CN110473806A - The deep learning identification of photovoltaic cell sorting and control method and device - Google Patents
The deep learning identification of photovoltaic cell sorting and control method and device Download PDFInfo
- Publication number
- CN110473806A CN110473806A CN201910632273.0A CN201910632273A CN110473806A CN 110473806 A CN110473806 A CN 110473806A CN 201910632273 A CN201910632273 A CN 201910632273A CN 110473806 A CN110473806 A CN 110473806A
- Authority
- CN
- China
- Prior art keywords
- deep learning
- defect
- image
- photovoltaic cell
- sorting
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B07—SEPARATING SOLIDS FROM SOLIDS; SORTING
- B07C—POSTAL SORTING; SORTING INDIVIDUAL ARTICLES, OR BULK MATERIAL FIT TO BE SORTED PIECE-MEAL, e.g. BY PICKING
- B07C5/00—Sorting according to a characteristic or feature of the articles or material being sorted, e.g. by control effected by devices which detect or measure such characteristic or feature; Sorting by manually actuated devices, e.g. switches
- B07C5/34—Sorting according to other particular properties
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
- H01L21/00—Processes or apparatus adapted for the manufacture or treatment of semiconductor or solid state devices or of parts thereof
- H01L21/67—Apparatus specially adapted for handling semiconductor or electric solid state devices during manufacture or treatment thereof; Apparatus specially adapted for handling wafers during manufacture or treatment of semiconductor or electric solid state devices or components ; Apparatus not specifically provided for elsewhere
- H01L21/67005—Apparatus not specifically provided for elsewhere
- H01L21/67242—Apparatus for monitoring, sorting or marking
- H01L21/67253—Process monitoring, e.g. flow or thickness monitoring
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
- H01L21/00—Processes or apparatus adapted for the manufacture or treatment of semiconductor or solid state devices or of parts thereof
- H01L21/67—Apparatus specially adapted for handling semiconductor or electric solid state devices during manufacture or treatment thereof; Apparatus specially adapted for handling wafers during manufacture or treatment of semiconductor or electric solid state devices or components ; Apparatus not specifically provided for elsewhere
- H01L21/67005—Apparatus not specifically provided for elsewhere
- H01L21/67242—Apparatus for monitoring, sorting or marking
- H01L21/67271—Sorting devices
-
- H—ELECTRICITY
- H01—ELECTRIC ELEMENTS
- H01L—SEMICONDUCTOR DEVICES NOT COVERED BY CLASS H10
- H01L21/00—Processes or apparatus adapted for the manufacture or treatment of semiconductor or solid state devices or of parts thereof
- H01L21/67—Apparatus specially adapted for handling semiconductor or electric solid state devices during manufacture or treatment thereof; Apparatus specially adapted for handling wafers during manufacture or treatment of semiconductor or electric solid state devices or components ; Apparatus not specifically provided for elsewhere
- H01L21/67005—Apparatus not specifically provided for elsewhere
- H01L21/67242—Apparatus for monitoring, sorting or marking
- H01L21/67288—Monitoring of warpage, curvature, damage, defects or the like
Abstract
The deep learning identification and control method of photovoltaic cell vision sorting provided by the invention, the defect based on deep learning is classified and detection method is integrated with industrial manufacturing process, intelligence with higher.The present apparatus replaces conventional learning algorithms with deep learning, has stronger ability in feature extraction, and particularly with working as, background interference is serious, when conventional method is difficult to out defect.Furthermore deep learning identification can carry out defect characteristic extraction with control device automatically, improve processing speed, reduce artificial dependence.
Description
Technical field
The present invention relates to the invention belongs to solar battery Image Acquisition fields, and in particular to a kind of sorting of photovoltaic cell
Deep learning identification and control method and device
Background technique
In industrial manufacturing process, the crack defect on the surface defect of workpiece such as photovoltaic cell surface；Some photovoltaic electrics
Pond piece surface appearance defects such as dirty, disconnected grid, scratch；Steel strip surface defect etc. has critically important shadow to industrial production efficiency
It rings.The classification of surface defect and detection are to improving technique, improve assembly quality, improve efficiency and steady production has important work
With.
