CN105427237B - A kind of steel mesh image registration of large format optical measuring system and detection method - Google Patents
A kind of steel mesh image registration of large format optical measuring system and detection method Download PDFInfo
- Publication number
- CN105427237B CN105427237B CN201510833458.XA CN201510833458A CN105427237B CN 105427237 B CN105427237 B CN 105427237B CN 201510833458 A CN201510833458 A CN 201510833458A CN 105427237 B CN105427237 B CN 105427237B
- Authority
- CN
- China
- Prior art keywords
- image
- steel mesh
- shooting
- vertex
- triangle
- 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.)
- Active
Links
- 229910000831 Steel Inorganic materials 0.000 title claims abstract description 60
- 239000010959 steel Substances 0.000 title claims abstract description 60
- 238000001514 detection method Methods 0.000 title claims abstract description 34
- 230000003287 optical effect Effects 0.000 title claims abstract description 22
- 238000000034 method Methods 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000013461 design Methods 0.000 claims abstract description 8
- 238000013519 translation Methods 0.000 claims abstract description 3
- 238000000605 extraction Methods 0.000 claims description 8
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000007547 defect Effects 0.000 claims description 4
- 239000000203 mixture Substances 0.000 claims description 4
- 230000010339 dilation Effects 0.000 claims description 3
- 210000002837 heart atrium Anatomy 0.000 claims description 2
- 238000012216 screening Methods 0.000 claims description 2
- 230000003628 erosive effect Effects 0.000 claims 1
- 238000012360 testing method Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 3
- 238000005259 measurement Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000003486 chemical etching Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 239000010903 husk Substances 0.000 description 1
- 238000003698 laser cutting Methods 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G06T3/14—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/48—Extraction of image or video features by mapping characteristic values of the pattern into a parameter space, e.g. Hough transformation
Abstract
A kind of steel mesh image registration of large format optical measuring system and detection method, are related to image processing field, solve the problems, such as conventional images registration with detection method existing for it is registering it is low with accuracy of detection, difficulty is big.Priori design parameter of the invention by the use of large format optical measuring system is as prior information come the algorithm of guide image registration, creative is introduced into method for recognising star map in steel mesh method for registering images, three the calculative scaling of steel mesh image registration, rotation and translation crucial parameters are used into different method separate computations, complicated steel mesh image registration problem is simplified.The present invention takes full advantage of prior information of the design parameter as image detection of large format optical measuring system, reduce the registration and the difficulty of detection of originally extremely complex steel mesh image, the method of importance in star map recognition is introduced into steel mesh method for registering images first, with being mutually authenticated using the method for registering of Hough transform and priori putting angle, the success rate of registration is effectively guaranteed.
Description
Technical field
The present invention relates to technical field of image processing, and in particular to a kind of steel mesh image of large format optical measuring system is matched somebody with somebody
Accurate and detection method.
Background technology
In recent years, electronics industry obtains development at full speed, and pcb board circuit level and complexity are also higher and higher, patch
The proportion of piece element on circuit boards is also increasing.Surface mounting technology has become skill most general in current Denso technology
Art, and Solder-Paste Printing is one of critical process of surface mounting technology, its quality directly affects the quality and effect of pcb board assembling
Rate.The making of printed steel mesh mainly has 3 kinds of techniques at present：Chemical etching, laser cutting and electrotyping process.With subtracting for spacing
Small, the size of perforate is also less and less on steel mesh, at the same the size of perforate is required it is also higher and higher, this cause at present mainly by
Artificial vision's detection means is increasingly difficult to meet the requirements to detect the quality of steel mesh.For large-sized steel mesh (such as 900mm ×
800mm), the number of openings is up to tens of thousands of, and artificial vision's detection is it is difficult to ensure that the accuracy of detection, while artificial vision detects
It is difficult to provide accurate measurement data.Current most manufacturer is still using vernier caliper, micrometer and microscope etc.
Instrument carries out size detection to printed steel mesh, this will consume the substantial amounts of time, and real-time and correctness are all difficult to ensure that.Can be pre-
See, it is steel mesh detection technique hair that the Machine Vision Detection based on Digital Image Processing and computer technology, which replaces artificial detection,
The inexorable trend of exhibition.
