CN110148130A - Method and apparatus for detecting part defect - Google Patents
Method and apparatus for detecting part defect Download PDFInfo
- Publication number
- CN110148130A CN110148130A CN201910446993.8A CN201910446993A CN110148130A CN 110148130 A CN110148130 A CN 110148130A CN 201910446993 A CN201910446993 A CN 201910446993A CN 110148130 A CN110148130 A CN 110148130A
- Authority
- CN
- China
- Prior art keywords
- sample surface
- surface image
- detected
- sample
- defects detection
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- 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
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30164—Workpiece; Machine component
Abstract
The embodiment of the present application discloses the method and apparatus for detecting part defect.One specific embodiment of this method includes: the surface image set to be detected obtained in the part to be detected of multiple special angles acquisition of part to be detected;The surface characteristics set to be detected of part to be detected is extracted from surface image set to be detected using feature extraction algorithm;Surface characteristics set to be detected is detected using defects detection model set trained in advance, obtain the defects detection result of part to be detected, wherein, for defects detection model set for detecting defect existing for part, testing result includes defect classification existing for part to be detected.The embodiment is related to field of cloud calculation, improves the accuracy in detection to part defect.
Description
Technical field
The invention relates to field of computer technology, and in particular to the method and apparatus for detecting part defect.
Background technique
As modernization of industry degree is higher and higher, machine is ubiquitous in life.For example, machine electronics, chemical industry,
It has all been widely used in the various industries such as aerospace.Part is the underlying dimension and machine-building of machine
Basic unit in the process.Since the profile of part, shape, size all consistent with precision when initial design must be just able to satisfy
Machine assembles demand, and therefore, under the industrial environment of high speed development, detection part defect is the indispensable link of secondary industry
One of.
Currently, part defect detection mode is mainly artificial quality inspection.That is, observing by artificial eye piece surface
To determine part with the presence or absence of defect.
Summary of the invention
The embodiment of the present application proposes the method and apparatus for detecting part defect.
In a first aspect, the embodiment of the present application provides a kind of method for detecting part defect, comprising: obtain to be checked
Survey the surface image set to be detected of the part to be detected of multiple special angles acquisition of part;Using feature extraction algorithm to
The surface characteristics set to be detected of part to be detected is extracted in detection surface image set;Utilize defects detection mould trained in advance
Type set detects surface characteristics set to be detected, obtains the defects detection result of part to be detected, wherein defects detection
For model set for detecting defect existing for part, testing result includes defect classification existing for part to be detected.
In some embodiments, part to be detected is being extracted from surface image set to be detected using feature extraction algorithm
Surface characteristics set to be detected before, further includes: surface image set to be detected is pre-processed using preprocess method.
In some embodiments, testing result further includes the position of defect classification existing for part to be detected and to be detected zero
There are the confidence levels of the defects of predetermined defect category set classification for part.
In some embodiments, using defects detection model set trained in advance to surface image features set to be detected
It is detected, obtains the defects detection result of part to be detected, comprising: for the defects of defect category set classification, if lacking
There are the defect classifications to part to be detected for multiple defects detection model inspections in sunken detection model set, are based on multiple defects
There are the confidence levels of the defect classification for the part to be detected that the accuracy rate of detection model and multiple defects detection model inspections arrive, raw
At part to be detected, there are the confidence levels of the defect classification.
In some embodiments, in utilization defects detection model set trained in advance to surface image features collection to be detected
Conjunction is detected, after obtaining the defects detection result of part to be detected, further includes: the defects detection knot based on part to be detected
Fruit sends to terminal device and shows information.
In some embodiments, it is sent based on the defects detection result of part to be detected to terminal device and shows information, packet
It includes: obtaining the defect annotation results of part to be detected;Defects detection result and defect annotation results based on part to be detected, it is raw
At the accuracy rate and recall rate of defects detection model set;The accuracy rate of defects detection model set is sent to terminal device, is called together
The rate of returning, at least one quasi- called together in rate curve and Receiver operating curve.
In some embodiments, it is sent based on the defects detection result of part to be detected to terminal device and shows information, also
Include: defects detection result and defect annotation results based on part to be detected, generates the yields and defect of part to be detected
Rate;At least one of in the yields curve and ratio of defects curve for sending part to be detected to terminal device.
In some embodiments, in the yields curve and ratio of defects curve for sending part to be detected to terminal device
After at least one, comprising: determine whether yields curve is in the first preset range, or determine whether ratio of defects curve is located
In the second preset range;If yields curve exceeds the second preset range, hair beyond the first preset range or ratio of defects curve
Send alarm command.
In some embodiments, defect category set determines as follows: obtaining in the multiple default of sample parts
The sample surface image collection of the sample parts of angle acquisition;It is extracted from sample surface image collection using feature extraction algorithm
The sample surface characteristic set of sample parts;Sample surface characteristic set is clustered using clustering algorithm, it is poly- to obtain first
Class result;Based on the first cluster result, defect category set is generated.
In some embodiments, multiple special angles determine as follows: sample surface characteristic set is divided into
The corresponding multiple sample surface character subsets of multiple predetermined angles;Using clustering algorithm respectively to multiple sample surface character subsets
It is clustered, obtains corresponding multiple second cluster results of multiple predetermined angles;Based on multiple second cluster results, from multiple pre-
If determining multiple special angles in angle.
In some embodiments, clustering algorithm includes k means clustering algorithm, mean shift clustering algorithm, pixel cluster calculation
In method, hierarchical clustering algorithm, spectral clustering and the embedded clustering algorithm of depth based on depth convolutional neural networks at least
One.
In some embodiments, training obtains defects detection model set as follows: for defects detection model
The defects of set detection model, chooses the corresponding sample surface image of the defects detection model from sample surface image collection
Subset;Sample surface feature is extracted from the corresponding sample surface image subset of the defects detection model using feature extraction algorithm
Subset;Using the sample surface feature in the corresponding sample surface character subset of the defects detection model as input, by input
The defect classification of the corresponding sample parts of sample surface feature obtains the defects detection model as output, training.
In some embodiments, the corresponding sample surface image of a defects detection model in defects detection model set
Subset is sample surface image of the sample parts of the special angle acquisition in multiple special angles of sample parts
Collection.
In some embodiments, feature extraction algorithm includes that candidate region generates network algorithm, feature pyramid network is calculated
At least one of in method and deep net algorithm.
In some embodiments, sub from the corresponding sample surface image of the defects detection model using feature extraction algorithm
It concentrates before extracting sample surface character subset, further includes: using preprocess method to the corresponding sample of defects detection model
Surface image subset is pre-processed.
In some embodiments, preprocess method includes at least one of the following: adjustment brightness, gray scale, in contrast
At least one of;At least one of in adjustment size, offset;Adaptively cut figure.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, which is converted into corresponding to
Gray scale surface image;Binaryzation is carried out to the corresponding gray scale surface image of the sample surface image, obtains the sample surface figure
As corresponding binaryzation surface image;Determine the wheel of the sample parts in the corresponding binaryzation surface image of the sample surface image
It is wide;Profile based on the sample parts in the corresponding binaryzation surface image of the sample surface image to the sample surface image into
Row cuts figure.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Part inclination, calculates the affine transformation matrix and spin matrix of the sample surface image;Utilize the affine change of the sample surface image
It changes matrix and spin matrix and affine transformation is carried out to the sample surface image, generate the corresponding new sample of the sample surface image
Surface image.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is that edge is damaged or dirty, adjusts the number of the decision region of the sample surface image.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is peeling, adjusts the brightness and contrast of the sample surface image.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is deformation, adjusts the contrast of the sample surface image, expands the detection block of the sample surface image,
And pixel cluster is carried out to the detection block of the sample surface image.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is window burr, extracts the window outline of the sample parts in the sample surface image, by the sample
The pixel value of the ordinate and two vertical edges of the pixel value of two horizontal edges of the window outline of the sample parts in surface image
Abscissa storage carries out normalizing at one-dimension array, to the pixel value of the window outline of the sample parts in the sample surface image
Change, and generates the line chart of window outline based on normalized pixel value.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is flow liner, expands the detection block of the sample surface image, and the detection to the sample surface image
Frame carries out pixel cluster.
Second aspect, the embodiment of the present application provide a kind of for detecting the device of part defect, comprising: acquiring unit,
It is configured to obtain the surface image set to be detected in the part to be detected of multiple special angles acquisition of part to be detected;It mentions
Unit is taken, is configured to extract the table to be detected of part to be detected from surface image set to be detected using feature extraction algorithm
Region feature set;Detection unit is configured to using defects detection model set trained in advance to surface characteristics collection to be detected
Conjunction is detected, and the defects detection result of part to be detected is obtained, wherein defects detection model set is for detecting part presence
Defect, testing result includes defect classification existing for part to be detected.
In some embodiments, device further include: pretreatment unit is configured to using preprocess method to be detected
Surface image set is pre-processed.
In some embodiments, testing result further includes the position of defect classification existing for part to be detected and to be detected zero
There are the confidence levels of the defects of predetermined defect category set classification for part.
In some embodiments, detection unit is further configured to: for the defects of defect category set classification, if
There are the defect classifications to part to be detected for multiple defects detection model inspections in defects detection model set, are lacked based on multiple
The accuracy rate and the part to be detected that arrives of multiple defects detection model inspections for falling into detection model there are the confidence level of the defect classification,
Generating part to be detected, there are the confidence levels of the defect classification.
In some embodiments, device further include: transmission unit is configured to the defects detection based on part to be detected
As a result it is sent to terminal device and shows information.
In some embodiments, transmission unit is further configured to: obtaining the defect annotation results of part to be detected;Base
In the defects detection result and defect annotation results of part to be detected, generates the accuracy rate of defects detection model set and recall
Rate;Accuracy rate, recall rate, the standard for sending defects detection model set to terminal device call rate curve together and Receiver Operating Characteristics are bent
At least one of in line.