The sorting unit of existing photovoltaic cell, such as a kind of 108010864 A of device CN (photovoltaic cell defect sorting dress
Set and its method for separating) there is the following: (1) small dimensional defects can not be analyzed: in solar battery picture
Under the interference of grid line and lattice, small defect image can not be identified.(2) for defect form, picture quality has strict demand: if
Defect part and patterning or picture quality are low, and Conventional visual can not all detect such defect.(3) production environment
It is required that engineer is manually operated: if production environment changes, or changing for object, require engineer's optimization and set
Fixed or manual operation.(4) processing routine is complicated, and processing speed is slow: Conventional visual detection mode is extracted dependent on manual feature,
High-accuracy relies on high complexity program, and processing speed is caused to reduce.
Therefore, the systems approach for needing a kind of manufacturing defect identification can also be detected accurately small even if in the case where background is complicated
Dimensional defects can provide higher precision and the degree of automation in all kinds of defects detections, can also be with for high-resolution pictures
Realize rapid batch detection.
Summary of the invention
In view of this, deep learning identification and control method and device the present invention provides photovoltaic cell sorting, solve
Defect existing in the prior art, concrete scheme are as follows:
A kind of the deep learning identification and control method of the sorting of photovoltaic cell vision, this method comprises: following steps:
S1: Image Acquisition
S1-1 acquires image
Whether S1-2 judgement acquired image is more than 500, if it exceeds continue below step, if not above,
It is re-introduced into S1-1 step；
S1-3 stops acquisition image, and acquired image is saved to corresponding defect file and is pressed from both sides, by training set and verifying
Collection carries out artificial separation and forming label；
S2: depth school
S2-1 loads pre-training model；
S2-2 initializes Faster R-CNN model parameter, initiation parameter content include all weighted values,
Bias, batch Normalized Scale factor values, are arranged the initial learning rate of network；
The training set that the step S1-3 is obtained and verifying collection are input to S2-2 initialization according to its label substance by S2-3
Model in be trained, obtain surface defect deep learning model；
S3: deep learning executes
The real-time collecting test image of S3-1；
The surface defect deep learning model that S3-2 calls the step S2-3 to obtain, to the picture of the step S3-1 into
Row test；
S3-3 obtains real-time detection result: exact position and its confidence level including defect classification and positioning defect；
S4: robot sorting
The obtained real-time detection result of the step S3-3 is fed back to robot by S4-1, executes sorting by robot；
Specifically, the training set and the ratio integrated of verifying is 4:1 in the step S1-3.
Specifically, the mode for acquiring image in the step S3-1 can be continuous acquisition or individual acquisition.
Specifically, the defect file folder is EL defect file folder or open defect file.
Specifically, described device include image capture module, deep learning training module, deep learning execution module and
Robot communication module.
The deep learning identification and control method of photovoltaic cell vision sorting provided by the invention, will be based on deep learning
Defect classification and detection method are integrated with industrial manufacturing process, intelligence with higher.The present apparatus is with deep learning come generation
For conventional learning algorithms, there is stronger ability in feature extraction, background interference is serious, and conventional method is difficult to particularly with working as
Out when defect.Furthermore deep learning identification can carry out defect characteristic extraction with control device automatically, improve processing speed, reduce people
Work dependence.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is required attached drawing in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only the present invention
Some embodiments for those of ordinary skill in the art without creative efforts, can also basis
These attached drawings obtain other attached drawings.