Need to be imaged it using large format optical measuring system for larger-size steel mesh, the large format optical measurement
The core technology of system is registration and the detection of the full field of view image of big visual field multiaspect battle array CMOS splicings synthesis.Its steel mesh shot
The characteristics of image is visual field is big, pixel number is more, repetitive pattern information is more, grain details information is few etc., this to the registering of image with
Detection all brings sizable difficulty.In addition, the designer of existing image registration and detection method is typically only for figure
As itself is handled, do not handled with reference to information such as the priori design parameters of optical measuring system fully, this causes
Registration and accuracy of detection will reduce.
The content of the invention
In order to solve existing for existing image registration and detection method it is registering it is low with accuracy of detection, difficulty is big asks
Topic, the present invention propose steel mesh image registration and the detection method of a kind of large format optical measuring system, are effectively combined optics
The prior information of measuring system carries out registration and the detection of big view field image.
The present invention is as follows for technical scheme applied to solve the technical problem：
A kind of steel mesh image registration of large format optical measuring system of the present invention and detection method, comprise the following steps：
Step 1: according to the normative document of the design parameter of optical measuring system and steel mesh, image registration and detection are determined
Prior information：It is maximum that the vector file of full filed steel mesh image, shooting image correspond to the image planes distance of steel mesh, steel mesh per pixel
The band of position of full filed corresponding to tilt angle theta, every CMOS shooting area；
Step 2: the current frame image of shooting is carried out binary conversion treatment obtains image 1, N group areas before retaining image 1
Maximum connected region obtains image 2；
Step 3: image 2 is obtained into image 3 using canny operator extractions edge；
Step 4: Hough transform, all rotations obtained are carried out with 0.01 ° of precision in the range of ± θ to image 3
Rotation angle α 1 of the amount of maximum weight as steel mesh image in angle；
Step 5: to each closed communicating extracted region barycenter of image 2, and with the barycenter of all extractions in image 2
For vertex, all triangles of permutation and combination, and make the screening for meeting following condition：Three angles of composition triangle are greater than most
Low-angle thresholding threshold_1, composition three sides of a triangle are greater than the long thresholding threshold_2 of minimum edge, record
Meet all triangle information representings of both thresholdings；
Step 6: to each closed communicating extracted region barycenter of image 1, and the N number of barycenter position of whole that image 2 is extracted
Put and be mapped in image 1；In the image 1 centered on this N number of barycenter, radius searches for the quantity of other barycenter in the range of being R,
And it recorded in the corresponding vertex information of each triangle in step 5, i.e., the number of each vertex neighborhood barycenter；
Step 7: all triangle information representings of step 5 are ranked up according to the size order of triangle girth, to every
Order of three vertex of a triangle in array is recorded is ranked up according to following priority：Angle is big where vertex
Small, vertex opposite side length, the area of vertex correspondence connected region, the number of vertex correspondence neighborhood target；It is excellent according to this four
After first level record, the positional information of the vertex in the picture is finally recorded, thus obtains shooting asterism information table；
Step 8: correspond to the image planes distance of steel mesh per pixel according to shooting image to the vector file of full filed steel mesh image
Scaling output is that the standard picture of scaling relation is not present with shooting image, and is clapped according to the CMOS of shooting current frame image
The band of position for taking the photograph full filed corresponding to region provides standard picture corresponding to present frame；
Step 9: the processing of the standard picture repeat step two, three, five, six, seven to present frame, obtains standard asterism letter
Cease table；
Step 10: setting angle of aspect threshold value threshold_3, search neighborhood destination number threshold value threshold_4, search
Vertex connected region area threshold threshold_5 and the anglec of rotation compare threshold alpha；According to sequencing in shoot asterism information table
Order chooses the triangle of candidate successively, is searched in standard asterism information table and meets threshold_3, threshold_ at the same time
4th, the triangle of threshold_5, if the triangle can be found, is closed using the position on the Atria vertex of two pairings
System, calculates the rotation angle α 2 of two images, and meet | α 1- α 2 | ＜ α then search for success, otherwise search failure；
Step 11: if step 10 is searched for successfully, the rotation angle of current frame image and standard picture is α 2, if step
Ten search failures, then the rotation angle of current frame image and standard picture is α 1, by the current frame image of shooting according to calculating
Rotation angle and standard picture angular alignment；
Step 12: calculate the translational movement of two images after step 11 is handled, and by shooting image and standard
Image is perfectly aligned；
Step 13: morphology pretreatment is carried out to the shooting image after step 12 processing；
Step 14: XOR operation is carried out successively with standard picture to the shooting image after step 13 processing, corrosion is transported
Calculation and dilation operation, finally obtain the location and shape of all defect.