In some embodiments, transmission unit is further configured to: defects detection result based on part to be detected and
Defect annotation results generate the yields and ratio of defects of part to be detected;The yields of part to be detected is sent to terminal device
At least one of in curve and ratio of defects curve.
In some embodiments, which comprises determining that unit, is configured to determine whether yields curve is in first
Preset range, or determine whether ratio of defects curve is in the second preset range;Alarm unit, if being configured to yields curve
Exceed the second preset range beyond the first preset range or ratio of defects curve, sends alarm command.
In some embodiments, defect category set determines as follows: obtaining in the multiple default of sample parts
The sample surface image collection of the sample parts of angle acquisition;It is extracted from sample surface image collection using feature extraction algorithm
The sample surface characteristic set of sample parts;Sample surface characteristic set is clustered using clustering algorithm, it is poly- to obtain first
Class result;Based on the first cluster result, defect category set is generated.
In some embodiments, multiple special angles determine as follows: sample surface characteristic set is divided into
The corresponding multiple sample surface character subsets of multiple predetermined angles;Using clustering algorithm respectively to multiple sample surface character subsets
It is clustered, obtains corresponding multiple second cluster results of multiple predetermined angles;Based on multiple second cluster results, from multiple pre-
If determining multiple special angles in angle.
In some embodiments, clustering algorithm includes k means clustering algorithm, mean shift clustering algorithm, pixel cluster calculation
In method, hierarchical clustering algorithm, spectral clustering and the embedded clustering algorithm of depth based on depth convolutional neural networks at least
One.
In some embodiments, training obtains defects detection model set as follows: for defects detection model
The defects of set detection model, chooses the corresponding sample surface image of the defects detection model from sample surface image collection
Subset;Sample surface feature is extracted from the corresponding sample surface image subset of the defects detection model using feature extraction algorithm
Subset;Using the sample surface feature in the corresponding sample surface character subset of the defects detection model as input, by input
The defect classification of the corresponding sample parts of sample surface feature obtains the defects detection model as output, training.
In some embodiments, the corresponding sample surface image of a defects detection model in defects detection model set
Subset is sample surface image of the sample parts of the special angle acquisition in multiple special angles of sample parts
Collection.
In some embodiments, feature extraction algorithm includes that candidate region generates network algorithm, feature pyramid network is calculated
At least one of in method and deep net algorithm.
In some embodiments, sub from the corresponding sample surface image of the defects detection model using feature extraction algorithm
It concentrates before extracting sample surface character subset, further includes: using preprocess method to the corresponding sample of defects detection model
Surface image subset is pre-processed.
In some embodiments, preprocess method includes at least one of the following: adjustment brightness, gray scale, in contrast
At least one of;At least one of in adjustment size, offset;Adaptively cut figure.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, which is converted into corresponding to
Gray scale surface image;Binaryzation is carried out to the corresponding gray scale surface image of the sample surface image, obtains the sample surface figure
As corresponding binaryzation surface image;Determine the wheel of the sample parts in the corresponding binaryzation surface image of the sample surface image
It is wide;Profile based on the sample parts in the corresponding binaryzation surface image of the sample surface image to the sample surface image into
Row cuts figure.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Part inclination, calculates the affine transformation matrix and spin matrix of the sample surface image;Utilize the affine change of the sample surface image
It changes matrix and spin matrix and affine transformation is carried out to the sample surface image, generate the corresponding new sample of the sample surface image
Surface image.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is that edge is damaged or dirty, adjusts the number of the decision region of the sample surface image.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is peeling, adjusts the brightness and contrast of the sample surface image.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is deformation, adjusts the contrast of the sample surface image, expands the detection block of the sample surface image,
And pixel cluster is carried out to the detection block of the sample surface image.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is window burr, extracts the window outline of the sample parts in the sample surface image, by the sample
The pixel value of the ordinate and two vertical edges of the pixel value of two horizontal edges of the window outline of the sample parts in surface image
Abscissa storage carries out normalizing at one-dimension array, to the pixel value of the window outline of the sample parts in the sample surface image
Change, and generates the line chart of window outline based on normalized pixel value.
In some embodiments, using preprocess method to the corresponding sample surface image subset of the defects detection model into
Row pretreatment, comprising: for the sample surface image in sample surface image subset, if the sample zero in the sample surface image
Defect classification existing for part is flow liner, expands the detection block of the sample surface image, and the detection to the sample surface image
Frame carries out pixel cluster.
The third aspect, the embodiment of the present application provide a kind of server, which includes: one or more processors;
Storage device is stored thereon with one or more programs;When one or more programs are executed by one or more processors, so that
One or more processors realize the method as described in implementation any in first aspect.
Fourth aspect, the embodiment of the present application provide a kind of computer-readable medium, are stored thereon with computer program, should
The method as described in implementation any in first aspect is realized when computer program is executed by processor.
Method and apparatus provided by the embodiments of the present application for detecting part defect are obtained in part to be detected first
The surface image set to be detected of the part to be detected of multiple special angle acquisitions;Then using feature extraction algorithm to be detected
The surface characteristics set to be detected of part to be detected is extracted in surface image set;Finally utilize defects detection mould trained in advance
Type set detects surface characteristics set to be detected, obtains the defects detection result of part to be detected.Utilize defects detection
Model set detects defect existing for part automatically, and entire detection process is participated in without artificial, not only reduces human cost, also
Improve the detection efficiency to part defect.Also, multiple defects detection models detect defect existing for part jointly, reduce
To the false detection rate and omission factor of part defect, and then improve the accuracy in detection to part defect.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the application's is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is that this application can be applied to exemplary system architectures therein;
Fig. 2 is the flow chart according to one embodiment of the method for detecting part defect of the application;
Fig. 3 is the flow chart according to another embodiment of the method for detecting part defect of the application;
Fig. 4 is the flow chart according to one embodiment of the method for determining defect category set of the application;
Fig. 5 is the flow chart according to one embodiment of the method for determining special angle of the application;
Fig. 6 is the flow chart according to one embodiment of the method for training defects detection model set of the application;
Fig. 7 is the structural schematic diagram according to one embodiment of the device for detecting part defect of the application;
Fig. 8 is adapted for the structural schematic diagram for the computer system for realizing the server of the embodiment of the present application.
Specific embodiment
The application is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, part relevant to related invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the features in the embodiments and the embodiments of the present application can phase
Mutually combination.The application is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 is shown can the method for detecting part defect using the application or the dress for detecting part defect
The exemplary system architecture 100 for the embodiment set.
As shown in Figure 1, may include image capture device 101,102,103, network 104 and service in system architecture 100
Device 105.Network 104 between image capture device 101,102,103 and server 105 to provide the medium of communication link.
Network 104 may include various connection classifications, such as wired, wireless communication link or fiber optic cables etc..
Image capture device 101,102,103 can send the table for the part that it is acquired by network 104 to server 105
Face image.Image capture device 101,102,103 can be hardware, be also possible to software.When image capture device 101,102,
103 be hardware when, can be support image collecting function various electronic equipments.Including but not limited to camera, video camera, phase
Machine and smart phone etc..When image capture device 101,102,103 is software, may be mounted in above-mentioned electronic equipment.
Multiple softwares or software module may be implemented into it, and single software or software module also may be implemented into.Specific limit is not done herein
It is fixed.
Server 105 can be to provide the server of various services.Such as part defect detection service device.Part defect inspection
Survey the surface to be detected for the part to be detected that server can acquire the multiple special angles in part to be detected got
The data such as image collection carry out the processing such as analyzing, and generate processing result (such as defects detection result of part to be detected).
It should be noted that server 105 can be hardware, it is also possible to software.It, can when server 105 is hardware
To be implemented as the distributed server cluster that multiple servers form, individual server also may be implemented into.When server 105 is
When software, multiple softwares or software module (such as providing Distributed Services) may be implemented into, also may be implemented into single
Software or software module.It is not specifically limited herein.
It should be noted that for detecting the method for part defect generally by server provided by the embodiment of the present application
105 execute, and correspondingly, the device for detecting part defect is generally positioned in server 105.
It should be understood that the number of image capture device, network and server in Fig. 1 is only schematical.According to reality
It now needs, can have any number of image capture device, network and server.
With continued reference to Fig. 2, it illustrates according to one embodiment of the method for detecting part defect of the application
Process 200.The method for being used to detect part defect, comprising the following steps:
Step 201, the exterior view to be detected in the part to be detected of multiple special angles acquisition of part to be detected is obtained
Image set closes.
In the present embodiment, for detecting the executing subject (such as server 105 shown in FIG. 1) of the method for part defect
Can from be arranged near part to be detected more image capture devices (such as image capture device shown in FIG. 1 101,
102,103) surface image set to be detected is obtained.Wherein, an image capture device and part to be detected are in a specific angle
Degree, the surface image to be detected of part to be detected is acquired for the special angle in part to be detected.For example, to be checked in face
It surveys and three image capture devices is set at the front, side and the back side of part, this three image capture devices acquire to be checked respectively
Survey part direct picture, side image and back side image, and the surface image to be detected as part to be detected be added to
It detects in surface image set.
Step 202, the to be detected of part to be detected is extracted from surface image set to be detected using feature extraction algorithm
Surface characteristics set.
In the present embodiment, for every surface image to be detected in surface image set to be detected, above-mentioned execution master
Body can use feature extraction algorithm and extract the surface characteristics to be detected of part to be detected from the surface image to be detected, and add
It is added in the surface characteristics set to be detected of part to be detected.Wherein, feature extraction algorithm can include but is not limited to down toward
One item missing: RPN (Region Proposal Network, candidate region generate network) algorithm, FPN (Feature Pyramid
Networks, feature pyramid network) algorithm and DN (Deep Net, deep net) algorithm etc..Surface characteristics to be detected can be
For the information that feature possessed by the surface to the part to be detected in surface image to be detected is described, including but it is unlimited
In various fundamentals (such as surface shape, surface color, surface flatness etc.) relevant to the surface of part to be detected.It is logical
Often, surface characteristics to be detected can be indicated with multi-C vector.
