CN110443791A - A kind of workpiece inspection method and its detection device based on deep learning network - Google Patents
A kind of workpiece inspection method and its detection device based on deep learning network Download PDFInfo
- Publication number
- CN110443791A CN110443791A CN201910712712.9A CN201910712712A CN110443791A CN 110443791 A CN110443791 A CN 110443791A CN 201910712712 A CN201910712712 A CN 201910712712A CN 110443791 A CN110443791 A CN 110443791A
- Authority
- CN
- China
- Prior art keywords
- workpiece
- image
- detected
- network
- bounding box
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20092—Interactive image processing based on input by user
- G06T2207/20104—Interactive definition of region of interest [ROI]
-
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Quality & Reliability (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of workpiece inspection method and its detection device based on deep learning network obtains the location information of sample workpiece image specifically includes the following steps: collecting sample workpiece image, is labeled sample workpiece image;Workpiece image to be detected is acquired, workpiece image to be detected is sent into image buffer storage area, judges image buffer storage area with the presence or absence of workpiece;If there are workpiece on workpiece image to be detected, defects detection is carried out to workpiece image to be detected;If workpiece is not present on workpiece image to be detected, stop detecting.It is able to achieve while being classified to workpiece, defects detection, the flexibility suitable for multi items, small lot, the frequent workpiece of product renewing detects.
Description
Technical field
The invention belongs to workpiece inspection method technical fields, are related to a kind of workpiece sensing side based on deep learning network
Method;Further relate to the detection device of above-mentioned workpiece inspection method.
Background technique
Manufacturing industry is the basic activity of national economy, and with the development of science and technology, intelligence manufacture has become from manufacture
The breach and only way that big country changes to manufacturing power.Industrial robot is as important dress irreplaceable in intelligence manufacture
Standby and means, it has also become measure the important symbol of national a manufacturing industry level and scientific and technological level.Based on industrial robot
Robot industry, exactly crack manufacturing industry cost increase, environmental constraints problem important path selection.In intelligence manufacture
In industry, detection using industrial robot vision's technology to part is beneficial to the technological problems of discovery production process, changes in time
Into production technology problem, product quality is improved, is conducive to part classification and sorting.
It is well known that intelligence manufacture, to meet the flexible production of multi-varieties and small-batch feature, quality testing must adapt to
Variety classes part classification, positioning, feature extraction and the detection of the different surface defect of each category feature;Currently, in workpiece sensing
Artificial vision is relied primarily on to complete, although manual time's detection is adaptable, is able to carry out the inspection of multiple types workpiece features
It surveys, but detection speed is slow, relies on artificial experience, be difficult to meet the requirement of on-line quick detection;And existing robot vision detection
Speed is fast, but is difficult the problem of adapting to multiple types workpiece, different characteristic detection.
Summary of the invention
The object of the present invention is to provide a kind of workpiece inspection methods based on deep learning network, solve in the prior art
The problem of existing robot vision detection can not adapt to multiple types workpiece sensing.
The technical scheme adopted by the invention is that a kind of workpiece inspection method based on deep learning network, specifically includes
Following steps:
Step 1, collecting sample workpiece image, are labeled sample workpiece image, obtain the position of sample workpiece image
Information;
Step 2, acquisition workpiece image to be detected, send workpiece image to be detected into image buffer storage area, judge image buffer storage
Area whether there is workpiece;
If there are workpiece on step 3, workpiece image to be detected, defects detection is carried out to workpiece image to be detected;If to be checked
It surveys and workpiece is not present on workpiece image, then stop detecting.
The features of the present invention also characterized in that:
Step 1 specifically: collecting sample workpiece image is labeled sample workpiece image by image labeling tool,
Obtain workpiece classification, the pixel coordinate of sample workpiece image.
Step 2 specifically: workpiece image to be detected is sent into image buffer storage area, suggests network in work to be detected by region
After drawing suspicious region Suggestion box on part image, workpiece image to be detected is sent into feature pyramid and carries out image segmentation, is obtained
The suspicious region of workpiece image to be detected and sample workpiece image are compared, are sentenced by the suspicious region of workpiece image to be detected
Disconnected image buffer storage area whether there is workpiece.