Further, in step 12, two width after step 11 is handled are calculated using normalized crosscorrelation method
The translational movement of image.
Further, in step 13, the morphology pretreatment includes：Eliminate spine, the isolated stain of filling, remove orphan
Vertical bright spot.
The beneficial effects of the invention are as follows：
1st, the priori design parameter of the invention for taking full advantage of large format optical measuring system is instructed as prior information
The algorithm of image registration, this is mainly reflected in the following aspects：First, the image planes of steel mesh are corresponded to per pixel using shooting image
The vector file of distance and full filed steel mesh image can solve the problems, such as the scaling of image registration, and the solution of this problem
Certainly be conducive to the precision of subsequent registration algorithm；Second, the placing direction of steel mesh can conduct with the angle of image device maximum
Hough transform calculates the scope of angle, so that the calculation amount of Hough transform is greatly reduced, due to the side of the rectangle in steel mesh
Edge is mostly laid parallel, therefore the weight maximum of Hough transform calculating is exactly the angle of monoblock steel mesh and direction of transfer；The
Three, the mark corresponding to present frame is provided according to the band of position of full filed corresponding to the CMOS shooting areas of shooting current frame image
Quasi- image has been greatly reduced the difficulty of image registration so that even if having the figure largely repeated, steel mesh in full filed
The method of image registration can also realize preferable registration.
Method for recognising star map is introduced into steel mesh method for registering images by the 2nd, the invention, each by steel mesh connects
Logical region is considered as a star, and area is considered as magnitude, so a complicated bianry image registration problems can be converted into one
The problem of star catalogue is searched.But since the position of steel mesh figure has the characteristics that proper alignment, this with traditional star chart problem very much not
Equally.Therefore the barycenter structure star chart of the figure of extraction top n maximum, that is, calculation amount is reduced, and computational accuracy is added, and
The information of each asterism neighborhood is further utilized, registering accuracy can be increased.
3rd, the present invention uses the crucial parameter of the calculative scaling of steel mesh image registration, rotation and translation three different
Method separate computations, complicated steel mesh image registration problem is simplified.Hough transform and star chart are utilized when calculating rotation parameter
Registering two methods are mutually authenticated, and ensure that the accuracy of calculating, reduce the probability of registration failure, are a kind of highly effective and new
The image registration of grain husk and detection method.
4th, the present invention takes full advantage of prior information of the design parameter as image detection of large format optical measuring system,
The registration and the difficulty of detection of originally extremely complex steel mesh image are reduced, the method for importance in star map recognition is introduced into steel mesh figure first
As in method for registering, with being mutually authenticated using the method for registering of Hough transform and priori putting angle, being effectively guaranteed registration
Success rate.
Brief description of the drawings
Fig. 1 is the schematic diagram of 13 CMOS arrangements of large format optical measuring system.
Fig. 2 is that the present frame steel mesh image of shooting passes through the image 1 after binary conversion treatment.
Fig. 3 is the image 2 after the connected region for retaining image 1 preceding 50 groups of areas maximum.
Fig. 4 is using the image 3 after canny operator extractions edge to image 2.
Fig. 5 is the standard picture of current frame position.
Fig. 6 is the image carried out to the shooting image after step 12 processing after morphology pretreatment.
Fig. 7 is the image for the location and shape for obtaining all defect.
Fig. 8 is the flow chart of the steel mesh image registration and detection method of a kind of large format optical measuring system of the present invention.
Embodiment
The present invention is described in further detail below in conjunction with attached drawing.
As shown in figure 8, steel mesh image registration and the detection method of a kind of large format optical measuring system of the present invention, the party
Method is realized by following steps：
Step 1: according to the normative document of the design parameter of optical measuring system and steel mesh, image registration and detection are determined
Prior information：The vector file of full filed steel mesh image, is 900mm × 800mm for maximum steel mesh size；Shooting image is every
Pixel corresponds to the steel mesh region of 15 μm of 15 μ m, steel mesh maximum tilt angle is θ=3 °；13 CMOS are according to product word shown in Fig. 1
Shape arranges, horizontal 13 CMOS, and longitudinal direction, which exposes 30 frames, can cover full filed, therefore needs to divide full filed standard steel online article part
For 390 shooting areas.