In some optional implementations of the present embodiment, above-mentioned executing subject can be first with preprocess method pair
Surface image set to be detected is pre-processed, and then executes step 202 again.In general, being carried out to surface image set to be detected
Pretreated preprocess method with to for train defects detection model set sample surface image collection carry out it is pretreated
Preprocess method is identical, referring specifically to fig. 6 shown in embodiment.
Step 203, surface characteristics set to be detected is detected using defects detection model set trained in advance, is obtained
To the defects detection result of part to be detected.
In the present embodiment, above-mentioned executing subject can use defects detection model set to surface characteristics set to be detected
It is detected, to obtain the defects detection result of part to be detected.Wherein, defects detection model set can be used for detecting part
Existing defect characterizes the corresponding relationship between the surface characteristics set of part and the defects detection result of part.To be detected zero
The testing result of part may include defect classification existing for part to be detected.Optionally, the testing result of part to be detected may be used also
To include that there are in predetermined defect category set for the position of defect classification and part to be detected existing for part to be detected
Defect classification confidence level.It wherein, may include that crackle, scratch, bulge, sizing, trachoma, edge touch in defect category set
The defects of wound, dirty, peeling, deformation, window burr and flow liner classification.
It is adopted in general, a defects detection model in defects detection model set can be used for detecting in a special angle
Defect existing for part in the surface image of collection.Therefore, for each special angle in multiple special angles, above-mentioned execution
The corresponding surface characteristics to be detected of the special angle in surface characteristics set to be detected can be input to the spy first by main body
The corresponding defects detection model of angle is determined, to obtain the corresponding defects detection result of the special angle.Then, summarize multiple specific
The corresponding multiple defects detections of angle are as a result, obtain the defects detection result of part to be detected.
In the present embodiment, defects detection model set can be trained in several ways and be obtained.For example, being examined for defect
Each defects detection model in model set is surveyed, above-mentioned executing subject can collect the sample in the acquisition of a special angle in advance
Defect classification existing for sample parts in this surface image and sample surface image, and corresponding storage generates mapping table,
As the defects detection model.After getting surface characteristics set to be detected, above-mentioned executing subject can be first to be checked
Survey surface characteristics set in choose the corresponding surface characteristics to be detected of the special angle, then calculate the special angle it is corresponding to
Detect the similarity between surface characteristics and the sample surface feature of each sample surface image in mapping table;Then base
In the calculated similarity of institute, the corresponding defects detection result of the special angle is determined from mapping table;Finally, summarizing
The corresponding multiple defects detections of multiple special angles are as a result, obtain the defects detection result of part to be detected.
In some optional implementations of the present embodiment, for every kind of defect classification in defect category set, if
There are the defect classification, above-mentioned execution masters to part to be detected for multiple defects detection model inspections in defects detection model set
Body can the part to be detected that arrives of accuracy rate based on multiple defects detection models and multiple defects detection model inspections exist should
The confidence level of defect classification, generating part to be detected, there are the confidence levels of the defect classification.For example, above-mentioned executing subject can be selected
There are the confidence levels of the defect classification for the part to be detected that the defects detection model inspection for taking accuracy rate high arrives, as to be detected zero
There are the confidence levels of the defect classification for part.In another example above-mentioned executing subject can be according to the accuracy rate of multiple defects detection models
Normalized is done, and using the fractional value after normalized as corresponding weight factor.
In some optional implementations of the present embodiment, above-mentioned executing subject can be based on the defect of part to be detected
Testing result sends to terminal device and shows information.For example, above-mentioned executing subject can be directly to be detected to terminal device transmission
The defects detection result of part.In this way, terminal device can show the defects detection result of part to be detected.
Method provided by the embodiments of the present application for detecting part defect obtains multiple spies in part to be detected first
Determine the surface image set to be detected of the part to be detected of angle acquisition;Then utilize feature extraction algorithm from exterior view to be detected
The surface characteristics set to be detected of part to be detected is extracted in image set conjunction;Finally utilize defects detection model set trained in advance
Surface characteristics set to be detected is detected, the defects detection result of part to be detected is obtained.Utilize defects detection Models Sets
Defect existing for automatic detection part is closed, entire detection process participates in without artificial, not only reduces human cost, also improve
To the detection efficiency of part defect.Also, multiple defects detection models detect defect existing for part jointly, reduce to part
The false detection rate and omission factor of defect, and then improve the accuracy in detection to part defect.
With further reference to Fig. 3, it illustrates another implementations according to the method for detecting part defect of the application
The process 300 of example.The method for being used to detect part defect, comprising the following steps:
Step 301, the exterior view to be detected in the part to be detected of multiple special angles acquisition of part to be detected is obtained
Image set closes.
Step 302, the to be detected of part to be detected is extracted from surface image set to be detected using feature extraction algorithm
Surface characteristics set.
Step 303, surface characteristics set to be detected is detected using defects detection model set trained in advance, is obtained
To the defects detection result of part to be detected.
In the present embodiment, the concrete operations of step 301-303 are in the embodiment shown in Figure 2 in step 201-203
It is described in detail, details are not described herein.
Step 304, the defect annotation results of part to be detected are obtained.
In the present embodiment, for detecting the executing subject (such as server 105 shown in FIG. 1) of the method for part defect
The defect annotation results of available part to be detected.Wherein, the defect annotation results of part to be detected can be to be detected
It is that part mark after artificial quality inspection as a result, including defect classification existing for part to be detected that artificial quality inspection goes out.Optionally,
The defect annotation results of part to be detected can also include the position of defect classification existing for the part to be detected of artificial quality inspection out.
Step 305, defects detection result and defect annotation results based on part to be detected generate defects detection Models Sets
The accuracy rate and recall rate of conjunction.
In the present embodiment, above-mentioned executing subject can be marked based on the defects detection result and defect of part to be detected and be tied
Fruit calculates the accuracy rate and recall rate of defects detection model set.For example, above-mentioned executing subject can by defects detection result and
The ratio of the quantity of the quantity and all parts to be detected of the consistent part to be detected of defect annotation results is as defects detection mould
The accuracy rate of type set.The part to be detected that defects detection result and defect annotation results can be positive by above-mentioned executing subject
Recall rate as defects detection model set of quantity and the ratio of the quantity of part to be detected that is positive of defect annotation results.
Step 306, to terminal device send defects detection model set accuracy rate, recall rate, standard call together rate curve and by
At least one of in examination person's performance curve.
In the present embodiment, above-mentioned executing subject can to terminal device send defects detection model set accuracy rate,
Recall rate, standard call at least one in rate curve (PRC curve) and Receiver operating curve's (ROC curve) together.In this way, terminal
Equipment can show the displaying information received.In general, PRC curve and ROC curve are according to defects detection model set
What accuracy rate and recall rate were drawn out.
Step 307, defects detection result and defect annotation results based on part to be detected, generate the good of part to be detected
Product rate and ratio of defects.
In the present embodiment, above-mentioned executing subject can be marked based on the defects detection result and defect of part to be detected and be tied
Fruit calculates the yields and ratio of defects of part to be detected.For example, above-mentioned executing subject defects detection result can be positive to
Detect yields of the ratio of the quantity of part and the quantity of all parts to be detected as part to be detected.Above-mentioned executing subject
The ratio of the quantity for the part to be detected that defects detection result can be negative and the quantity of all parts to be detected is as to be checked
Survey the ratio of defects of part.Wherein, defects detection result is positive, and indicating the part to be detected, there is no defect, defects detection results
It is negative, indicates the part existing defects to be detected.
At least one of in the yields curve and ratio of defects curve for step 308, sending part to be detected to terminal device.
In the present embodiment, above-mentioned executing subject can send the yields curve of part to be detected and be lacked to terminal device
Fall at least one in rate curve.In this way, terminal device can show the displaying information received.
Step 309, it determines whether yields curve is in the first preset range, or determines whether ratio of defects curve is in
Second preset range.
In the present embodiment, above-mentioned executing subject can determine whether yields curve is in the first preset range, or
Determine whether ratio of defects curve is in the second preset range.If yields curve exceeds the first preset range or defect
Rate curve exceeds the second preset range, executes step 310.
Step 310, alarm command is sent.
In the present embodiment, if yields curve is default beyond second beyond the first preset range or ratio of defects curve
Range, above-mentioned executing subject can send alarm command to warning device.In this way, warning device can issue alarm sound,
Inside machine to prompt processing part or there are abnormal conditions in the production line of processing part.
From figure 3, it can be seen that being used to detect part defect in the present embodiment compared with the corresponding embodiment of Fig. 2
The process 300 of method increases the step of transmission shows information and alarm command.The scheme of the present embodiment description passes through end as a result,
End equipment show in the accuracy rate of defects detection model set, recall rate, PRC curve and ROC curve at least one of, and to
At least one in the yields and ratio of defects of part is detected, the defect for being more intuitive to see part to be detected is allowed users to
Situation.Also, the case where yields curve exceeds the second preset range beyond the first preset range or ratio of defects curve
Under, by warning device sending alarm sound, machine inside or the processing zero of part can be processed according to prompt discovery in time
The abnormal conditions that the production line of part occurs.
With further reference to Fig. 4, it illustrates the processes for the one embodiment for determining the method for defect category set
400.The method for being used to determine defect category set, comprising the following steps:
Step 401, the sample surface image collection in the sample parts of multiple predetermined angles acquisition of sample parts is obtained.
In the present embodiment, for determining executing subject (such as the server shown in FIG. 1 of the method for defect category set
105) sample surface image collection can be obtained from the great amount of images acquisition equipment being arranged near sample parts.Wherein, one
Platform image capture device and sample parts are in a predetermined angle, for the predetermined angle collecting sample part in sample parts
Sample surface image.For example, an image capture device is arranged every certain angle (such as 10 degree).Sample parts can be with
It is that there are the parts of various defect classifications.
Step 402, the sample surface for extracting sample parts from sample surface image collection using feature extraction algorithm is special
Collection is closed.