Step 3 specifically:
Workpiece image to be detected is inputted residual error network by step 3.1, obtains characteristic image;
Characteristic image input area suggestion network is obtained class parameter, logic of class parameter and bounding box parameter by step 3.2,
By class parameter, logic of class parameter and bounding box parameter by obtaining interest region after non-maxima suppression;
Step 3.3, the target detection that the location information input area of interest region and sample workpiece image is suggested to network
Layer carries out target detection, obtains initial workpiece classification, bounding box, segmentation mask, by workpiece classification, bounding box, segmentation mask
The alignment of region of interest domain is carried out after combining with characteristic image, is obtained feature pyramid network, is passed through the positioning of feature pyramid network
Network and segmentation network obtain workpiece classification, workpiece bounding box and the workpiece segmentation mask of workpiece image to be detected.
Step 3.4 inputs accurate workpiece classification, workpiece bounding box and workpiece segmentation mask in ROI classifier, obtains just
Then the defect classification of beginning and rough defect bounding box obtain final mask result by dividing mask network.
It is a further object to provide a kind of Workpiece detecting devices based on deep learning network.
Another technical solution adopted in the present invention is that a kind of Workpiece detecting device based on deep learning network, packet
It includes:
Locating module is used for collecting sample workpiece image, is labeled to sample workpiece image, obtains sample workpiece image
Location information;
Identification module judges image buffer storage area with the presence or absence of work for workpiece image to be detected to be sent into image buffer storage area
Part;
Detection module carries out defects detection to workpiece image to be detected if there are workpiece on workpiece image to be detected;If to
It detects and workpiece is not present on workpiece image, then stop detecting.
Locating module includes:
First acquisition module is used for collecting sample workpiece image;
Labeling module obtains the location information of sample workpiece image for being labeled to sample workpiece image.
Identification module includes:
Second acquisition module, for acquiring workpiece image to be detected;
Judgment module judges image buffer storage area with the presence or absence of work for workpiece image to be detected to be sent into image buffer storage area
Part.
Detection module includes:
First obtains module, for workpiece image to be detected to be inputted residual error network, obtains characteristic image;
Second obtains module, for characteristic image input area suggestion network to be obtained class parameter, logic of class parameter and side
Boundary's frame parameter, by class parameter, logic of class parameter and bounding box parameter by obtaining interest region after non-maxima suppression;
Third obtains module, for the location information input area in interest region and sample workpiece image to be suggested network
Target detection layer carry out target detection, obtain initial workpiece classification, bounding box, segmentation mask, by workpiece classification, bounding box,
The alignment of region of interest domain is carried out after dividing mask and characteristic image combination, feature pyramid network is obtained, passes through feature pyramid network
The positioning network and segmentation network of network obtain workpiece classification, workpiece bounding box and the workpiece segmentation mask of workpiece image to be detected;
4th obtains module, for accurate workpiece classification, workpiece bounding box and workpiece segmentation mask to be inputted ROI classifier
In, initial defect classification and rough defect bounding box are obtained, then obtains final mask knot by dividing mask network
Fruit.
The beneficial effects of the present invention are: workpiece inspection method of the invention, is able to achieve while being classified to workpiece, defect
Detection, the flexibility suitable for multi items, small lot, the frequent workpiece of product renewing detect.
Detailed description of the invention
Fig. 1 is a kind of flow chart of the workpiece inspection method based on deep learning network of the present invention;
Fig. 2 is a kind of image labeling tool box choosing figure of the workpiece inspection method based on deep learning network of the present invention;
Fig. 3 is a kind of workpiece coordinate that the locating module of the workpiece inspection method based on deep learning network obtains of the present invention
Schematic diagram;
Fig. 4 is a kind of flow chart of the detection module of the workpiece inspection method based on deep learning network of the present invention;
Fig. 5 is a kind of initial workpiece classification that the workpiece inspection method based on deep learning network obtains of the present invention, side
Boundary's frame, segmentation mask;
Fig. 6 is the structural representation that network is suggested in a kind of region of the workpiece inspection method based on deep learning network of the present invention
Figure;
Fig. 7 is that a kind of region of the workpiece inspection method based on deep learning network of the present invention suggests that network sweep generates
Anchor point;
Fig. 8 is that a kind of structure of the feature pyramid network of the workpiece inspection method based on deep learning network of the present invention is shown
It is intended to;
Fig. 9 is a kind of structural schematic diagram of the ROI classifier of the workpiece inspection method based on deep learning network of the present invention;
Figure 10 is a kind of segmentation mask netmask process of the workpiece inspection method based on deep learning network of the present invention
Figure;
Figure 11 is a kind of structural schematic diagram of the Workpiece detecting device based on deep learning network of the present invention.
In figure, 1. locating modules, 101. first acquisition modules, 102. labeling modules, 2. identification modules, 201. second acquisitions
Module, 202. judgment modules, 3. detection modules, 301. first obtain module, and 302. second obtain module, and 303. thirds obtain mould
Block, 304. the 4th obtain module.
Specific embodiment
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
A kind of workpiece inspection method based on deep learning network of the present invention, as shown in Figure 1, specifically includes the following steps:
Step 1, collecting sample workpiece image, are labeled sample workpiece image, obtain the position of sample workpiece image
Information;
Sample workpiece image is labeled by image labeling tool (labelImg), which can adopt
It is selected with clicking with frame, since the workpiece shapes that the present invention detects are irregular, so choice box selects, as shown in Figure 2.Then, pass through
LabelImg obtains the pixel coordinate of workpiece, as shown in table 1, i.e. the location information of workpiece, the work including sample workpiece image
Part classification, workpiece height, width and workpiece coordinate y-axis maximum value ymax, y-axis minimum value ymin, x-axis maximum value xmax、x
Axis minimum value xmin, as shown in figure 3, to realize the positioning to workpiece.
Coordinate information corresponding to 1 location of workpiece of table (unit: pixel)
Step 2, acquisition workpiece image to be detected, send workpiece image to be detected into image buffer storage area, judge image buffer storage
Area whether there is workpiece;
Workpiece image to be detected is sent into image buffer storage area, suggests that network is drawn on workpiece image to be detected by region
After suspicious region Suggestion box, workpiece image to be detected is sent into feature pyramid and carries out image segmentation, feature pyramid allows every
A grade of feature is mutually combined with advanced, low-level features, and the feature by merging these different layers reaches forecast image suspicious region
Effect, to obtain the suspicious region of workpiece image to be detected;By the suspicious region and sample workpiece of workpiece image to be detected
Image compares, and judges image buffer storage area with the presence or absence of workpiece.
If there are workpiece on step 3, workpiece image to be detected, defects detection is carried out to workpiece image to be detected, such as Fig. 4 institute
Show;If workpiece is not present on workpiece image to be detected, stop detecting.
Workpiece image to be detected is inputted residual error network by step 3.1, generates p2, p3 by 5 layers of convolution of residual error network,
P4, p5, p6 characteristic image;
Step 3.2, by p2, p3, p4, p5, p6 characteristic image input area suggests that network generates anchor point, centered on anchor point
Grid division traverses characteristic pattern, then splices all traversing results, and then obtain class parameter, logic of class parameter and bounding box
Parameter;By class parameter, logic of class parameter and bounding box parameter by obtaining interest region after non-maxima suppression;
Step 3.3.1, by (workpiece classification, callout box etc.) the location information input area in interest region and sample workpiece image
Domain suggests that the target detection layer of network carries out target detection, initial workpiece classification, bounding box, segmentation mask is obtained, such as Fig. 5 institute
Show.Suggest the structure of network as shown in fig. 6, by using sliding window scan image, to search target that may be present in region
Region.The rectangle being distributed on the region of RPN scanning is referred to as anchor point, as shown in Figure 7.There are nearly 200,000 different sizes on image
With the rectangle anchor point of length-width ratio, coverings as more as possible are intersected on the image.RPN generates sliding window by convolution process,
Can't direct scan image, but scan trunk characteristic pattern, in this way permission RPN effectively reuse the feature of extraction and avoid again
It is multiple to calculate, and final anchor point will be passed to next stage.
Step 3.3.2, the alignment of region of interest domain is carried out after combining workpiece classification, bounding box, segmentation mask and characteristic image,
All there is strong semantic information on all scales using bilinear interpolation building on the candidate region characteristic image of generation
Feature pyramid network passes through the positioning network and segmentation network of feature pyramid network, the shallow-layers information such as combination of edge, angle point
Divide mask with the deep informations, workpiece classification, workpiece bounding box and the workpiece for obtaining workpiece image to be detected such as feature, defects.