Step 2: the current frame image of shooting is carried out binary conversion treatment obtains image 1, as shown in Figure 2.Image 1 is protected
The connected region of preceding 50 groups of areas maximum is stayed to obtain image 2, as shown in Figure 3.
Step 3: image 2 is obtained into image 3, as shown in Figure 4 using canny operator extractions edge.
Step 4: Hough transform, all rotations obtained are carried out with 0.01 ° of precision in the range of ± 3 ° to image 3
Rotation angle α 1 of the amount of maximum weight as steel mesh image in angle, is as a result α 1=1.99 °.
Step 5: to each closed communicating extracted region barycenter of image 2, and with the barycenter of all extractions in image 2
For vertex, all possible triangle of permutation and combination, three angles for forming triangle are greater than 30 °, form the three of triangle
Bar side is greater than 1000 pixels.All triangle information representings for meeting both thresholdings are recorded, amount to 3748.
Step 6: to each closed communicating extracted region barycenter of image 1, and all 50 barycenter that image 2 is extracted
Position is mapped in image 1.In the image 1 centered on this 50 barycenter, radius be 300 pixels in the range of search for other
The quantity of barycenter, and recorded in the corresponding vertex information of each triangle of step 5, i.e., each vertex neighborhood barycenter
Number.
Step 7: 3748 triangle information representings of step 5 are ranked up according to the size order of triangle girth, it is right
Order of three vertex of each triangle in array is recorded is ranked up according to following priority：Angle is big where vertex
Small, vertex opposite side length, the area of vertex correspondence connected region, the number of vertex correspondence neighborhood target.It is excellent according to this four
After first level record, the positional information of the vertex in the picture is finally recorded, thus obtains shooting asterism information table.
Step 8: standard picture is scaled per pixel 15 μm according to 15 μ m to the vector file of full filed steel mesh image, and
Provided according to the band of position of full filed corresponding to the CMOS shooting areas of shooting current frame image corresponding to current frame position
Standard picture, as shown in Figure 5.
Step 9: the processing of the standard picture repeat step two, three, five, six, seven to present frame, obtains standard asterism letter
Cease table.
Step 10: 0.01 ° of setting angle of aspect threshold value, search neighborhood destination number threshold value 1, corresponding vertex connected region face
Product 5 pixels of threshold value and the anglec of rotation compare 0.02 ° of threshold value.From the beginning opened according to the order sequenced before in asterism information table is shot
Beginning chooses the triangle of candidate successively, is searched in standard asterism information table and meets that angle error is less than 0.01 °, neighborhood target at the same time
Number is identical, corresponding vertex connected region area error is less than the triangles of 5 pixels.Utilize the top of the pairing triangle found
The rotation angle that two images are calculated in point position is 2.00 °, differs 0.01 ° with 1.99 ° of step 4 calculating, is less than
0.02 ° of the anglec of rotation compares threshold value, it is believed that searches for successfully.
Step 11: the present frame of shooting is rotated into 2 ° and standard picture angular alignment.
Step 12: calculate the two images after step 11 is handled using normalized crosscorrelation method (NCC)
Translational movement, and shooting image is perfectly aligned with standard picture.
Step 13: carrying out morphology pretreatment to the shooting image after step 12 processing, mainly include：Eliminate point
Thorn, the isolated stain of filling, remove isolated bright spot, as shown in Figure 6.
Step 14: carrying out XOR operation with standard picture to the shooting image after step 13 processing, then carry out rotten
Computing and dilation operation are lost, finally obtains the location and shape of all defect, as shown in Figure 7.