In the present embodiment, for every sample surface image in sample surface image collection, above-mentioned executing subject can
To extract sample surface feature from the sample surface image using feature extraction algorithm, and it is added to sample surface characteristic set
In.Wherein, sample surface feature can be for feature possessed by the surface to the sample parts in sample surface image into
The information of row description, including but not limited to relevant to the surface of sample parts various fundamentals (such as surface shape, surface
Color, surface flatness etc.).In general, sample surface feature can be indicated with multi-C vector.
It should be noted that feature extraction algorithm here is identical as the feature extraction algorithm in embodiment shown in Fig. 2,
Which is not described herein again.
Step 403, sample surface characteristic set is clustered using clustering algorithm, obtains the first cluster result.
In the present embodiment, above-mentioned executing subject can use clustering algorithm and cluster to sample surface characteristic set,
To obtain the first cluster result.Wherein, cluster is usually and is divided into a large amount of physics or abstract object to be made of similar object
The process of multiple set.These objects and the object in the same set are similar to each other, different with the object in other set.It is poly-
Class algorithm can include but is not limited at least one of following: k-means (k mean value) clustering algorithm, mean shift clustering algorithm, as
Plain clustering algorithm, hierarchical clustering algorithm, spectral clustering and DEC_DCNN (Deep Embedded Clustering based
On Deep Convolutional Neural Network, the embedded cluster of depth based on depth convolutional neural networks) it calculates
Method.First cluster result may include multiple sample surface feature clusters.Sample surface feature in same sample surface feature cluster
Similar, the sample surface feature in different sample surface feature clusters is different.
Step 404, it is based on the first cluster result, generates defect category set.
In the present embodiment, above-mentioned executing subject can be based on the first cluster result, generate defect category set.Practice
In, there are the surface characteristics of the part of the defect of same defect classification is similar, there are the parts of the defect of different defect classifications
Surface characteristics is different, and therefore, a sample surface feature cluster in the first cluster result can correspond to a kind of defect classification.This
In, for each sample surface feature cluster in the first cluster result, above-mentioned executing subject can be to the sample surface feature cluster
It is analyzed, to determine the corresponding defect classification of sample surface feature cluster.
In some optional implementations of the present embodiment, for each predetermined angle in multiple predetermined angles, on
State the sample surface feature that the sample surface feature of the sample surface image of the predetermined angle can also be distributed by executing subject
The image capture device institute that the corresponding defect classification of cluster is determined as the predetermined angle can collected defect classification.
With further reference to Fig. 5, it illustrates the processes 500 of one embodiment for determining the method for special angle.It should
Method for determining special angle, comprising the following steps:
Step 501, sample surface characteristic set is divided into corresponding multiple sample surface feature of multiple predetermined angles
Collection.
In the present embodiment, for determining the executing subject (such as server 105 shown in FIG. 1) of the method for special angle
Sample surface characteristic set can be divided into the corresponding multiple sample surface character subsets of multiple predetermined angles.In general, in sample
The sample surface image of the sample parts of the same predetermined angle acquisition of this part belongs to same sample surface character subset.
Step 502, multiple sample surface character subsets are clustered respectively using clustering algorithm, obtains multiple preset angles
Spend corresponding multiple second cluster results.
In the present embodiment, above-mentioned executing subject can use clustering algorithm respectively to multiple sample surface character subsets into
Row cluster, to obtain corresponding multiple second cluster results of multiple predetermined angles.Wherein, each predetermined angle corresponding second is gathered
Class result may include multiple sample surface feature clusters.
It should be noted that clustering algorithm here is identical as the clustering algorithm in embodiment shown in Fig. 4, here no longer
It repeats.
Step 503, multiple second cluster results are based on, multiple special angles are determined from multiple predetermined angles.
In the present embodiment, above-mentioned executing subject can be based on multiple second cluster results, from multiple predetermined angles really
Fixed multiple special angles.For example, above-mentioned executing subject can be based on each predetermined angle in multiple predetermined angles corresponding
The Clustering Effect of two cluster results, it is good to select Clustering Effect from multiple predetermined angles, and being capable of collected defect classification
Multiple special angles of all defect classification in set.In general, the sample surface feature phase in same sample surface feature cluster
Higher like spending, the sample surface characteristic similarity in different sample surface feature clusters is lower, and Clustering Effect is better.
With further reference to Fig. 6, it illustrates the streams for the one embodiment for training the method for defects detection model set
Journey 600.The method for being used to train defects detection model set, comprising the following steps:
Step 601, it for the defects of defects detection model set detection model, is chosen from sample surface image collection
The corresponding sample surface image subset of the defects detection model.
In the present embodiment, for each defects detection model in defects detection model set, for training defect to examine
The executing subject (such as server 105 shown in FIG. 1) for surveying the method for model set can be selected from sample surface image collection
Take the corresponding sample surface image subset of the defects detection model.In general, a defects detection in defects detection model set
The corresponding sample surface image subset of model can be the acquisition of a special angle in multiple special angles of sample parts
Sample parts sample surface image subset.
Step 602, it is extracted from the corresponding sample surface image subset of the defects detection model using feature extraction algorithm
Sample surface character subset.
In the present embodiment, every sample surface in sample surface image subset corresponding for the defects detection model
Image, above-mentioned executing subject can use feature extraction algorithm and extract sample surface feature from the sample surface image, and add
It is added in the corresponding sample surface character subset of the defects detection model.
It should be noted that feature extraction algorithm here is identical as the feature extraction algorithm in embodiment shown in Fig. 2,
Which is not described herein again.
In some optional implementations of the present embodiment, above-mentioned executing subject can be first first with pretreatment side
Method pre-processes the corresponding sample surface image subset of the defects detection model, then executes step 602 again.
In some optional implementations of the present embodiment, preprocess method can include but is not limited to following at least one
: adjustment brightness, gray scale, in contrast at least one of;At least one of in adjustment size, offset;Adaptively cut figure.
In some optional implementations of the present embodiment, for the sample surface figure in sample surface image subset
The sample surface image can be converted into corresponding gray scale surface image first by picture, above-mentioned executing subject;Later to the sample
The corresponding gray scale surface image of this surface image carries out binaryzation, obtains the corresponding binaryzation exterior view of the sample surface image
Picture;Then the profile of the sample parts in the corresponding binaryzation surface image of the sample surface image is determined;Finally it is based on the sample
The profile of sample parts in the corresponding binaryzation surface image of this surface image carries out cutting figure to the sample surface image.
In some optional implementations of the present embodiment, for the sample surface figure in sample surface image subset
Picture, if the sample parts in the sample surface image tilt, above-mentioned executing subject can calculate the sample surface image first
Affine transformation matrix and spin matrix;Then using the affine transformation matrix of the sample surface image and spin matrix to the sample
Surface image carries out affine transformation, generates the corresponding new sample surface image of the sample surface image.By affine transformation come
It realizes translation, scaling, the rotation etc. of image, to generate new sample surface image, realizes training sample augmentation, help to improve
The generalization ability of the defects detection model trained.
In some optional implementations of the present embodiment, for the sample surface figure in sample surface image subset
Picture, if defect classification existing for sample parts in the sample surface image is that edge is damaged or dirty, above-mentioned executing subject can
To adjust the number of the decision region of the sample surface image.In general, edge is damaged or dirty similar in Image Visual Feature
Degree is higher, and above-mentioned executing subject can suitably adjust the number of the decision region of sample surface image, then delete choosing by threshold value
False detection rate can be effectively reduced in method.For example, (Region-based Fully Convolutional Networks, is based on RFCN
The full convolutional network in region) algorithm by the way that target area is divided into k × k decision region, calculates separately k × k decision
The similarity in region is averaged as overall similarity, participates in defect classification similarity calculation.Damaged due to edge and
Dirty visual signature difference maximum on the image is: the damage edge of this kind of defect of edge is irregular zigzag, and
Dirty edge is then smooth curve.Therefore, more fringe region participative decision makings are allowed by suitably tuning up k value, can allow side
Edge is damaged to be distinguished in terms of similarity with dirty, eventually by threshold value delete choosing method can effectively prevent it is dirty by erroneous detection at edge
It damages.
In some optional implementations of the present embodiment, for the sample surface figure in sample surface image subset
Picture, if defect classification existing for sample parts in the sample surface image is peeling, the adjustable sample of above-mentioned executing subject
The brightness and contrast of this surface image.Since peeling is sensitive to brightness variation characteristic, if bright in image acquisition process
Brightness, which changes, be easy to cause missing inspection and erroneous detection.Due to the unique images of peeling be characterized in than normal segments color more deeply, than
Dirty part is more of light color, the brightness of image and the characteristics of image of the prominent peeling of contrast is adjusted by appropriateness, then pass through
The method of image segmentation improves the accuracy in detection of peeling.
In some optional implementations of the present embodiment, for the sample surface figure in sample surface image subset
Picture, if defect classification existing for sample parts in the sample surface image is deformation, the adjustable sample of above-mentioned executing subject
The contrast of this surface image, expands the detection block of the sample surface image, and to the detection block of the sample surface image into
Row pixel cluster.In general, the Image Visual Feature of deformation is very unobvious, image capturing angle and polishing are required all very
It is high.By, in 10 degree of relatively acquisition images, deformation being allowed to have the characteristics of image of shading value variation on the image from textured surface.Pass through
Appropriate adjustment contrast, the characteristics of image for allowing light and shade to change is more obvious, reuses deep learning method and carries out to deformation position
Detection.Further, since deformation is generally present in the centre on some surface of part, and area is larger with respect to other class defects, position
Variation range is smaller, in order to improve accuracy rate, by expanding according to a certain percentage to detection block, keeps detection range wider, accurately
Du Genggao.Pixel cluster is carried out in the detection block after expansion, by being analyzed to determine whether there is change to cluster result
Shape.
