
Capable computer program, processor realize that material surface as claimed in any one of claims 1 to 6 lacks when executing computer program
Fall into detection method.
14. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
Material surface defect detection method as claimed in any one of claims 1 to 6 is realized when being executed by processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910747956.0A CN110473197A (en) | 2019-08-13 | 2019-08-13 | Material surface defect detection method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910747956.0A CN110473197A (en) | 2019-08-13 | 2019-08-13 | Material surface defect detection method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110473197A true CN110473197A (en) | 2019-11-19 |
Family
ID=68511691
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910747956.0A Pending CN110473197A (en) | 2019-08-13 | 2019-08-13 | Material surface defect detection method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110473197A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111353983A (en) * | 2020-02-28 | 2020-06-30 | 腾讯科技（深圳）有限公司 | Defect detection and identification method and device, computer readable medium and electronic equipment |
| CN111413350A (en) * | 2020-03-24 | 2020-07-14 | 江苏斯德雷特通光光纤有限公司 | Method and device for detecting defects of optical fiber flat cable |
| CN111598084A (en) * | 2020-05-11 | 2020-08-28 | 北京阿丘机器人科技有限公司 | Defect segmentation network training method, device and equipment and readable storage medium |
| CN112651941A (en) * | 2020-12-25 | 2021-04-13 | 北京巅峰科技有限公司 | Vehicle defect identification method and device, electronic device and storage medium |
| CN112734693A (en) * | 2020-12-18 | 2021-04-30 | 平安科技（深圳）有限公司 | Pipeline weld defect detection method and related device |
| CN113112442A (en) * | 2019-12-25 | 2021-07-13 | 精锐视觉智能科技（深圳）有限公司 | Defect detection method and device and terminal equipment |
| CN113469299A (en) * | 2021-09-06 | 2021-10-01 | 常州微亿智造科技有限公司 | Defect detection method and defect detection device in industrial detection |
| CN113723219A (en) * | 2021-08-10 | 2021-11-30 | 浙江大华技术股份有限公司 | Method for identifying breakage of insulating member and related device |
| CN114742837A (en) * | 2022-06-14 | 2022-07-12 | 华南农业大学 | Deep information image interaction-based wood artwork processing auxiliary instrument and method |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105241923A (en) * | 2015-10-08 | 2016-01-13 | 电子科技大学 | Flip-chip bonding welded spot defect detection method |
| US20180300865A1 (en) * | 2017-04-13 | 2018-10-18 | Instrumental, Inc. | Method for predicting defects in assembly units |
| CN109493339A (en) * | 2018-11-20 | 2019-03-19 | 北京嘉恒中自图像技术有限公司 | A kind of casting inner surface gas hole defect detection method based on endoscopic imaging |
| US20190087684A1 (en) * | 2017-09-15 | 2019-03-21 | International Business Machines Corporation | Fast joint template matching |
| CN109711474A (en) * | 2018-12-24 | 2019-05-03 | 中山大学 | A kind of aluminium material surface defects detection algorithm based on deep learning |
| CN109884070A (en) * | 2019-03-08 | 2019-06-14 | 武汉大学 | A kind of aluminium aluminium blister package tablet defect inspection method based on machine vision |
-
2019
- 2019-08-13 CN CN201910747956.0A patent/CN110473197A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105241923A (en) * | 2015-10-08 | 2016-01-13 | 电子科技大学 | Flip-chip bonding welded spot defect detection method |
| US20180300865A1 (en) * | 2017-04-13 | 2018-10-18 | Instrumental, Inc. | Method for predicting defects in assembly units |
| US20190087684A1 (en) * | 2017-09-15 | 2019-03-21 | International Business Machines Corporation | Fast joint template matching |
| CN109493339A (en) * | 2018-11-20 | 2019-03-19 | 北京嘉恒中自图像技术有限公司 | A kind of casting inner surface gas hole defect detection method based on endoscopic imaging |
| CN109711474A (en) * | 2018-12-24 | 2019-05-03 | 中山大学 | A kind of aluminium material surface defects detection algorithm based on deep learning |
| CN109884070A (en) * | 2019-03-08 | 2019-06-14 | 武汉大学 | A kind of aluminium aluminium blister package tablet defect inspection method based on machine vision |
Non-Patent Citations (3)
| Title |
|---|
| DU-MING TSAI 等: "Defect Detection in Electronic Surfaces Using Template-Based Fourier Image Reconstruction", 《IEEE》 * |
| 朱志宇 等: "《流形粒子滤波算法及其在视频目标跟踪中的应用》", 31 January 2015, 国防工业出版社 * |
| 王成明 等: "冷轧带钢表面缺陷的二级检测与快速判定方法研究", 《中原工学院》 * |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113112442A (en) * | 2019-12-25 | 2021-07-13 | 精锐视觉智能科技（深圳）有限公司 | Defect detection method and device and terminal equipment |
| CN111353983A (en) * | 2020-02-28 | 2020-06-30 | 腾讯科技（深圳）有限公司 | Defect detection and identification method and device, computer readable medium and electronic equipment |
| CN111413350A (en) * | 2020-03-24 | 2020-07-14 | 江苏斯德雷特通光光纤有限公司 | Method and device for detecting defects of optical fiber flat cable |
| CN111598084B (en) * | 2020-05-11 | 2023-06-02 | 北京阿丘机器人科技有限公司 | Defect segmentation network training method, device, equipment and readable storage medium |
| CN111598084A (en) * | 2020-05-11 | 2020-08-28 | 北京阿丘机器人科技有限公司 | Defect segmentation network training method, device and equipment and readable storage medium |
| CN112734693A (en) * | 2020-12-18 | 2021-04-30 | 平安科技（深圳）有限公司 | Pipeline weld defect detection method and related device |
| CN112734693B (en) * | 2020-12-18 | 2024-06-07 | 平安科技（深圳）有限公司 | Pipeline weld defect detection method and related device |
| CN112651941A (en) * | 2020-12-25 | 2021-04-13 | 北京巅峰科技有限公司 | Vehicle defect identification method and device, electronic device and storage medium |