Claims (3)
1. steel mesh image registration and the detection method of a kind of large format optical measuring system, it is characterised in that comprise the following steps：
Step 1: according to the normative document of the design parameter of optical measuring system and steel mesh, image registration and the elder generation detected are determined
Test information：The vector file of full filed steel mesh image, shooting image correspond to image planes distance, the steel mesh maximum inclination of steel mesh per pixel
The band of position of full filed corresponding to angle, θ, every CMOS shooting area；
Step 2: the current frame image of shooting is carried out binary conversion treatment obtains image 1, N groups area is maximum before retaining image 1
Connected region obtain image 2；
Step 3: image 2 is obtained into image 3 using canny operator extractions edge；
Step 4: Hough transform, all rotation angles obtained are carried out with 0.01 ° of precision in the range of ± θ to image 3
Rotation angle α 1 of the amount of middle maximum weight as steel mesh image；
Step 5: barycenter is extracted to the connected region of each area maximum of image 2, and with the matter of all extractions in image 2
The heart is vertex, all triangles of permutation and combination, and makees the screening for meeting following condition：Three angles of composition triangle are greater than
Minimum angles thresholding threshold_1, composition three sides of a triangle are greater than the long thresholding threshold_2 of minimum edge, remember
Record meets all triangle information representings of both thresholdings；
Step 6: to each closed communicating extracted region barycenter of image 1, and the N number of centroid position of whole that image 2 is extracted reflects
It is mapped in image 1；In the image 1 centered on this N number of barycenter, radius searches for the quantity of other barycenter in the range of being R, and remembers
Record in the corresponding vertex information of each triangle in step 5, i.e., the number of each vertex neighborhood barycenter；
Step 7: all triangle information representings of step 5 are ranked up according to the size order of triangle girth, to each three
Order of the three angular vertex in array is recorded is ranked up according to following priority：The size at angle, top where vertex
The point length of opposite side, the area of vertex correspondence connected region, the number of vertex correspondence neighborhood target；According to this four priority notes
After record, the positional information of the vertex in the picture is finally recorded, thus obtains shooting asterism information table；
Step 8: the image planes of steel mesh are corresponded to per pixel apart from scaling according to shooting image to the vector file of full filed steel mesh image
Export as the standard picture of scaling relation is not present with shooting image, and area is shot according to the CMOS of shooting current frame image
The band of position of full filed corresponding to domain provides the standard picture corresponding to present frame；
Step 9: the processing of the standard picture repeat step two, three, five, six, seven to present frame, obtains standard asterism information table；
Step 10: setting angle of aspect threshold value threshold_3, search neighborhood destination number threshold value threshold_4, search vertex
Connected region area threshold threshold_5 and the anglec of rotation compare threshold alpha；According to the order sequenced in asterism information table is shot
The triangle of candidate is chosen successively, searched in standard asterism information table while meet threshold_3, threshold_4,
The triangle of threshold_5, if the triangle can be found, is closed using the position on the Atria vertex of two pairings
System, calculates the rotation angle α 2 of two images, and meet | α 1- α 2 | ＜ α then search for success, otherwise search failure；
Step 11: if step 10 is searched for successfully, the rotation angle of current frame image and standard picture is α 2, if step 10 is searched
Rope fails, then the rotation angle of current frame image and standard picture is α 1, the rotation by the current frame image of shooting according to calculating
Angle and standard picture angular alignment；
Step 12: calculate the translational movement of two images after step 11 is handled, and by shooting image and standard picture
It is perfectly aligned；
Step 13: morphology pretreatment is carried out to the shooting image after step 12 processing；
Step 14: shooting image and the standard picture after step 13 processing are carried out successively XOR operation, erosion operation and
Dilation operation, finally obtains the location and shape of all defect.
2. steel mesh image registration and the detection method of a kind of large format optical measuring system according to claim 1, it is special
Sign is, in step 12, the translation of the two images after step 11 is handled is calculated using normalized crosscorrelation method
Amount.