In some optional implementations of the present embodiment, for the sample surface figure in sample surface image subset
Picture, if defect classification existing for sample parts in the sample surface image is window burr, above-mentioned executing subject can be extracted
The window outline of sample parts in the sample surface image, by the window outline of the sample parts in the sample surface image
The abscissa of the pixel value of the ordinate of the pixel value of two horizontal edges and two vertical edges is stored at one-dimension array, to the sample surface
The pixel value of the window outline of sample parts in image is normalized, and generates window outline based on normalized pixel value
Line chart.In general, normal window profile is rectangle, there are the window outlines of burr to have concave-convex phenomenon, takes window outline
The abscissa of the pixel value of the ordinate of the pixel value of two horizontal edges and two vertical edges is stored at one-dimension array, at normalization
Reason, all pixels value of window outline is transformed into 0 to 1 range, line chart is drawn.The line chart of normal window is four sections
Straight line, and there are the line charts of the window of burr then will appear exceptional value.For example, Yi Huocheng high or low compared with the pixel value of surrounding
Curve distribution.By being detected to these exceptional values, to realize the detection to window burr.
In some optional implementations of the present embodiment, for the sample surface figure in sample surface image subset
Picture, if defect classification existing for sample parts in the sample surface image is flow liner, above-mentioned executing subject can expand the sample
The detection block of this surface image, and pixel cluster is carried out to the detection block of the sample surface image.In general, the image of flow liner is special
It levies obvious, if directly detected using deep learning method, haves the shortcomings that false detection rate is high.In order to reduce erroneous detection
Rate expands detection block, carries out pixel cluster to the detection block after expansion, flow liner passes through meeting and normal region after pixel cluster
By apparent boundary, so that the false detection rate of flow liner be effectively reduced.
Step 603, using the sample surface feature in the corresponding sample surface character subset of the defects detection model as defeated
Enter, using the defect classification of the corresponding sample parts of sample surface feature of input as output, training obtains the defects detection mould
Type.
In the present embodiment, above-mentioned executing subject can will be in the corresponding sample surface character subset of the defects detection model
Sample surface feature as input, using the defect classification of the corresponding sample parts of sample surface feature of input as export,
The defects detection model is obtained using the training of deep learning method.
In general, above-mentioned executing subject can input sample surface feature from the input side of target detection model, through looking over so as to check
The processing for marking detection model, from the testing result of outlet side output sample parts.Then, above-mentioned executing subject can be based on sample
The accuracy in detection of the defect classification target detection model of the testing result and sample parts of part.If accuracy in detection is unsatisfactory for
Preset constraint condition then adjusts the parameter of target detection model, subsequently inputs sample surface feature and continues model
Training.If accuracy in detection meets preset constraint condition, model training is completed, and sample surface feature at this time is
The defects detection model.
With further reference to Fig. 7, as the realization to method shown in above-mentioned each figure, this application provides one kind for detecting zero
One embodiment of the device of part defect, the Installation practice is corresponding with embodiment of the method shown in Fig. 2, which specifically may be used
To be applied in various electronic equipments.
As shown in fig. 7, the device 700 for detecting part defect of the present embodiment may include: acquiring unit 701, mention
Take unit 702 and detection unit 703.Wherein, acquiring unit 701 are configured to obtain multiple specific angles in part to be detected
Spend the surface image set to be detected of the part to be detected of acquisition;Extraction unit 702, be configured to using feature extraction algorithm from
The surface characteristics set to be detected of part to be detected is extracted in surface image set to be detected;Detection unit 703, is configured to benefit
Surface characteristics set to be detected is detected with defects detection model set trained in advance, obtains the defect of part to be detected
Testing result, wherein for defects detection model set for detecting defect existing for part, testing result includes that part to be detected is deposited
Defect classification.
In the present embodiment, in the device 700 for detecting part defect: acquiring unit 701, extraction unit 702 and inspection
The specific processing and its brought technical effect for surveying unit 703 can be respectively with reference to step 201, the steps in Fig. 2 corresponding embodiment
Rapid 202 and step 203 related description, details are not described herein.
In some optional implementations of the present embodiment, for detecting the device 700 of part defect further include: pre- place
Unit (not shown) is managed, is configured to pre-process surface image set to be detected using preprocess method.
In some optional implementations of the present embodiment, testing result further includes defect class existing for part to be detected
There are the confidence levels of the defects of predetermined defect category set classification for other position and part to be detected.
In some optional implementations of the present embodiment, detection unit 703 is further configured to: for defect class
Not Ji He the defects of classification, if multiple defects detection model inspections in defects detection model set exist to part to be detected
The defect classification, accuracy rate and multiple defects detection model inspections based on multiple defects detection models to part to be detected deposit
In the confidence level of the defect classification, generating part to be detected, there are the confidence levels of the defect classification.
In some optional implementations of the present embodiment, for detecting the device 700 of part defect further include: send
Unit (not shown) is configured to the defects detection result based on part to be detected to terminal device and sends displaying information.
In some optional implementations of the present embodiment, transmission unit is further configured to: obtaining to be detected zero
The defect annotation results of part;Defects detection result and defect annotation results based on part to be detected generate defects detection model
The accuracy rate and recall rate of set;Accuracy rate, recall rate, the standard for sending defects detection model set to terminal device call rate curve together
With at least one in Receiver operating curve.
In some optional implementations of the present embodiment, transmission unit is further configured to: being based on to be detected zero
The defects detection result and defect annotation results of part, generate the yields and ratio of defects of part to be detected;It is sent to terminal device
At least one of in the yields curve and ratio of defects curve of part to be detected.
In some optional implementations of the present embodiment, the device 700 for detecting part defect comprises determining that list
First (not shown), is configured to determine whether yields curve is in the first preset range, or determines ratio of defects curve
Whether the second preset range is in;Alarm unit (not shown), if being configured to yields curve beyond the first default model
It encloses or ratio of defects curve is beyond the second preset range, send alarm command.
In some optional implementations of the present embodiment, defect category set determines as follows: obtaining
The sample surface image collection of the sample parts of multiple predetermined angles acquisition of sample parts;Using feature extraction algorithm from sample
The sample surface characteristic set of sample parts is extracted in surface image set;Using clustering algorithm to sample surface characteristic set into
Row cluster, obtains the first cluster result;Based on the first cluster result, defect category set is generated.
In some optional implementations of the present embodiment, multiple special angles determine as follows: by sample
Surface characteristics set is divided into the corresponding multiple sample surface character subsets of multiple predetermined angles;Using clustering algorithm respectively to more
A sample surface character subset is clustered, and corresponding multiple second cluster results of multiple predetermined angles are obtained;Based on multiple
Two clusters from multiple predetermined angles as a result, determine multiple special angles.
In some optional implementations of the present embodiment, clustering algorithm includes that k means clustering algorithm, average drifting are poly-
Class algorithm, pixel cluster algorithm, hierarchical clustering algorithm, spectral clustering and depth based on depth convolutional neural networks are embedded
At least one of in clustering algorithm.
In some optional implementations of the present embodiment, defects detection model set is trained as follows
It arrives: for the defects of defects detection model set detection model, the defects detection mould is chosen from sample surface image collection
The corresponding sample surface image subset of type;It is sub from the corresponding sample surface image of the defects detection model using feature extraction algorithm
It concentrates and extracts sample surface character subset;Sample surface in the corresponding sample surface character subset of the defects detection model is special
Sign is as input, and using the defect classification of the corresponding sample parts of sample surface feature of input as output, training obtains this and lacks
Fall into detection model.
A defects detection model in some optional implementations of the present embodiment, in defects detection model set
Corresponding sample surface image subset is the sample zero of the special angle acquisition in multiple special angles of sample parts
The sample surface image subset of part.
In some optional implementations of the present embodiment, feature extraction algorithm includes that candidate region generates network calculation
At least one of in method, feature pyramid network algorithm and deep net algorithm.
In some optional implementations of the present embodiment, using feature extraction algorithm from the defects detection model pair
In the sample surface image subset answered before extraction sample surface character subset, further includes: using preprocess method to the defect
The corresponding sample surface image subset of detection model is pre-processed.
In some optional implementations of the present embodiment, preprocess method includes at least one of the following: that adjustment is bright
Degree, gray scale, in contrast at least one of;At least one of in adjustment size, offset;Adaptively cut figure.
It is corresponding to the defects detection model using preprocess method in some optional implementations of the present embodiment
Sample surface image subset is pre-processed, comprising: for the sample surface image in sample surface image subset, by the sample
Surface image is converted into corresponding gray scale surface image;Two-value is carried out to the corresponding gray scale surface image of the sample surface image
Change, obtains the corresponding binaryzation surface image of the sample surface image;Determine the corresponding binaryzation surface of the sample surface image
The profile of sample parts in image;Wheel based on the sample parts in the corresponding binaryzation surface image of the sample surface image
Exterior feature carries out cutting figure to the sample surface image.
It is corresponding to the defects detection model using preprocess method in some optional implementations of the present embodiment
Sample surface image subset is pre-processed, comprising: for the sample surface image in sample surface image subset, if the sample
Sample parts inclination in surface image, calculates the affine transformation matrix and spin matrix of the sample surface image;Utilize the sample
The affine transformation matrix and spin matrix of this surface image carry out affine transformation to the sample surface image, generate the sample surface
The corresponding new sample surface image of image.
It is corresponding to the defects detection model using preprocess method in some optional implementations of the present embodiment
Sample surface image subset is pre-processed, comprising: for the sample surface image in sample surface image subset, if the sample
Defect classification existing for sample parts in surface image is that edge is damaged or dirty, adjusts the decision area of the sample surface image
The number in domain.
It is corresponding to the defects detection model using preprocess method in some optional implementations of the present embodiment
Sample surface image subset is pre-processed, comprising: for the sample surface image in sample surface image subset, if the sample
Defect classification existing for sample parts in surface image is peeling, adjusts the brightness and contrast of the sample surface image.
It is corresponding to the defects detection model using preprocess method in some optional implementations of the present embodiment
Sample surface image subset is pre-processed, comprising: for the sample surface image in sample surface image subset, if the sample
Defect classification existing for sample parts in surface image is deformation, adjusts the contrast of the sample surface image, expands the sample
The detection block of this surface image, and pixel cluster is carried out to the detection block of the sample surface image.