Fig. 1 is flow chart of the method for the present invention；
Fig. 2 is the deep learning training surface chart of apparatus of the present invention；
Fig. 3 is the deep learning real-time monitoring surface chart of apparatus of the present invention；
Fig. 4 is the flow diagram of the method Faster R-CNN in positioning target defect region used；
Fig. 5 is the logic diagram for acquiring image；
Fig. 6 is the logic diagram of real-time testing；
Fig. 7 is defects detection initial pictures and detection result image comparison diagram of the invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
Fig. 1-7 is respectively flow chart of the method for the present invention, the deep learning of apparatus of the present invention training surface chart, present invention dress
The deep learning real-time monitoring surface chart set, the flow diagram of the method FasterR-CNN in positioning target defect region used, is adopted
Collect the logic diagram of image, the logic diagram of real-time testing and defects detection initial pictures of the invention and detection result image
Comparison diagram, referring to Fig.1 shown in -7, the deep learning identification and controlling party of a kind of photovoltaic cell vision sorting is claimed in the present invention
Method, this method comprises: following steps:
S1: Image Acquisition
S1-1 acquires image
Whether S1-2 judgement acquired image is more than 500, if it exceeds continue below step, if not above,
It is re-introduced into S1-1 step；
S1-3 stops acquisition image, and acquired image is saved to corresponding defect file and is pressed from both sides, by training set and verifying
Collection carries out artificial separation and forming label；
Wherein, the production of label is executed referring to following table
Described image acquires for EL defect and component defect, and triggering gray scale camera carries out under near infrared light
Acquisition, collecting image size is 1024 × 1024；For open defect, triggering color camera is acquired, and is obtained
Image size be 1868 × 1868.The forming label is manually to be labeled defect area using LabelImg, described
Artificial separation needs manually to separate acquired image with the ratio of training sample set and verifying sample set 4:1.
S2: depth school
S2-1 loads pre-training model；
S2-2 initializes Faster R-CNN model parameter, initiation parameter content include all weighted values,
Bias, batch Normalized Scale factor values, are arranged the initial learning rate of network；
The training set that the step S1-3 is obtained and verifying collection are input to S2-2 initialization according to its label substance by S2-3
Model in be trained, obtain surface defect deep learning model；
Learning rate is set as 0.001, and the descending factors of learning rate are 0.1, maximum number of iterations 30000, regular terms
Weight decays to 0.0005, and the size of batch processing picture number is 256.
The Faster R-CNN proposes a kind of method of effective position target area: specifically, as shown in figure 4, (1)
The characteristic pattern of candidate original image image is extracted using " convolution+activation+pond " layer on one group of basis first.This feature figure is shared use
In subsequent sections candidate network RPN (Region Proposal Network) layer and full connection (fully connection) layer.
(2) region candidate network RPN (Region Proposal Network) layer judges that anchor point (anchors) belongs to by softmax
Prospect (foreground) or background (background) recycle bounding box to return amendment anchor point acquisition accurate
Proposals ultimately produces region candidate image block.(3) target area pond (Roi Pooling) layer collects the characteristic pattern of input
With candidate target area, the characteristic pattern of target area is extracted after these comprehensive information, subsequent full articulamentum is sent into and determines target
Classification.(4) target classification (Classification).The classification of target area is calculated using target area characteristic pattern, while again
Secondary bounding box, which returns, obtains the final exact position of detection block.
S3: deep learning executes
The real-time collecting test image of S3-1；
The surface defect deep learning model that S3-2 calls the step S2-3 to obtain, to the picture of the step S3-1 into
Row test；
S3-3 obtains real-time detection result: exact position and its confidence level including defect classification and positioning defect；
Specifically, the second width figure of Fig. 7 is testing result as the first width of Fig. 7 figure show a kind of original image of disconnected grid defect,
Defective locations are come out by box precise marking, and side number represents the confidence level of testing result, and maximum value 1, digital value is got over
It is more accurate that Gao represents testing result.
S4: robot sorting
The obtained real-time detection result of the step S3-3 is fed back to robot by S4-1, executes sorting by robot, tool
Body, real-time detection result result is sent to robot controller by ICP/IP protocol or serial communication and executes sorting.
Specifically, the mode for acquiring image in the step S3-1 can be continuous acquisition or individual acquisition.