| CN113723219A (en) * | 2021-08-10 | 2021-11-30 | 浙江大华技术股份有限公司 | Method for identifying breakage of insulating member and related device |
| CN113469299A (en) * | 2021-09-06 | 2021-10-01 | 常州微亿智造科技有限公司 | Defect detection method and defect detection device in industrial detection |
| CN113469299B (en) * | 2021-09-06 | 2021-12-07 | 常州微亿智造科技有限公司 | Defect detection method and defect detection device in industrial detection |
| CN114742837B (en) * | 2022-06-14 | 2022-09-16 | 华南农业大学 | Deep information image interaction-based wood artwork processing auxiliary instrument and method |
| CN114742837A (en) * | 2022-06-14 | 2022-07-12 | 华南农业大学 | Deep information image interaction-based wood artwork processing auxiliary instrument and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110473197A (en) | Material surface defect detection method, device, equipment and storage medium | |
| Truong et al. | Automatic image thresholding using Otsu’s method and entropy weighting scheme for surface defect detection | |
| Sezgin et al. | Selection of thresholding methods for nondestructive testing applications | |
| CN116188475B (en) | Intelligent control method, system and medium for automatic optical detection of appearance defects | |
| CN105224937B (en) | Fine granularity semanteme color pedestrian recognition methods again based on human part position constraint | |
| CN116611748B (en) | Titanium alloy furniture production quality monitoring system | |
| WO2022247162A1 (en) | Transparency detection method based on machine vision | |
| CN114511718B (en) | Intelligent management method and system for materials for building construction | |
| Wu et al. | Automated visual inspection of surface mounted chip components | |
| CN112613569B (en) | Image recognition method, training method and device for image classification model | |
| CN110188694B (en) | Method for identifying shoe wearing footprint sequence based on pressure characteristics | |
| CN107610119B (en) | The accurate detection method of steel strip surface defect decomposed based on histogram | |
| CN110057828A (en) | A kind of surface inspecting method and system based on machine vision | |
| CN112419260A (en) | PCB character area defect detection method | |
| CN113793332B (en) | Experimental instrument defect identification and classification method and system | |
| CN111310682B (en) | Universal detection analysis and recognition method for text file forms | |
| Cho et al. | A computer vision system for automated grading of rough hardwood lumber using a knowledge-based approach | |
| US11436506B2 (en) | Method and devices for determining metrology sites | |
| CN117115569A (en) | Automatic object image identification and classification method and system based on machine learning | |
| CN112131477A (en) | Library book recommendation system and method based on user portrait | |
| Zhang et al. | Saliency detection for surface defects of ceramic tile | |
| Noh et al. | An effective and efficient defect inspection system for TFT-LCD polarised films using adaptive thresholds and shape-based image analyses | |
| CN109190489A (en) | A kind of abnormal face detecting method based on reparation autocoder residual error | |
| CN115953463A (en) | Package marking method, device and equipment and storage medium | |
| CN115908344A (en) | Micro LED chip defect detection method based on MLCT-YOLO |
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 | ||
| CB02 | Change of applicant information |
Address after: Room 303-309, No.3, Pazhou Avenue East Road, Haizhu District, Guangzhou City, Guangdong Province 510000 Applicant after: Shugen Internet Co.,Ltd. Applicant after: SHANGHAI ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant after: JIANSU ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant after: GUANGZHOU ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant after: BEIJING ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant after: CHANGSHA ROOTCLOUD TECHNOLOGY Co.,Ltd. Address before: 100094 No. six, No. 5, 5 East Tower, 10 northwestern Wang Dong Road, Haidian District, Beijing. Applicant before: IROOTECH TECHNOLOGY Co.,Ltd. Applicant before: SHANGHAI ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant before: JIANSU ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant before: GUANGZHOU ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant before: BEIJING ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant before: CHANGSHA ROOTCLOUD TECHNOLOGY Co.,Ltd. |
|
| CB02 | Change of applicant information | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20210514 Address after: Room 303-309, 3 Pazhou Avenue East Road, Haizhu District, Guangzhou, Guangdong 510000 Applicant after: Shugen Internet Co.,Ltd. Applicant after: SHANGHAI ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant after: JIANSU ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant after: BEIJING ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant after: CHANGSHA ROOTCLOUD TECHNOLOGY Co.,Ltd. Address before: Room 303-309, 3 Pazhou Avenue East Road, Haizhu District, Guangzhou, Guangdong 510000 Applicant before: Shugen Internet Co.,Ltd. Applicant before: SHANGHAI ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant before: JIANSU ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant before: GUANGZHOU ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant before: BEIJING ROOTCLOUD TECHNOLOGY Co.,Ltd. Applicant before: CHANGSHA ROOTCLOUD TECHNOLOGY Co.,Ltd. |
|
| TA01 | Transfer of patent application right | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191119 |
|
| RJ01 | Rejection of invention patent application after publication |