It is corresponding to the defects detection model using preprocess method in some optional implementations of the present embodiment
Sample surface image subset is pre-processed, comprising: for the sample surface image in sample surface image subset, if the sample
Defect classification existing for sample parts in surface image is window burr, extracts the sample parts in the sample surface image
Window outline, by the ordinate and two of the pixel value of two horizontal edges of the window outline of the sample parts in the sample surface image
The abscissa storage of the pixel value of vertical edge is at one-dimension array, to the window outlines of the sample parts in the sample surface image
Pixel value is normalized, and the line chart of window outline is generated based on normalized pixel value.
It is corresponding to the defects detection model using preprocess method in some optional implementations of the present embodiment
Sample surface image subset is pre-processed, comprising: for the sample surface image in sample surface image subset, if the sample
Defect classification existing for sample parts in surface image is flow liner, expands the detection block of the sample surface image, and to this
The detection block of sample surface image carries out pixel cluster.
Below with reference to Fig. 8, it illustrates the server for being suitable for being used to realize the embodiment of the present application (such as clothes shown in FIG. 1
Be engaged in device 105) computer system 800 structural schematic diagram.Server shown in Fig. 8 is only an example, should not be to this Shen
Please embodiment function and use scope bring any restrictions.
As shown in figure 8, computer system 800 includes central processing unit (CPU) 801, it can be read-only according to being stored in
Program in memory (ROM) 802 or be loaded into the program in random access storage device (RAM) 803 from storage section 808 and
Execute various movements appropriate and processing.In RAM 803, also it is stored with system 800 and operates required various programs and data.
CPU 801, ROM 802 and RAM 803 are connected with each other by bus 804.Input/output (I/O) interface 805 is also connected to always
Line 804.
I/O interface 805 is connected to lower component: the importation 806 including keyboard, mouse etc.;It is penetrated including such as cathode
The output par, c 807 of spool (CRT), liquid crystal display (LCD) etc. and loudspeaker etc.;Storage section 808 including hard disk etc.;
And the communications portion 809 of the network interface card including LAN card, modem etc..Communications portion 809 via such as because
The network of spy's net executes communication process.Driver 810 is also connected to I/O interface 805 as needed.Detachable media 811, such as
Disk, CD, magneto-optic disk, semiconductor memory etc. are mounted on as needed on driver 810, in order to read from thereon
Computer program be mounted into storage section 808 as needed.
Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description
Software program.For example, embodiment of the disclosure includes a kind of computer program product comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed from network by communications portion 809, and/or from detachable media
811 are mounted.When the computer program is executed by central processing unit (CPU) 701, limited in execution the present processes
Above-mentioned function.
It should be noted that computer-readable medium described herein can be computer-readable signal media or meter
Calculation machine readable storage medium storing program for executing either the two any combination.Computer readable storage medium for example can be --- but not
Be limited to --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor system, device or device, or any above combination.Meter
The more specific example of calculation machine readable storage medium storing program for executing can include but is not limited to: have the electrical connection, just of one or more conducting wires
Taking formula computer disk, hard disk, random access storage device (RAM), read-only memory (ROM), erasable type may be programmed read-only storage
Device (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic memory device,
Or above-mentioned any appropriate combination.In this application, computer readable storage medium can be it is any include or storage journey
The tangible medium of sequence, the program can be commanded execution system, device or device use or in connection.And at this
In application, computer-readable signal media may include in a base band or as carrier wave a part propagate data-signal,
Wherein carry computer-readable program code.The data-signal of this propagation can take various forms, including but unlimited
In electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal media can also be that computer can
Any computer-readable medium other than storage medium is read, which can send, propagates or transmit and be used for
By the use of instruction execution system, device or device or program in connection.Include on computer-readable medium
Program code can transmit with any suitable medium, including but not limited to: wireless, electric wire, optical cable, RF etc. are above-mentioned
Any appropriate combination.
The calculating of the operation for executing the application can be write with one or more programming languages or combinations thereof
Machine program code, described program design language include object-oriented programming language-such as Java, Smalltalk, C+
+, further include conventional procedural programming language-such as " C " language or similar programming language.Program code can
Fully to execute, partly execute on the user computer on the user computer, be executed as an independent software package,
Part executes on the remote computer or executes on a remote computer or server completely on the user computer for part.
In situations involving remote computers, remote computer can pass through the network of any kind --- including local area network (LAN)
Or wide area network (WAN)-is connected to subscriber computer, or, it may be connected to outer computer (such as utilize Internet service
Provider is connected by internet).
Flow chart and block diagram in attached drawing are illustrated according to the system of the various embodiments of the application, method and computer journey
The architecture, function and operation in the cards of sequence product.In this regard, each box in flowchart or block diagram can generation
A part of one module, program segment or code of table, a part of the module, program segment or code include one or more use
The executable instruction of the logic function as defined in realizing.It should also be noted that in some implementations as replacements, being marked in box
The function of note can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are actually
It can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it to infuse
Meaning, the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart can be with holding
The dedicated hardware based system of functions or operations as defined in row is realized, or can use specialized hardware and computer instruction
Combination realize.
Being described in unit involved in the embodiment of the present application can be realized by way of software, can also be by hard
The mode of part is realized.Described unit also can be set in the processor, for example, can be described as: a kind of processor packet
Include acquiring unit, extraction unit and detection unit.Wherein, the title of these units is not constituted under certain conditions to the unit
The restriction of itself, for example, acquiring unit be also described as " obtain part to be detected multiple special angles acquire to
Detect the unit of the surface image set to be detected of part ".
As on the other hand, present invention also provides a kind of computer-readable medium, which be can be
Included in server described in above-described embodiment;It is also possible to individualism, and without in the supplying server.It is above-mentioned
Computer-readable medium carries one or more program, when said one or multiple programs are executed by the server,
So that the server: obtaining the exterior view image set to be detected in the part to be detected of multiple special angles acquisition of part to be detected
It closes;The surface characteristics set to be detected of part to be detected is extracted from surface image set to be detected using feature extraction algorithm;
Surface characteristics set to be detected is detected using defects detection model set trained in advance, obtains lacking for part to be detected
Fall into testing result, wherein for defects detection model set for detecting defect existing for part, testing result includes part to be detected
Existing defect classification.
Above description is only the preferred embodiment of the application and the explanation to institute's application technology principle.Those skilled in the art
Member is it should be appreciated that invention scope involved in the application, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature
Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed herein
Can technical characteristic replaced mutually and the technical solution that is formed.
Claims (32)
1. a kind of method for detecting part defect, comprising:
Obtain the surface image set to be detected in the part to be detected of multiple special angles acquisition of part to be detected;
The surface to be detected of the part to be detected is extracted from the surface image set to be detected using feature extraction algorithm
Characteristic set;
The surface characteristics set to be detected is detected using defects detection model set trained in advance, obtain it is described to
Detect the defects detection result of part, wherein the defects detection model set is for detecting defect existing for part, the inspection
Surveying result includes defect classification existing for the part to be detected.
2. according to the method described in claim 1, wherein, utilizing feature extraction algorithm from the surface image to be detected described
Before the surface characteristics set to be detected for extracting the part to be detected in set, further includes:
The surface image set to be detected is pre-processed using preprocess method.
3. method according to claim 1 or 2, wherein the testing result further includes existing for the part to be detected
There are the confidence levels of the defects of predetermined defect category set classification for the position of defect classification and the part to be detected.
4. according to the method described in claim 3, wherein, it is described using defects detection model set trained in advance to it is described to
Detection surface characteristics set is detected, and the defects detection result of the part to be detected is obtained, comprising:
For the defects of defect category set classification, if multiple defects detection moulds in the defects detection model set
Type detects the part to be detected, and there are the defect classifications, accuracy rate based on the multiple defects detection model and described more
There are the confidence levels of the defect classification for the part to be detected that a defects detection model inspection arrives, and generate the part to be detected
There are the confidence levels of the defect classification.
5. according to the method described in claim 1, wherein, utilizing defects detection model set trained in advance to described described
Surface characteristics set to be detected is detected, after obtaining the defects detection result of the part to be detected, further includes:
Defects detection result based on the part to be detected sends to terminal device and shows information.
6. according to the method described in claim 5, wherein, the defects detection result based on the part to be detected is to terminal
Equipment, which is sent, shows information, comprising:
Obtain the defect annotation results of the part to be detected;
Defects detection result and defect annotation results based on the part to be detected, generate the defects detection model set
Accuracy rate and recall rate;
Accuracy rate, recall rate, the standard for sending the defects detection model set to the terminal device call rate curve and subject together
At least one of in performance curve.
7. according to the method described in claim 6, wherein, the defects detection result based on the part to be detected is to terminal
Equipment, which is sent, shows information, further includes:
Defects detection result and defect annotation results based on the part to be detected generate the yields of the part to be detected
And ratio of defects;
At least one in the yields curve and ratio of defects curve of the part to be detected is sent to the terminal device.
8. according to the method described in claim 7, wherein, sending the good of the part to be detected to the terminal device described
After at least one in product rate curve and ratio of defects curve, comprising:
It determines whether the yields curve is in the first preset range, or determines whether the ratio of defects curve is in second
Preset range;
If the yields curve exceeds the described second default model beyond first preset range or the ratio of defects curve
It encloses, sends alarm command.
9. according to the method described in claim 3, wherein, the defect category set determines as follows:
Obtain the sample surface image collection in the sample parts of multiple predetermined angles acquisition of sample parts;
The sample surface for extracting the sample parts from the sample surface image collection using the feature extraction algorithm is special
Collection is closed;
The sample surface characteristic set is clustered using clustering algorithm, obtains the first cluster result;
Based on first cluster result, the defect category set is generated.