A kind of dress of deep learning identification and control method using the sorting of photovoltaic cell vision is also claimed in the present invention
It sets, described device includes image capture module, deep learning training module, deep learning execution module and robot communication mould
Block.
This method experiment is completed under the platform of Ubuntu16.04, is programmed and is realized using TensorFlow, what training used
Computer CPU is Intel Core i7 series, inside saves as 64GB, and video card is double GTX2080TI video cards,
This method is specifically realized by following mode of operation:
It acquires image and carries out model training: if Fig. 5 is acquisition image logic block diagram, specifically, selection Fig. 2 deep learning instruction
Practice defect training type in interface, image will be automatically saved in selected defect training set file, and click the interface Fig. 2
In start acquire button, i.e., triggering camera enter state to be collected.Since training needs more sample size, it is defaulted as
Continuous acquisition observes total acquisition number, and when collecting quantity reaches requirement (> 500), click stops acquisition button, is manually divided
Training set is separated with verifying collection with 4:1 ratio, defect area is manually then carried out picture frame mark using LabelImg and is added by choosing
Add defect kind label.Frame number reset button is clicked before carrying out next type acquisition.
Start to train: clicking and start to train button in Fig. 2 deep learning training interface, this is the deep learning in interface
Training module interface supports the deep learning model of the platform trainings such as tensorflow, cafe, matlab and Baidu, the depth
Learning training module is spent by load pre-training model Faster R-CNN, and the initialization including parameter initializes all weights
Value, bias, batch Normalized Scale factor values, are arranged the initial learning rate of network；Then by obtained training set and verifying
Collection and its corresponding label are input in pre-training model and are trained, and finally obtain corresponding surface defect deep learning mould
Type.In training process, training matched curve is shown in training curve region.
Real-time testing is carried out, if Fig. 6 is real-time testing logic diagram, is executed in interface specifically, clicking Fig. 3 deep learning
Start to acquire button i.e. and can trigger camera enter state to be collected and acquire picture in real time for deep learning execution module, select simultaneously
Trigger control mode (triggering control button, soft trigger button, rising edge trigger button, failing edge trigger button) is selected, wherein touching
It sends out for camera continuous acquisition after close button selects, wherein rising edge failing edge triggering mode is hard triggering mode, that is, works as camera
When the sensor of lower section detects workpiece, PLC level signal is risen or fallen along signal triggering camera acquisition image, wherein soft
Triggering acquires beat for the artificial camera by software set；Simultaneously by selection defect kind, i.e., (EL is lacked thread control mode
Sunken or open defect), the surface defect deep learning model that training obtains is loaded according to actually detected content.Pass through real-time testing
Real-time collected picture can be input in the surface defect deep learning model and surface defect is measured in real time,
Real-time detection is obtained as a result, original image and testing result figure are shown respectively in the initial pictures region and detection image region of Fig. 3,
The comparison of two figures is shown simultaneously in Fig. 7, is finally sent result to robot communication module and is executed sorting.
Test result of the invention shows that the present invention has arranged in pairs or groups deep learning intelligent vision software, the depth on interface
Module interface is practised, the deep learning model of the platform trainings such as tensorflow, cafe, matlab and Baidu is supported, with depth
Practising algoritic module replaces traditional algorithm module to can satisfy high speed, complicated vision-based detection task.It is provided simultaneously with various communications
Interface, may be implemented the flexible configuration with industry spot various cameras and robot, surface EL defect and open defect part
Detection accuracy has also reached 95% or more.
The embodiment of the present invention is described with above attached drawing, but the invention is not limited to above-mentioned specific
Embodiment, the above mentioned embodiment is only schematical, rather than restrictive, those skilled in the art
Under the inspiration of the present invention, without breaking away from the scope protected by the purposes and claims of the present invention, it can also make very much
Form, all of these belong to the protection of the present invention.