Feature pyramid network (also known as feature extractor), as shown in figure 8, feature pyramid network is by bottom-up and top-down
Two path compositions, by generating multilayer feature mapping (Analysis On Multi-scale Features mapping), information quality obtained is special better than tradition
Sign detection pyramid, feature pyramid network promote standard feature by second pyramid of addition and extract pyramidal performance,
Second pyramid can select advanced features from first pyramid and be transmitted on bottom.
Step 3.4 inputs workpiece classification, workpiece bounding box and workpiece segmentation mask in ROI classifier, as shown in figure 9,
Initial defect classification, defect bounding box are obtained, then obtains final mask as a result, completing to examine by dividing mask network
It surveys.
Step 3.4.1, firstly, it is special using the information extraction workpiece of workpiece classification, workpiece bounding box and workpiece segmentation mask
Sign figure, inputs in the layer of the pond ROI after depth process of convolution, is selected workpiece image bounding box to be detected by the pond ROI layer
On the area maps selected to convolution characteristic pattern;Then, output layer is reached by two layers of full articulamentum.Finally, distinguishing from output layer
It obtains classification maximum value and frame returns to get initial defect classification, defect bounding box is arrived.If not deposited on workpiece image to be detected
In defect, then output layer will not output category maximum value and frame recurrence.
Step 3.4.2, segmentation mask netmask process is as shown in Figure 10, with initial defect classification, defect bounding box
As input, then it is divided into two branches, upper branch carries out convolution to the characteristic pattern of 7x7x256 and obtains defect classification and bounding box,
Lower branch carries out convolution twice to 14x14x256 characteristic pattern, then generates 256 feature of 28x 28x using deconvolution operation
Figure increases image and increases pixel to improve resolution ratio, then is connected the fully-connected network of a 28x 28x 80, obtains low point
The segmentation mask of resolution calculates loss function, and it is final to provide that the mask of prediction is finally enlarged into the size of ROI bounding box
Mask result.
A kind of Workpiece detecting device based on deep learning network, as shown in figure 11, comprising:
Locating module is used for collecting sample workpiece image, is labeled to sample workpiece image, obtains sample workpiece image
Location information;
Identification module judges image buffer storage area with the presence or absence of work for workpiece image to be detected to be sent into image buffer storage area
Part;
Detection module carries out defects detection to workpiece image to be detected if there are workpiece on workpiece image to be detected;If to
It detects and workpiece is not present on workpiece image, then stop detecting.
Locating module includes:
First acquisition module is used for collecting sample workpiece image;
Labeling module obtains the location information of sample workpiece image for being labeled to sample workpiece image.
Identification module includes:
Second acquisition module, for acquiring workpiece image to be detected;
Judgment module judges whether image buffer storage area deposits for the workpiece image to be detected to be sent into image buffer storage area
In workpiece.
Detection module includes:
First obtains module, for the workpiece image to be detected to be inputted residual error network, obtains characteristic image;
Second obtains module, for characteristic image input area suggestion network to be obtained class parameter, logic of class parameter
With bounding box parameter, region of interest is obtained after the class parameter, logic of class parameter and bounding box parameter are passed through non-maxima suppression
Domain;
Third obtains module, for the location information input area in the interest region and sample workpiece image to be suggested net
The target detection layer of network carries out target detection, obtains initial workpiece classification, bounding box, segmentation mask, by the workpiece classification,
Bounding box, segmentation mask and characteristic image carry out the alignment of region of interest domain after combining, and obtain feature pyramid network, pass through feature gold
The positioning network and segmentation network of word tower network, obtain workpiece classification, workpiece bounding box and the workpiece point of workpiece image to be detected
Cut mask;
4th obtains module, for by ROI points of input of the accurate workpiece classification, workpiece bounding box and workpiece segmentation mask
In class device, initial defect classification and rough defect bounding box are obtained, final cover then is obtained by segmentation mask network
Code result.
In the above manner, workpiece inspection method of the invention, is able to achieve while being classified to workpiece, defects detection,
Flexibility suitable for multi items, small lot, the frequent workpiece of product renewing detects.
Claims (8)
1. a kind of workpiece inspection method based on deep learning network, which is characterized in that specifically includes the following steps:
Step 1, collecting sample workpiece image, are labeled sample workpiece image, obtain the location information of sample workpiece image;
Step 2, acquisition workpiece image to be detected, send workpiece image to be detected into image buffer storage area, judge that image buffer storage area is
It is no that there are workpiece;
If there are workpiece on step 3, workpiece image to be detected, defects detection is carried out to workpiece image to be detected;If work to be detected
Workpiece is not present on part image, then stops detecting.
2. a kind of workpiece inspection method based on deep learning network as described in claim 1, which is characterized in that step 1 tool
Body are as follows: collecting sample workpiece image is labeled sample workpiece image by image labeling tool, obtains sample workpiece image
Workpiece classification, pixel coordinate.
3. a kind of workpiece inspection method based on deep learning network as described in claim 1, which is characterized in that step 2 tool
Body are as follows: workpiece image to be detected is sent into image buffer storage area, drawing on workpiece image to be detected by region suggestion network can
After doubting region Suggestion box, workpiece image to be detected is sent into feature pyramid and carries out image segmentation, obtains workpiece image to be detected
Suspicious region, the suspicious region of workpiece image to be detected and sample workpiece image are compared, judge that image buffer storage area is
It is no that there are workpiece.
4. a kind of workpiece inspection method based on deep learning network as described in claim 1, which is characterized in that step 3 tool
Body are as follows:
The workpiece image to be detected is inputted residual error network by step 3.1, obtains characteristic image;
Characteristic image input area suggestion network is obtained class parameter, logic of class parameter and bounding box parameter by step 3.2,
By the class parameter, logic of class parameter and bounding box parameter by obtaining interest region after non-maxima suppression;
Step 3.3, the target detection that the location information input area of the interest region and sample workpiece image is suggested to network
Layer carries out target detection, initial workpiece classification, bounding box, segmentation mask is obtained, by the workpiece classification, bounding box, segmentation
Mask and characteristic image carry out the alignment of region of interest domain after combining, and obtain feature pyramid network, pass through feature pyramid network
Network and segmentation network are positioned, workpiece classification, workpiece bounding box and the workpiece segmentation mask of workpiece image to be detected are obtained;
Step 3.4 inputs the accurate workpiece classification, workpiece bounding box and workpiece segmentation mask in ROI classifier, obtains just
Then the defect classification of beginning and rough defect bounding box obtain final mask result by dividing mask network.
5. a kind of Workpiece detecting device based on deep learning network characterized by comprising
Locating module is used for collecting sample workpiece image, is labeled to the sample workpiece image, obtains sample workpiece image
Location information;
Identification module judges image buffer storage area with the presence or absence of workpiece for workpiece image to be detected to be sent into image buffer storage area;
Detection module carries out defects detection to workpiece image to be detected if there are workpiece on workpiece image to be detected;If to be detected
Workpiece is not present on workpiece image, then stops detecting.
6. a kind of Workpiece detecting device based on deep learning network as claimed in claim 5, which is characterized in that locating module
Include:
First acquisition module is used for collecting sample workpiece image;
Labeling module obtains the location information of sample workpiece image for being labeled to the sample workpiece image.
7. a kind of Workpiece detecting device based on deep learning network as claimed in claim 5, which is characterized in that identification module
Include:
Second acquisition module, for acquiring workpiece image to be detected;
Judgment module judges image buffer storage area with the presence or absence of work for the workpiece image to be detected to be sent into image buffer storage area
Part.
8. a kind of Workpiece detecting device based on deep learning network as claimed in claim 5, which is characterized in that detection module
Include:
First obtains module, for the workpiece image to be detected to be inputted residual error network, obtains characteristic image;
Second obtains module, for characteristic image input area suggestion network to be obtained class parameter, logic of class parameter and side
Boundary's frame parameter, by the class parameter, logic of class parameter and bounding box parameter by obtaining interest region after non-maxima suppression;
Third obtains module, for the location information input area in the interest region and sample workpiece image to be suggested network
Target detection layer carries out target detection, initial workpiece classification, bounding box, segmentation mask is obtained, by the workpiece classification, boundary
Frame, segmentation mask and characteristic image carry out the alignment of region of interest domain after combining, and obtain feature pyramid network, pass through feature pyramid
The positioning network and segmentation network of network, workpiece classification, workpiece bounding box and the workpiece segmentation for obtaining workpiece image to be detected are covered
Code;
4th obtains module, for the accurate workpiece classification, workpiece bounding box and workpiece segmentation mask to be inputted ROI classifier
In, initial defect classification and rough defect bounding box are obtained, then obtains final mask knot by dividing mask network
Fruit.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910712712.9A CN110443791B (en) | 2019-08-02 | 2019-08-02 | Workpiece detection method and device based on deep learning network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910712712.9A CN110443791B (en) | 2019-08-02 | 2019-08-02 | Workpiece detection method and device based on deep learning network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110443791A true CN110443791A (en) | 2019-11-12 |
CN110443791B CN110443791B (en) | 2023-04-07 |
Family
ID=68433073
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910712712.9A Active CN110443791B (en) | 2019-08-02 | 2019-08-02 | Workpiece detection method and device based on deep learning network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110443791B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111178424A (en) * | 2019-12-26 | 2020-05-19 | 中国石油大学(北京) | Petrochemical production site safety compliance real-time detection system and method |
CN111402255A (en) * | 2020-04-10 | 2020-07-10 | 山东大学 | Side formwork failure detection method and system based on deep learning |
CN111523540A (en) * | 2020-04-17 | 2020-08-11 | 中南大学 | Metal surface defect detection method based on deep learning |
CN111951232A (en) * | 2020-07-24 | 2020-11-17 | 上海微亿智造科技有限公司 | Metal powder injection molding appearance defect detection method and system |
CN112862770A (en) * | 2021-01-29 | 2021-05-28 | 珠海迪沃航空工程有限公司 | Defect analysis and diagnosis system, method and device based on artificial intelligence |
CN113467914A (en) * | 2021-07-08 | 2021-10-01 | 征图新视(江苏)科技股份有限公司 | Method for improving processing efficiency of image defect detection software |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106053479A (en) * | 2016-07-21 | 2016-10-26 | 湘潭大学 | System for visually detecting workpiece appearance defects based on image processing |
WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
CN107966447A (en) * | 2017-11-14 | 2018-04-27 | 浙江大学 | A kind of Surface Flaw Detection method based on convolutional neural networks |
CN108144873A (en) * | 2017-12-29 | 2018-06-12 | 深圳市智能机器人研究院 | A kind of vision detection system and method |
CN109035232A (en) * | 2018-07-24 | 2018-12-18 | 桂林电子科技大学 | A kind of intermediate plate defect inspection method merged based on deep learning and machine vision |
CN109063708A (en) * | 2018-06-21 | 2018-12-21 | 歌尔股份有限公司 | Industrial picture characteristic recognition method and system based on contours extract |
CN109509172A (en) * | 2018-09-25 | 2019-03-22 | 无锡动视宫原科技有限公司 | A kind of liquid crystal display flaw detection method and system based on deep learning |
CN109859163A (en) * | 2018-12-19 | 2019-06-07 | 重庆邮电大学 | A kind of LCD defect inspection method based on feature pyramid convolutional neural networks |
CN109900711A (en) * | 2019-04-02 | 2019-06-18 | 天津工业大学 | Workpiece, defect detection method based on machine vision |
-
2019
- 2019-08-02 CN CN201910712712.9A patent/CN110443791B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2018000731A1 (en) * | 2016-06-28 | 2018-01-04 | 华南理工大学 | Method for automatically detecting curved surface defect and device thereof |
CN106053479A (en) * | 2016-07-21 | 2016-10-26 | 湘潭大学 | System for visually detecting workpiece appearance defects based on image processing |
CN107966447A (en) * | 2017-11-14 | 2018-04-27 | 浙江大学 | A kind of Surface Flaw Detection method based on convolutional neural networks |
CN108144873A (en) * | 2017-12-29 | 2018-06-12 | 深圳市智能机器人研究院 | A kind of vision detection system and method |
CN109063708A (en) * | 2018-06-21 | 2018-12-21 | 歌尔股份有限公司 | Industrial picture characteristic recognition method and system based on contours extract |
CN109035232A (en) * | 2018-07-24 | 2018-12-18 | 桂林电子科技大学 | A kind of intermediate plate defect inspection method merged based on deep learning and machine vision |
CN109509172A (en) * | 2018-09-25 | 2019-03-22 | 无锡动视宫原科技有限公司 | A kind of liquid crystal display flaw detection method and system based on deep learning |
CN109859163A (en) * | 2018-12-19 | 2019-06-07 | 重庆邮电大学 | A kind of LCD defect inspection method based on feature pyramid convolutional neural networks |
CN109900711A (en) * | 2019-04-02 | 2019-06-18 | 天津工业大学 | Workpiece, defect detection method based on machine vision |
Non-Patent Citations (5)
Title |
---|
FAN YANG等: "Feature Pyramid and Hierarchical Boosting Network for Pavement Crack Detection", 《IEEE TRANSACTIONS ON INTELLIGENT TRANSPORTATION SYSTEMS》 * |
SHENGQI GUAN: "Strip Steel Defect Detection Based on Saliency Map Construction Using Gaussian Pyramid Decomposition", 《ISIJ INTERNATIONAL》 * |
常海涛: "基于Faster R-CNN的工业CT图像缺陷检测研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》 * |
旷可嘉: "深度学习及其在航空发动机缺陷检测中的应用研究", 《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》 * |
李欣遥: "基于深度学习的工件微小缺陷诊断方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111178424A (en) * | 2019-12-26 | 2020-05-19 | 中国石油大学(北京) | Petrochemical production site safety compliance real-time detection system and method |
CN111402255A (en) * | 2020-04-10 | 2020-07-10 | 山东大学 | Side formwork failure detection method and system based on deep learning |
CN111523540A (en) * | 2020-04-17 | 2020-08-11 | 中南大学 | Metal surface defect detection method based on deep learning |
CN111951232A (en) * | 2020-07-24 | 2020-11-17 | 上海微亿智造科技有限公司 | Metal powder injection molding appearance defect detection method and system |
CN112862770A (en) * | 2021-01-29 | 2021-05-28 | 珠海迪沃航空工程有限公司 | Defect analysis and diagnosis system, method and device based on artificial intelligence |
CN112862770B (en) * | 2021-01-29 | 2023-02-14 | 珠海迪沃航空工程有限公司 | Defect analysis and diagnosis system, method and device based on artificial intelligence |
CN113467914A (en) * | 2021-07-08 | 2021-10-01 | 征图新视(江苏)科技股份有限公司 | Method for improving processing efficiency of image defect detection software |
Also Published As
Publication number | Publication date |
---|---|
CN110443791B (en) | 2023-04-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110443791A (en) | A kind of workpiece inspection method and its detection device based on deep learning network | |
CN108520274B (en) | High-reflectivity surface defect detection method based on image processing and neural network classification | |
CN106952258B (en) | A kind of bottle mouth defect detection method based on gradient orientation histogram | |
CN106056118B (en) | A kind of identification method of counting for cell | |
CN109636772A (en) | The defect inspection method on the irregular shape intermetallic composite coating surface based on deep learning | |
Zhou et al. | Automated visual inspection of glass bottle bottom with saliency detection and template matching | |
CN114140679B (en) | Defect fusion method, device, recognition system and storage medium | |
CN108090423B (en) | Depth license plate detection method based on thermodynamic diagram and key point regression | |
CN107389701A (en) | A kind of PCB visual defects automatic checkout system and method based on image | |
CN109584227A (en) | A kind of quality of welding spot detection method and its realization system based on deep learning algorithm of target detection | |
CN108898610A (en) | A kind of object contour extraction method based on mask-RCNN | |
CN111693534B (en) | Surface defect detection method, model training method, device, equipment and medium | |
CN109859207B (en) | Defect detection method of high-density flexible substrate | |
CN109166098A (en) | Work-piece burr detection method based on image procossing | |
CN105844621A (en) | Method for detecting quality of printed matter | |
CN107064170A (en) | One kind detection phone housing profile tolerance defect method | |
CN102998316B (en) | Transparent liquid impurity detection system and detection method thereof | |
CN102496161B (en) | Method for extracting contour of image of printed circuit board (PCB) | |
CN108665458A (en) | Transparent body surface defect is extracted and recognition methods | |
CN103940708B (en) | A kind of Quick Measurement, sophisticated category method of steel holotype state crystal grain | |
CN106290392A (en) | A kind of little micro-bearing surface pitting defects online test method and system thereof | |
Li et al. | Application of multi-scale feature fusion and deep learning in detection of steel strip surface defect | |
KR20220012217A (en) | Machine Learning-Based Classification of Defects in Semiconductor Specimens | |
Wang et al. | Attention-based deep learning for chip-surface-defect detection | |
CN113393426A (en) | Method for detecting surface defects of rolled steel plate |
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 |