10. according to the method described in claim 9, wherein, the multiple special angle determines as follows:
The sample surface characteristic set is divided into the corresponding multiple sample surface character subsets of the multiple predetermined angle;
The multiple sample surface character subset is clustered respectively using the clustering algorithm, obtains the multiple preset angle
Spend corresponding multiple second cluster results;
Based on the multiple second cluster result, the multiple special angle is determined from the multiple predetermined angle.
11. method according to claim 9 or 10, wherein the clustering algorithm includes k means clustering algorithm, mean value drift
It is embedding to move clustering algorithm, pixel cluster algorithm, hierarchical clustering algorithm, spectral clustering and depth based on depth convolutional neural networks
Enter at least one in formula clustering algorithm.
12. according to the method described in claim 9, wherein, training obtains the defects detection model set as follows:
For the defects of defects detection model set detection model, this is chosen from the sample surface image collection and is lacked
Fall into the corresponding sample surface image subset of detection model;
Sample surface is extracted from the corresponding sample surface image subset of the defects detection model using the feature extraction algorithm
Character subset;
Using the sample surface feature in the corresponding sample surface character subset of the defects detection model as input, by the sample of input
The defect classification of the corresponding sample parts of this surface characteristics obtains the defects detection model as output, training.
13. according to the method for claim 12, wherein a defects detection model in the defects detection model set
Corresponding sample surface image subset is the special angle acquisition in the multiple special angle of the sample parts
The sample parts sample surface image subset.
14. according to the method for claim 12, wherein the feature extraction algorithm includes that candidate region generates network calculation
At least one of in method, feature pyramid network algorithm and deep net algorithm.
15. according to the method for claim 12, wherein utilize the feature extraction algorithm from the defects detection mould described
In the corresponding sample surface image subset of type before extraction sample surface character subset, further includes:
The corresponding sample surface image subset of the defects detection model is pre-processed using the preprocess method.
16. according to the method for claim 15, wherein the preprocess method includes at least one of the following:
Adjust brightness, gray scale, in contrast at least one of;
At least one of in adjustment size, offset;
Adaptively cut figure.
17. according to the method for claim 16, wherein described to utilize the preprocess method to the defects detection model pair
The sample surface image subset answered is pre-processed, comprising:
For the sample surface image in the sample surface image subset, which is converted into corresponding gray scale
Surface image;
Binaryzation is carried out to the corresponding gray scale surface image of the sample surface image, obtains the corresponding two-value of sample surface image
Change surface image;
Determine the profile of the sample parts in the corresponding binaryzation surface image of the sample surface image;
Profile based on the sample parts in the corresponding binaryzation surface image of the sample surface image is to the sample surface image
It carries out cutting figure.
18. according to the method for claim 16, wherein described to utilize the preprocess method to the defects detection model pair
The sample surface image subset answered is pre-processed, comprising:
For the sample surface image in the sample surface image subset, if the sample parts in the sample surface image incline
Tiltedly, the affine transformation matrix and spin matrix of the sample surface image are calculated;
Affine transformation is carried out to the sample surface image using the affine transformation matrix and spin matrix of the sample surface image, it is raw
At the corresponding new sample surface image of the sample surface image.
19. according to the method for claim 16, wherein described to utilize the preprocess method to the defects detection model pair
The sample surface image subset answered is pre-processed, comprising:
For the sample surface image in the sample surface image subset, if the sample parts in the sample surface image exist
Defect classification be that edge is damaged or dirty, adjust the number of the decision region of the sample surface image.
20. according to the method for claim 16, wherein described to utilize the preprocess method to the defects detection model pair
The sample surface image subset answered is pre-processed, comprising:
For the sample surface image in the sample surface image subset, if the sample parts in the sample surface image exist
Defect classification be peeling, adjust the brightness and contrast of the sample surface image.
21. according to the method for claim 16, wherein described to utilize the preprocess method to the defects detection model pair
The sample surface image subset answered is pre-processed, comprising:
For the sample surface image in the sample surface image subset, if the sample parts in the sample surface image exist
Defect classification be deformation, adjust the contrast of the sample surface image, expand the detection block of the sample surface image and right
The detection block of the sample surface image carries out pixel cluster.
22. according to the method for claim 16, wherein described to utilize the preprocess method to the defects detection model pair
The sample surface image subset answered is pre-processed, comprising:
For the sample surface image in the sample surface image subset, if the sample parts in the sample surface image exist
Defect classification be window burr, the window outline of the sample parts in the sample surface image is extracted, by the sample surface figure
The abscissa of the pixel value of the ordinate and two vertical edges of the pixel value of two horizontal edges of the window outline of the sample parts as in
One-dimension array is stored into, the pixel value of the window outline of the sample parts in the sample surface image is normalized, and base
The line chart of window outline is generated in normalized pixel value.
23. according to the method for claim 16, wherein described to utilize the preprocess method to the defects detection model pair
The sample surface image subset answered is pre-processed, comprising:
For the sample surface image in the sample surface image subset, if the sample parts in the sample surface image exist
Defect classification be flow liner, expand the detection block of the sample surface image, and carry out to the detection block of the sample surface image
Pixel cluster.
24. a kind of for detecting the device of part defect, comprising:
Acquiring unit, be configured to obtain part to be detected multiple special angles acquisition the part to be detected it is to be checked
Survey surface image set;
Extraction unit is configured to extract from the surface image set to be detected using feature extraction algorithm described to be detected
The surface characteristics set to be detected of part;
Detection unit, be configured to using defects detection model set trained in advance to the surface characteristics set to be detected into
Row detection, obtains the defects detection result of the part to be detected, wherein the defects detection model set is for detecting part
Existing defect, the testing result include defect classification existing for the part to be detected.
25. device according to claim 24, wherein described device further include:
Pretreatment unit is configured to pre-process the surface image set to be detected using preprocess method.
26. device according to claim 24, wherein described device further include:
Transmission unit is configured to the defects detection result based on the part to be detected to terminal device and sends displaying information.
27. device according to claim 25, wherein the defect category set determines as follows:
Obtain the sample surface image collection in the sample parts of multiple predetermined angles acquisition of sample parts;
The sample surface for extracting the sample parts from the sample surface image collection using the feature extraction algorithm is special
Collection is closed;
The sample surface characteristic set is clustered using clustering algorithm, obtains the first cluster result;
Based on first cluster result, the defect category set is generated.
28. device according to claim 27, wherein the multiple special angle determines as follows:
The sample surface characteristic set is divided into the corresponding multiple sample surface character subsets of the multiple predetermined angle;
The multiple sample surface character subset is clustered respectively using the clustering algorithm, obtains the multiple preset angle
Spend corresponding multiple second cluster results;
Based on the multiple second cluster result, the multiple special angle is determined from the multiple predetermined angle.
29. device according to claim 27, wherein the defects detection model set is trained as follows
It arrives:
For the defects of defects detection model set detection model, this is chosen from the sample surface image collection and is lacked
Fall into the corresponding sample surface image subset of detection model;
Sample surface is extracted from the corresponding sample surface image subset of the defects detection model using the feature extraction algorithm
Character subset;
Using the sample surface feature in the corresponding sample surface character subset of the defects detection model as input, by the sample of input
The defect classification of the corresponding sample parts of this surface characteristics obtains the defects detection model as output, training.
30. device according to claim 29, wherein utilize the feature extraction algorithm from the defects detection mould described
In the corresponding sample surface image subset of type before extraction sample surface character subset, further includes:
The corresponding sample surface image subset of the defects detection model is pre-processed using the preprocess method.
31. a kind of server, comprising:
One or more processors;
Storage device is stored thereon with one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors are real
The now method as described in any in claim 1-23.
32. a kind of computer-readable medium, is stored thereon with computer program, wherein the computer program is held by processor
The method as described in any in claim 1-23 is realized when row.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910446993.8A CN110148130B (en) | 2019-05-27 | 2019-05-27 | Method and device for detecting part defects |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910446993.8A CN110148130B (en) | 2019-05-27 | 2019-05-27 | Method and device for detecting part defects |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110148130A true CN110148130A (en) | 2019-08-20 |
CN110148130B CN110148130B (en) | 2022-02-11 |
Family
ID=67592030
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910446993.8A Active CN110148130B (en) | 2019-05-27 | 2019-05-27 | Method and device for detecting part defects |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110148130B (en) |
Cited By (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110261392A (en) * | 2019-06-19 | 2019-09-20 | 北京百度网讯科技有限公司 | Quality determining method, device, electronic equipment and system |
CN110569899A (en) * | 2019-09-03 | 2019-12-13 | 嘉陵江亭子口水利水电开发有限公司 | Dam face defect classification model training method and device |
CN111008954A (en) * | 2019-11-04 | 2020-04-14 | 联想(北京)有限公司 | Information processing method and device, electronic equipment and storage medium |
CN111105399A (en) * | 2019-12-19 | 2020-05-05 | 苏州经贸职业技术学院 | Switch surface defect detection method and system |
CN111210421A (en) * | 2020-01-10 | 2020-05-29 | 苏州杰锐思智能科技股份有限公司 | Batch generation method and system for lithium battery surface defect images |
CN111402220A (en) * | 2020-03-11 | 2020-07-10 | 北京百度网讯科技有限公司 | Method and device for acquiring information |
CN111681220A (en) * | 2020-06-04 | 2020-09-18 | 阿丘机器人科技(苏州)有限公司 | Defect detection model construction method, device and system and storage medium |
CN111768404A (en) * | 2020-07-08 | 2020-10-13 | 北京滴普科技有限公司 | Mask appearance defect detection system, method and device and storage medium |
WO2021082921A1 (en) * | 2019-10-28 | 2021-05-06 | 上海悦易网络信息技术有限公司 | Back cover appearance defect detection method for electronic apparatus, and apparatus |
CN113689431A (en) * | 2021-10-27 | 2021-11-23 | 常州微亿智造科技有限公司 | Industrial product appearance defect detection method and device |
CN113936000A (en) * | 2021-12-16 | 2022-01-14 | 武汉欧易塑胶包装有限公司 | Injection molding wave flow mark identification method based on image processing |
CN115049641A (en) * | 2022-08-10 | 2022-09-13 | 启东普力马机械有限公司 | Electric data processing method and system for anomaly detection of mechanical parts |
CN116718616A (en) * | 2023-08-04 | 2023-09-08 | 苏州视谷视觉技术有限公司 | Machine vision detection system and detection method for flaw detection |
CN117409010A (en) * | 2023-12-15 | 2024-01-16 | 菲特(天津)检测技术有限公司 | Paint surface defect detection model training, detecting and encoding method and detecting system |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0742431A1 (en) * | 1995-05-10 | 1996-11-13 | Mahlo GmbH & Co. KG | Method and apparatus for detecting flaws in moving fabrics or the like |
US20090202135A1 (en) * | 2008-02-07 | 2009-08-13 | Keyence Corporation | Defect Detection Apparatus, Defect Detection Method and Computer Program |
CN104794491A (en) * | 2015-04-28 | 2015-07-22 | 重庆大学 | Fuzzy clustering steel plate surface defect detection method based on pre classification |
KR20170022346A (en) * | 2015-08-20 | 2017-03-02 | 충북대학교 산학협력단 | System and Method for Classification of PCB fault and Type of Fault |
CN106570853A (en) * | 2015-10-08 | 2017-04-19 | 上海深邃智能科技有限公司 | Shape and color integration insulator identification and defect detection method |
CN107506702A (en) * | 2017-08-08 | 2017-12-22 | 江西高创保安服务技术有限公司 | Human face recognition model training and test system and method based on multi-angle |
CN108776808A (en) * | 2018-05-25 | 2018-11-09 | 北京百度网讯科技有限公司 | A kind of method and apparatus for detecting ladle corrosion defect |
CN108830837A (en) * | 2018-05-25 | 2018-11-16 | 北京百度网讯科技有限公司 | A kind of method and apparatus for detecting ladle corrosion defect |
CN109300102A (en) * | 2017-07-21 | 2019-02-01 | 凌云光技术集团有限责任公司 | A kind of defect inspection method for PCB |
CN109461149A (en) * | 2018-10-31 | 2019-03-12 | 泰州市创新电子有限公司 | The intelligent checking system and method for lacquered surface defect |
US20190147586A1 (en) * | 2017-11-15 | 2019-05-16 | Omron Corporation | Defect inspection apparatus, defect inspection method, and program thereof |
-
2019
- 2019-05-27 CN CN201910446993.8A patent/CN110148130B/en active Active
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0742431A1 (en) * | 1995-05-10 | 1996-11-13 | Mahlo GmbH & Co. KG | Method and apparatus for detecting flaws in moving fabrics or the like |
US20090202135A1 (en) * | 2008-02-07 | 2009-08-13 | Keyence Corporation | Defect Detection Apparatus, Defect Detection Method and Computer Program |
CN104794491A (en) * | 2015-04-28 | 2015-07-22 | 重庆大学 | Fuzzy clustering steel plate surface defect detection method based on pre classification |
KR20170022346A (en) * | 2015-08-20 | 2017-03-02 | 충북대학교 산학협력단 | System and Method for Classification of PCB fault and Type of Fault |
CN106570853A (en) * | 2015-10-08 | 2017-04-19 | 上海深邃智能科技有限公司 | Shape and color integration insulator identification and defect detection method |
CN109300102A (en) * | 2017-07-21 | 2019-02-01 | 凌云光技术集团有限责任公司 | A kind of defect inspection method for PCB |
CN107506702A (en) * | 2017-08-08 | 2017-12-22 | 江西高创保安服务技术有限公司 | Human face recognition model training and test system and method based on multi-angle |
US20190147586A1 (en) * | 2017-11-15 | 2019-05-16 | Omron Corporation | Defect inspection apparatus, defect inspection method, and program thereof |
CN108776808A (en) * | 2018-05-25 | 2018-11-09 | 北京百度网讯科技有限公司 | A kind of method and apparatus for detecting ladle corrosion defect |
CN108830837A (en) * | 2018-05-25 | 2018-11-16 | 北京百度网讯科技有限公司 | A kind of method and apparatus for detecting ladle corrosion defect |
CN109461149A (en) * | 2018-10-31 | 2019-03-12 | 泰州市创新电子有限公司 | The intelligent checking system and method for lacquered surface defect |
Non-Patent Citations (2)
Title |
---|
YAO RONG ET.AL: "《Rapid detection method for fabric defects based on machine vision》", 《2010 INTERNATIONAL CONFERENCE ON COMPUTER APPLICATION AND SYSTEM MODELING》 * |
甘胜丰: "《带钢表面缺陷图像检测与分类方法研究》", 《CNKI博士学位论文全文库》 * |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110261392A (en) * | 2019-06-19 | 2019-09-20 | 北京百度网讯科技有限公司 | Quality determining method, device, electronic equipment and system |
CN110569899A (en) * | 2019-09-03 | 2019-12-13 | 嘉陵江亭子口水利水电开发有限公司 | Dam face defect classification model training method and device |
CN110569899B (en) * | 2019-09-03 | 2022-06-10 | 嘉陵江亭子口水利水电开发有限公司 | Dam face defect classification model training method and device |
WO2021082921A1 (en) * | 2019-10-28 | 2021-05-06 | 上海悦易网络信息技术有限公司 | Back cover appearance defect detection method for electronic apparatus, and apparatus |
CN111008954A (en) * | 2019-11-04 | 2020-04-14 | 联想(北京)有限公司 | Information processing method and device, electronic equipment and storage medium |
CN111105399A (en) * | 2019-12-19 | 2020-05-05 | 苏州经贸职业技术学院 | Switch surface defect detection method and system |
CN111210421A (en) * | 2020-01-10 | 2020-05-29 | 苏州杰锐思智能科技股份有限公司 | Batch generation method and system for lithium battery surface defect images |
CN111210421B (en) * | 2020-01-10 | 2023-12-01 | 苏州杰锐思智能科技股份有限公司 | Batch generation method and system for surface defect images of lithium battery |
WO2021179565A1 (en) * | 2020-03-11 | 2021-09-16 | 北京百度网讯科技有限公司 | Method and apparatus for acquiring information |
CN111402220B (en) * | 2020-03-11 | 2023-06-09 | 北京百度网讯科技有限公司 | Method and device for acquiring information |
CN111402220A (en) * | 2020-03-11 | 2020-07-10 | 北京百度网讯科技有限公司 | Method and device for acquiring information |
CN111681220B (en) * | 2020-06-04 | 2024-02-13 | 阿丘机器人科技(苏州)有限公司 | Method, device, system and storage medium for constructing defect detection model |
CN111681220A (en) * | 2020-06-04 | 2020-09-18 | 阿丘机器人科技(苏州)有限公司 | Defect detection model construction method, device and system and storage medium |
CN111768404A (en) * | 2020-07-08 | 2020-10-13 | 北京滴普科技有限公司 | Mask appearance defect detection system, method and device and storage medium |
CN113689431A (en) * | 2021-10-27 | 2021-11-23 | 常州微亿智造科技有限公司 | Industrial product appearance defect detection method and device |
CN113936000A (en) * | 2021-12-16 | 2022-01-14 | 武汉欧易塑胶包装有限公司 | Injection molding wave flow mark identification method based on image processing |
CN115049641A (en) * | 2022-08-10 | 2022-09-13 | 启东普力马机械有限公司 | Electric data processing method and system for anomaly detection of mechanical parts |
CN116718616A (en) * | 2023-08-04 | 2023-09-08 | 苏州视谷视觉技术有限公司 | Machine vision detection system and detection method for flaw detection |
CN116718616B (en) * | 2023-08-04 | 2023-11-03 | 苏州视谷视觉技术有限公司 | Machine vision detection system and detection method for flaw detection |
CN117409010A (en) * | 2023-12-15 | 2024-01-16 | 菲特(天津)检测技术有限公司 | Paint surface defect detection model training, detecting and encoding method and detecting system |
CN117409010B (en) * | 2023-12-15 | 2024-04-26 | 菲特(天津)检测技术有限公司 | Paint surface defect detection model training, detecting and encoding method and detecting system |
Also Published As
Publication number | Publication date |
---|---|
CN110148130B (en) | 2022-02-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110148130A (en) | Method and apparatus for detecting part defect | |
CN108510472B (en) | Method and apparatus for handling image | |
CN108154196B (en) | Method and apparatus for exporting image | |
CN109086718A (en) | Biopsy method, device, computer equipment and storage medium | |
CN109117831A (en) | The training method and device of object detection network | |
CN107274402A (en) | A kind of Lung neoplasm automatic testing method and system based on chest CT image | |
CN109635627A (en) | Pictorial information extracting method, device, computer equipment and storage medium | |
CN108229418B (en) | Human body key point detection method and apparatus, electronic device, storage medium, and program | |
CN103871046B (en) | Setting apparatus, control method thereof, image processing apparatus, and control method thereof | |
CN107491771A (en) | Method for detecting human face and device | |
CN109949286A (en) | Method and apparatus for output information | |
CN109446990A (en) | Method and apparatus for generating information | |
CN108171191B (en) | Method and apparatus for detecting face | |
CN109389589A (en) | Method and apparatus for statistical number of person | |
CN109308490A (en) | Method and apparatus for generating information | |
CN109800698A (en) | Icon detection method based on depth network | |
CN108416379A (en) | Method and apparatus for handling cervical cell image | |
CN109145828A (en) | Method and apparatus for generating video classification detection model | |
Zhang et al. | Efficient system of cracking-detection algorithms with 1-mm 3D-surface models and performance measures | |
CN108876756A (en) | The measure and device of image similarity | |
CN108171211A (en) | Biopsy method and device | |
CN110335313A (en) | Audio collecting device localization method and device, method for distinguishing speek person and system | |
CN109766818A (en) | Pupil center's localization method and system, computer equipment and readable storage medium storing program for executing | |
CN108830842A (en) | A kind of medical image processing method based on Corner Detection | |
CN109363697A (en) | A kind of method and device of breast image lesion identification |
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 |