Claims (5)
1. the deep learning identification and control method of a kind of photovoltaic cell vision sorting, which is characterized in that this method comprises: following
Step:
S1: Image Acquisition
S1-1 acquires image
S1-2 determines acquired image whether more than 500, if it exceeds continuing below step, if not above again
It enters in S1-1 step；
S1-3 stops acquisition image, and acquired image is saved to corresponding defect file and is pressed from both sides, by training set and verifying collect into
Row artificial separation and forming label；
S2: depth school
S2-1 loads pre-training model；
S2-2 initializes Faster R-CNN model parameter, and initiation parameter content includes all weighted values, biasing
Value, batch Normalized Scale factor values, are arranged the initial learning rate of network；
The training set that the step S1-3 is obtained and verifying collection are input to the mould of S2-2 initialization by S2-3 according to its label substance
It is trained in type, obtains surface defect deep learning model；
S3: deep learning executes
The real-time collecting test image of S3-1；
The surface defect deep learning model that S3-2 calls the step S2-3 to obtain, surveys the picture of the step S3-1
Examination；
S3-3 obtains real-time detection result: exact position and its confidence level including defect classification and positioning defect；
S4: robot sorting
The obtained real-time detection result of the step S3-3 is fed back to robot by S4-1, executes sorting by robot.
2. the deep learning identification and control method of a kind of photovoltaic cell vision sorting according to claim 1, feature
Be: in the step S1-3, the training set and the ratio integrated of verifying is 4:1.
3. the deep learning identification and control method of a kind of photovoltaic cell vision sorting according to claim 1, feature
Be: the mode that image is acquired in the step S3-1 can be continuous acquisition or individual acquisition.
4. the deep learning identification and controlling party of a kind of photovoltaic cell vision sorting according to claim 1-3
Method, it is characterised in that: the defect file folder is EL defect file folder or open defect file.
5. a kind of device of deep learning identification and control method using the sorting of photovoltaic cell vision, it is characterised in that: described
Device includes image capture module, deep learning training module, deep learning execution module and robot communication module.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910632273.0A CN110473806A (en) | 2019-07-13 | 2019-07-13 | The deep learning identification of photovoltaic cell sorting and control method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910632273.0A CN110473806A (en) | 2019-07-13 | 2019-07-13 | The deep learning identification of photovoltaic cell sorting and control method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110473806A true CN110473806A (en) | 2019-11-19 |
Family
ID=68508560
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910632273.0A Pending CN110473806A (en) | 2019-07-13 | 2019-07-13 | The deep learning identification of photovoltaic cell sorting and control method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110473806A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111054638A (en) * | 2019-11-27 | 2020-04-24 | 奥特斯科技(重庆)有限公司 | Method for manufacturing a component carrier and device for handling panels during manufacturing |
| CN111507325A (en) * | 2020-03-16 | 2020-08-07 | 重庆大学 | Industrial visual OCR recognition system and method based on deep learning |
| CN111830048A (en) * | 2020-07-17 | 2020-10-27 | 苏州凌创电子系统有限公司 | Automobile fuel spray nozzle defect detection equipment based on deep learning and detection method thereof |
| CN114627122A (en) * | 2022-05-16 | 2022-06-14 | 北京东方国信科技股份有限公司 | Defect detection method and device |
| CN115088125A (en) * | 2020-03-20 | 2022-09-20 | 舍弗勒技术股份两合公司 | Method and inspection device for inspecting bipolar plates of electrochemical cells, in particular fuel cells |
| US11935221B2 (en) | 2019-11-27 | 2024-03-19 | AT&S (Chongqing) Company Limited | User interface for judgment concerning quality classification of displayed arrays of component carriers |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006023130A (en) * | 2004-07-06 | 2006-01-26 | Daido Steel Co Ltd | End face position detector for semifinished product of steel |
| CN106873566A (en) * | 2017-03-14 | 2017-06-20 | 东北大学 | A kind of unmanned logistic car based on deep learning |
| CN107831173A (en) * | 2017-10-17 | 2018-03-23 | 哈尔滨工业大学（威海） | Photovoltaic component defect detection method and system |
-
2019
- 2019-07-13 CN CN201910632273.0A patent/CN110473806A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006023130A (en) * | 2004-07-06 | 2006-01-26 | Daido Steel Co Ltd | End face position detector for semifinished product of steel |
| CN106873566A (en) * | 2017-03-14 | 2017-06-20 | 东北大学 | A kind of unmanned logistic car based on deep learning |
| CN107831173A (en) * | 2017-10-17 | 2018-03-23 | 哈尔滨工业大学（威海） | Photovoltaic component defect detection method and system |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111054638A (en) * | 2019-11-27 | 2020-04-24 | 奥特斯科技(重庆)有限公司 | Method for manufacturing a component carrier and device for handling panels during manufacturing |
| US11935221B2 (en) | 2019-11-27 | 2024-03-19 | AT&S (Chongqing) Company Limited | User interface for judgment concerning quality classification of displayed arrays of component carriers |
| CN111507325A (en) * | 2020-03-16 | 2020-08-07 | 重庆大学 | Industrial visual OCR recognition system and method based on deep learning |
| CN111507325B (en) * | 2020-03-16 | 2023-04-07 | 重庆大学 | Industrial visual OCR recognition system and method based on deep learning |
| CN115088125A (en) * | 2020-03-20 | 2022-09-20 | 舍弗勒技术股份两合公司 | Method and inspection device for inspecting bipolar plates of electrochemical cells, in particular fuel cells |
| CN111830048A (en) * | 2020-07-17 | 2020-10-27 | 苏州凌创电子系统有限公司 | Automobile fuel spray nozzle defect detection equipment based on deep learning and detection method thereof |
| CN114627122A (en) * | 2022-05-16 | 2022-06-14 | 北京东方国信科技股份有限公司 | Defect detection method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110473806A (en) | The deep learning identification of photovoltaic cell sorting and control method and device | |
| CN108280856A (en) | The unknown object that network model is inputted based on mixed information captures position and orientation estimation method | |
| CN101464946B (en) | Detection method based on head identification and tracking characteristics | |
| CN109584227A (en) | A kind of quality of welding spot detection method and its realization system based on deep learning algorithm of target detection | |
| CN108710919A (en) | A kind of crack automation delineation method based on multi-scale feature fusion deep learning | |
| CN109800697B (en) | Transformer target detection and appearance defect identification method based on VGG-net style migration | |
| CN108711148B (en) | Tire defect intelligent detection method based on deep learning | |
| CN104574389A (en) | Battery piece chromatism selection control method based on color machine vision | |
| CN108961235A (en) | A kind of disordered insulator recognition methods based on YOLOv3 network and particle filter algorithm | |
| CN106127780A (en) | A kind of curved surface defect automatic testing method and device thereof | |
| CN103324937A (en) | Method and device for labeling targets | |
| CN109544522A (en) | A kind of Surface Defects in Steel Plate detection method and system | |
| CN107328787A (en) | A kind of metal plate and belt surface defects detection system based on depth convolutional neural networks | |
| CN106934333A (en) | A kind of gesture identification method and system | |
| CN110490842A (en) | A kind of steel strip surface defect detection method based on deep learning | |
| CN110302981B (en) | Solid waste sorting online grabbing method and system | |
| CN110096980A (en) | Character machining identifying system | |
| CN110909660A (en) | Plastic bottle detection and positioning method based on target detection | |
| CN110443791A (en) | A kind of workpiece inspection method and its detection device based on deep learning network | |
| CN208092786U (en) | A kind of the System of Sorting Components based on convolutional neural networks by depth | |
| CN109997199A (en) | Tuberculosis inspection method based on deep learning | |
| CN114155610B (en) | Panel assembly key action identification method based on upper half body posture estimation | |
| CN103455826A (en) | Efficient matching kernel body detection method based on rapid robustness characteristics | |
| CN101908150B (en) | Human body detection method | |
| CN109584208B (en) | Inspection method for intelligent identification model of industrial structure defects |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20191119 |
|
| RJ01 | Rejection of invention patent application after publication |