3. steel mesh image registration and the detection method of a kind of large format optical measuring system according to claim 1, it is special
Sign is, in step 13, the morphology pretreatment includes：Eliminate spine, the isolated stain of filling, remove isolated bright spot.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510833458.XA CN105427237B (en) | 2015-11-25 | 2015-11-25 | A kind of steel mesh image registration of large format optical measuring system and detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510833458.XA CN105427237B (en) | 2015-11-25 | 2015-11-25 | A kind of steel mesh image registration of large format optical measuring system and detection method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105427237A CN105427237A (en) | 2016-03-23 |
| CN105427237B true CN105427237B (en) | 2018-04-24 |
Family
ID=55505419
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510833458.XA Active CN105427237B (en) | 2015-11-25 | 2015-11-25 | A kind of steel mesh image registration of large format optical measuring system and detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105427237B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108470081B (en) * | 2018-01-17 | 2021-10-12 | 天津大学 | Method for positioning multiple grids of supersonic velocity boundary layer and quickly interpolating flow field |
| CN109785324B (en) * | 2019-02-01 | 2020-11-27 | 佛山市南海区广工大数控装备协同创新研究院 | Large-format PCB positioning method |
| CN110977073B (en) * | 2020-01-03 | 2022-04-01 | 广州市鸿利显示电子有限公司 | Method for brushing tin at fixed points of soldering tin points |
| CN112950680B (en) * | 2021-02-20 | 2022-07-05 | 哈尔滨学院 | Satellite remote sensing image registration method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103901416A (en) * | 2014-03-31 | 2014-07-02 | 西安电子科技大学 | Multichannel clutter suppression method based on robust principal component analysis |
| CN104867104A (en) * | 2015-05-20 | 2015-08-26 | 天津大学 | Method for obtaining anatomical structural atlas for target mouse based on XCT image non-rigid registration |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP3747866B2 (en) * | 2002-03-05 | 2006-02-22 | 日産自動車株式会社 | Image processing apparatus for vehicle |
-
2015
- 2015-11-25 CN CN201510833458.XA patent/CN105427237B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103901416A (en) * | 2014-03-31 | 2014-07-02 | 西安电子科技大学 | Multichannel clutter suppression method based on robust principal component analysis |
| CN104867104A (en) * | 2015-05-20 | 2015-08-26 | 天津大学 | Method for obtaining anatomical structural atlas for target mouse based on XCT image non-rigid registration |
Non-Patent Citations (1)
| Title |
|---|
| 柔性印制电路板自动生产设备关键技术研究;陈安;《中国博士学位论文全文数据库 信息科技辑》;20140515(第5期);第I138-45页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105427237A (en) | 2016-03-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105427237B (en) | A kind of steel mesh image registration of large format optical measuring system and detection method | |
| CN108830213A (en) | Car plate detection and recognition methods and device based on deep learning | |
| CN108961236A (en) | Training method and device, the detection method and device of circuit board defect detection model | |
| CN105261022B (en) | A kind of pcb board card matching process and device based on outer profile | |
| CN104881665A (en) | Chip character identification and verification method and apparatus | |
| CN105184225B (en) | A kind of multinational banknote image recognition methods and device | |
| CN115184380B (en) | Method for detecting abnormity of welding spots of printed circuit board based on machine vision | |
| CN108765349A (en) | A kind of image repair method and system with watermark | |
| CN109558844A (en) | The method and apparatus of self-defined template discrimination is promoted based on image normalization | |
| CN109738450A (en) | The detection method and device of keyboard of notebook computer | |
| CN106529368A (en) | Method for localizing recognition region of two-dimensional code | |
| Xie et al. | A method for surface defect detection of printed circuit board based on improved YOLOv4 | |
| CN105354816B (en) | A kind of electronic units fix method and device | |
| CN106709870A (en) | Close-range image straight-line segment matching method | |
| Chen et al. | An adaptive deep learning framework for fast recognition of integrated circuit markings | |
| CN115797359A (en) | Detection method and device based on solder paste on circuit board and storage medium | |
| CN113705564A (en) | Pointer type instrument identification reading method | |
| CN107016414A (en) | A kind of recognition methods of footprint | |
| CN110288040A (en) | A kind of similar evaluation method of image based on validating topology and equipment | |
| CN113033559A (en) | Text detection method and device based on target detection and storage medium | |
| CN108345817A (en) | A kind of recognition methods of cylindrical surface Quick Response Code | |
| CN113378847A (en) | Character segmentation method, system, computer device and storage medium | |
| CN112528982A (en) | Method, device and system for detecting water gauge line of ship | |
| CN109801334A (en) | Workpiece localization method, standard point determine method, device and equipment | |
| JP3381225B2 (en) | Positioning method and positioning device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB03 | Change of inventor or designer information |
Inventor after: Wu Yunge Inventor before: Zhang Xingxiang Inventor before: Wang Wenhua |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |