CN108760766A - A kind of image split-joint method of large-aperture optical plane of crystal microdefect detection - Google Patents
A kind of image split-joint method of large-aperture optical plane of crystal microdefect detection Download PDFInfo
- Publication number
- CN108760766A CN108760766A CN201810517270.8A CN201810517270A CN108760766A CN 108760766 A CN108760766 A CN 108760766A CN 201810517270 A CN201810517270 A CN 201810517270A CN 108760766 A CN108760766 A CN 108760766A
- Authority
- CN
- China
- Prior art keywords
- image
- crystal
- detection
- defect
- coordinate system
- 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
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/95—Investigating the presence of flaws or contamination characterised by the material or shape of the object to be examined
Landscapes
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Biochemistry (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Immunology (AREA)
- Pathology (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Image Processing (AREA)
Abstract
A kind of image split-joint method of large-aperture optical plane of crystal microdefect detection, is related to a kind of image split-joint method of microdefect detection.The present invention is longer in order to which the Image Acquisition and defect recognition link that solve the problems, such as current defects detection Large diameter crystal expend the time.The present invention is first scanned heavy caliber crystal element surface region to be measured, and carries out real time image collection to scanning area using detection microscope and detection CCD, and determines the size range and overlapping region size of single picture:It is then based on coordinate system translation transformation method and realizes the splicing of acquisition image and the coordinate conversion of defect point, determine position of each defect point under global coordinate system in each image, and establish defect database.The present invention is suitable for the image mosaic of optical crystal surface microdefect detection.
Description
Technical field
The invention belongs to optical engineering fields, and in particular to a kind of image split-joint method of microdefect detection.
Background technology
In short supply with the serious problem of environmental pollution in order to alleviate fossil fuel that the mankind are faced, countries in the world are just opened one after another
The development work for opening up Laser Driven inertial confinement fusion device, in the hope of obtaining controllable cleaning fusion energy resource.Using KDP as representative
Non-linear optical crystal material be used to make optoelectronic switch and frequency multiplier because having unique optical property, become current
Irreplaceable core element in laser fusion engineering.However, the Water-soluble growth and ultraprecise of large-aperture optical crystal add
Work is extremely difficult (traditional manufacturing cycle of monolithic crystal is just up to 2 years).Also, the mechanical processing of optical crystal element and
Laser pre-treated can its surface introduce micron dimension defect point, these defect points under high energy laser use environment easily
Damage from laser is induced, and is drastically extended during follow-up light laser is practiced shooting, scrapping for bulk crystal element is caused, it is final serious
Limit the optical property and service life of heavy caliber crystal element.At this stage, laser caused by the microdefect of optical crystal surface damages
Hinder the technical bottleneck that problem has become limitation laser fusion device output energy lift, exploitation large-aperture optical plane of crystal is micro-
Defects controlling and removal technology, to realizing that laser fusion igniting target has important meaning.
It is to alleviate defect to draw currently, carrying out precise repairing to optical crystal surface microdefect using micromachining technology
Element damage from laser is played, a kind of most promising strategy of large-aperture optical crystal element service life is extended, which can be achieved
In high precision, the recycling of high quality, heavy caliber costliness optical crystal, to ensure the high-energy load of laser fusion device
Stable operation.Quick, the accurate detection of optical crystal surface microdefect is the key that realize its precise repairing.First, optics
The size and shape information of plane of crystal microdefect directly determines formulation and the technological parameter of follow-up micromechanics correcting strategy
It chooses.In addition, the accuracy of detection of optical crystal surface microdefect point can seriously affect cutter and defect to be repaired in repair process
The determination of point relative position, whether eventually influencing the successful reparation of defect point.It is of particular importance that laser fusion device
Target practice density requirements must complete the replacement of a heavy caliber crystal element, detection, reparation and installation process again in 4 hours,
I.e. defects detection must have the characteristics that efficient.However large-aperture optical plane of crystal microdefect size is small, different, distribution
It is uneven, only in the Image Acquisition and defect recognition link of defects detection Large diameter crystal, a few hours just usually need to be taken.
Also, it can only obtain that element is a large amount of, defect information of regional area by the efficient detection of microdefect, the position coordinates of defect point
Also it is only the pixel coordinate relative to single picture, and during the precise repairing of microdefect, defect is in entire optics crystalline substance
Position coordinates in the global coordinate system of body surface face are only the information to repairing most worthy.Therefore, there is an urgent need for develop a kind of heavy caliber
The image split-joint method of optical crystal surface microdefect detection, by detected batch, the progress of local defect image
Efficiently, high-precision splicing obtains the defects of unified optical crystal surface range position distribution, is large-aperture optical crystal element
Effective repair important parameter information is provided.
Invention content
The present invention expends for the Image Acquisition and defect recognition link that solve current defects detection Large diameter crystal
Time longer problem.
A kind of image split-joint method of large-aperture optical plane of crystal microdefect detection is to be based on large-aperture KDP crystal
Surface microdefect fast searching is realized with micro- milling prosthetic device, is based on large-aperture KDP crystal surface microdefect fast searching
Be equipped with marble platform below the air supporting frame of micro- milling prosthetic device, in marble platform, immediately below corresponding air supporting frame
Position is provided with reparation window, and microscope mobile platform holder (121) is equipped with microscope and detection CCD (881);
A kind of image split-joint method of large-aperture optical plane of crystal microdefect detection, includes the following steps:
Step 1 is scanned heavy caliber crystal element surface region to be measured, and utilizes detection microscope and detection CCD
Real time image collection is carried out to scanning area, obtains part, individual scan image of heavy caliber crystal element batch;
The detection CCD is charge coupling device imaging sensor;
During acquiring image, detect the field range of CCD X, Y-axis moving direction on overlapping dimension be respectively
Δ m and Δ n;
Step 2, the size range for determining single picture and overlapping region size:
Mobile heavy caliber crystal element makes crystal element marginal position be located at and repairs in window, ensures to observe in detection CCD
The intact trapping spot of plane of crystal arrived;
Then micro- milling cutter is made to start to rotate, be lifted cutter three-shaft linkage platform, until in detecting CCD it is observed that cutter
Front end profile stops tool feeding;Cutter is located at certain position P in detecting CCD at this time1, acquisition image I1;Next mobile knife
Tool is moved in X-direction, displacement distance x, in-position P2, acquisition image I2;It continues to move to x and reaches P3, acquisition image I3;So
Twice, displacement distance x acquires image I to continuous moving respectively in the Y direction afterwards4、I5;
The image of five positions is handled, using Image Feature Matching algorithm, extraction characteristic point calculates knife in image
Have mobile pixel distance,
Using the characteristic point for concentrating on tool nose position, according to characteristic point movement sequence successively image progress two-by-two
Match;The pixel distance that characteristic point is moved in X, Y-direction is calculated separately, and calculates the mean deviation of characteristic point in the x, y direction
Amount, i.e. characteristic point mean pixel displacement distance Δ P;
Thus the practical enlargement ratio K of image is obtained:
K=(α/x) Δs P
Wherein α is Pixel Dimensions;
The size range of single picture is calculated by the practical enlargement ratio K of image, W, H are respectively the width per pictures
Degree and height;And determine the overlapping dimension of each acquisition picture;
Step 3 realizes that the splicing of acquisition image and the coordinate of defect point are converted based on coordinate system translation transformation method:
Each image is with " X-m in single image acquisitionx-Y-ny" mode name, wherein mx、nyX, the side Y are indicated respectively
To the Scanning step number passed by, that is, the real time position residing for captured images is represented, there is identical mxOr nyThe image of number has
Same X or Y-direction coordinate position;Assuming that there is n images in the Y direction, X-direction has m scan images, in this m × n scannings
In the global coordinate system of image composition, it is assumed that still using the upper left corner as origin;Use Ii,jNumber is respectively i on expression X, Y-direction, j's
Image, i, j=0,1,2..., then Ii,jCoordinate origin Oi,jPosition under global coordinate system is O'i,j, each image is complete
Coordinate value (i (W- Δs m), j (H- Δs n)) under office's coordinate system;Oi,jIndicate the coordinate per pictures correspondence image coordinate system
Origin;
And then determine each position of the defect point under global coordinate system in each image, and establish defect database.
Further, the image split-joint method of a kind of large-aperture optical plane of crystal microdefect detection, is also wrapped
Include following steps:
Step 4 carries out image mosaic using coordinate system translation transformation method:
First according to amount of images and arrangement situation, establish " painting canvas " of a blank, the size of painting canvas by image and
Overlapped portion size determines;
Then, by the location information of defect point in defect database, all defect point position in every image is carried out
Coordinate is converted, the corresponding position on " painting canvas ", is drawn according to defect point size and is illustrated its shape, the fitted ellipse of size;It is complete
At the image mosaic for the microdefect detection for containing only defective information.
Further, it is determined that needing to survey using laser interferometer before the size range and overlapping region size of single picture
The position error of amount X, Y-axis, and carry out tracking error compensation.
Further, the x is 500 μm.
The invention has the advantages that:
(1) by compare based on Image Feature Matching algorithm and the present invention efficiency and precision, the present invention for 90mm ×
The acquisition image of 90mm is completed image mosaic overall process and is taken no more than 1min.
(2) image coordinate system proposed by the invention conversion joining method can be according to the overlapping portion between adjacent acquisition image
Divide size, the splicing of batch topography is fast implemented by coordinate system translation transformation and defect coordinate is converted, every is acquired
Defect point information extracts and summarizes in the global coordinate system on bulk crystal surface in image;
(3) present invention calibrates defects detection microscope magnifications, it is determined that microscope, the visual field for detecting CCD
The lap size of range size and individual acquisition picture, necessary parameter information is provided for image mosaic;
Description of the drawings
Fig. 1 is detection microscope and the inspection of large-aperture KDP crystal surface microdefect fast searching and micro- milling prosthetic device
Survey CCD vertical views;
Overlapping region schematic diagrames of the Fig. 2 between large-aperture optical crystal scan image;
Fig. 3 is the Defect Scanning image mosaic figure based on Image Feature Matching;
Fig. 4 is magnification calibration process schematic;
Fig. 5 is characterized a matching schematic diagram;
Fig. 6 is that coordinate converts splicing effect schematic diagram;
Fig. 7 is the data information sectional view that every image is read in ergodic data library;
Fig. 8 is that the finally formed stitching image schematic diagram of the present invention is utilized in embodiment.
Specific implementation mode
Specific implementation mode one:
A kind of image split-joint method of large-aperture optical plane of crystal microdefect detection is to be based on large-aperture KDP crystal
Surface microdefect fast searching and micro- milling prosthetic device (application number:201310744691.1) realize, it is based on heavy caliber KDP
Plane of crystal microdefect fast searching is with micro- milling prosthetic device as shown in Figure 1, fast based on large-aperture KDP crystal surface microdefect
Fast search is equipped with marble platform below the air supporting frame with micro- milling prosthetic device, in marble platform, correspond to air supporting frame
Following position directly is provided with reparation window, and microscope mobile platform holder 121 is equipped with microscope and detection CCD881;
A kind of image split-joint method of large-aperture optical plane of crystal microdefect detection, includes the following steps:
Step 1 is based on heavy caliber crystal element " continuous motion pick " raster scanning scheme, to heavy caliber crystal to be measured
Element surface region is scanned, and carries out real time image collection to scanning area using detection microscope and detection CCD, is obtained
The part of heavy caliber crystal element batch, individual scan image;
The detection CCD (Charge Coupled Device) is charge coupling device imaging sensor;
During acquiring image, detect the field range of CCD X, Y-axis moving direction on overlapping dimension be respectively
Δ m and Δ n;
Described " continuous motion pick " raster scanning refers to that optical crystal does continuous fortune along raster scanning path
It is dynamic, while (i.e. crystal often moves a Scanning step) acquires an image to top detection CCD at a time interval;In crystal
In scanning process, there are certain surpluses between the field range distance of X, the Scanning step of Y-direction and image respective direction (respectively
Indicated with Δ m, Δ n), the surplus be in order to determine continuous scanning every image between relative position relation, can be image
Between splicing necessary public overlapping region is provided, as shown in Figure 2;
Step 2, the size range for determining single picture and overlapping region size:
Mobile heavy caliber crystal element makes crystal element marginal position be located at and repairs in window, ensures to observe in detection CCD
The plane of crystal no significant defect point arrived;
Then make micro- milling cutter start to rotate, cutter three-shaft linkage platform is lifted, until can more clearly in detecting CCD
It observes cutter front end profile, stops tool feeding, crystal lower surface on tool contact cannot be made;The knife in detecting CCD at this time
Tool is located at certain position P1, acquisition image I1;It is moved in X-direction followed by motor movement cutter, displacement distance x reaches position
Set P2, acquisition image I2;It continues to move to x and reaches P3, acquisition image I3;Then twice, displacement distance is continuous moving in the Y direction
X acquires image I respectively4、I5, as shown in Figure 4;
Since the kinematic accuracy of cutter mobile motor is very high, error can effectively be controlled for full-length with this distance;Inspection
Although surveying the imaging unit theoretically standard square of CCD, it is also likely to be present certain error, passes through X, Y both direction
While calibrate, can inhibit detect CCD image deformation;
The image of five positions is handled, using Image Feature Matching algorithm, extraction characteristic point calculates knife in image
Have a mobile pixel distance, in this way can to avoid it is artificial choose reference point when error, accuracy higher;
Since the defect point with obvious characteristic is not present in plane of crystal, the spy for concentrating on tool nose position is utilized
Point is levied, image is matched two-by-two successively according to characteristic point movement sequence, and effect is as shown in Figure 5;Four times are carried out to 5 images
With operation, the pixel distance that characteristic point is moved in X, Y-direction is calculated separately, and calculates characteristic point being averaged in the x, y direction
Offset, i.e. characteristic point mean pixel displacement distance Δ P;
The time of each images match operation is about 13s, since magnification calibration process need to only be scanned in optical crystal
It is preceding to carry out once, therefore the efficiency can receive;Thus the practical enlargement ratio K of image is obtained:
K=(α/x) Δs P=(3.45/500) Δ P
Wherein α is Pixel Dimensions;
The size range of single picture is calculated by the practical enlargement ratio K of image, W, H are respectively the width per pictures
Degree and height;And determine the overlapping dimension of each acquisition picture;By taking enlargement ratio is accurate 2.25X as an example, picture traverse W at this time
=3.766mm, Δ m=0.266mm (corresponding 196pixels), height H=3.156mm, Δ n=0.156mm are (corresponding
102pixels);
Step 3 realizes that the splicing of acquisition image and the coordinate of defect point are converted based on coordinate system translation transformation method:
Splicing principle based on coordinate system translation transformation method as shown in fig. 6, wherein single image acquisition in each image
With " X-mx-Y-ny" mode name, wherein mx、nyThe Scanning step number that X, Y-direction pass by is indicated respectively, that is, is represented and captured
Real time position residing for image has identical mxOr nyThe image of number has same X or Y-direction coordinate position;Assuming that in Y
Direction has n images, X-direction to have m scan images, in the global coordinate system of this m × n scan image compositions, it is assumed that still
Using the upper left corner as origin;Use Ii,jIndicate X, number is respectively i in Y-direction, the image of j, i, j=0,1,2..., then Ii,jSeat
Mark origin Oi,jPosition under global coordinate system is O'i,j, coordinate value (i (W- Δ of each image under global coordinate system
M), j (H- Δs n)), as shown in Figure 6;Oi,jIndicate the coordinate origin per pictures correspondence image coordinate system;
And then determine each position of the defect point under global coordinate system in each image, and establish defect database.
Specific implementation mode two:
A kind of image split-joint method of large-aperture optical plane of crystal microdefect detection described in present embodiment, it is special
Sign is, further comprising the steps of:
Step 4 carries out image mosaic using coordinate system translation transformation method:
Crystal scan image is substantially the carrier of an information, and the most information for including in image are can to ignore
, and real important information is only only by the parameter information of detected defect point;It therefore can from Fig. 6
Go out, convert carrying out the essence of image mosaic and be to the extraction of limited a defect point information in every image, summarize by coordinate system,
And have ignored the complete image for including a large amount of garbages;
First according to amount of images and arrangement situation, establish " painting canvas " of a blank, the size of painting canvas by image and
Overlapped portion size determines;
Then, by the location information of defect point in defect database, all defect point position in every image is carried out
Coordinate is converted, the corresponding position on " painting canvas ", is drawn according to defect point size and is illustrated its shape, the fitted ellipse of size;This
Sample, you can from the information for intuitively reading all defect point in entire scanning range in " painting canvas ";Completion contains only defective information
The image mosaic of microdefect detection.
Other steps and parameter are same as the specific embodiment one
Specific implementation mode three:
Present embodiment needs to utilize laser interference before the size range and overlapping region size for determining single picture
The position error of instrument measurement X, Y-axis, and carry out tracking error compensation.
In order to ensure the application and accuracy of the present invention, the X-axis and Y-axis straight line units that improve crystal scanning motion are needed
Kinetic characteristic, measure its position error and compensate:
Premise based on image coordinate system conversion splicing is to ensure that optical crystal scanning motion positional precision is very high, and energy
Enough lap sizes determined between the image acquired every time;
Grating scale is equipped with inside the servo motor and linear motor of optical crystal shifting axle can realize that position signal is fed back,
By the pid parameter of regulation motor closed-loop control system, to improve the movement steady-state characteristic and dynamic characteristic of X, Y-axis;
Using laser interferometer measurement X, the position error of Y-axis, and carry out tracking error compensation, it can be achieved that two axis positioning
Control errors (300mm strokes) within 3 μm;At this point, the positional precision of optical crystal scanning motion can be met the requirements.
Other steps and parameter are the same as one or two specific embodiments
Embodiment
To image mosaic directly be carried out by Image Feature Matching algorithm and the present invention carries out comparative illustration:
Microdefect fast searching and micro- milling prosthetic device (application number on large-aperture KDP crystal surface:
201310744691.1) on optical crystal element to be measured is installed;
Based on heavy caliber crystal element " continuous motion pick " raster scanning scheme, to heavy caliber crystal element table to be measured
Face region is scanned, and carries out real time image collection to scanning area using detection microscope and detection CCD, obtains heavy caliber
The part of crystal element batch, individual scan image;
The detection CCD (Charge Coupled Device) is charge coupling device imaging sensor;
Then it is respectively adopted and image mosaic and present invention progress image is carried out based on Image Feature Matching algorithm joining method
Splicing:
(1), using based on Image Feature Matching algorithm joining method:
The Image Feature Matching algorithm joining method is a comprehensive complicated realization process, needs to call big spirogram
As algorithm support realization, is mainly found including characteristics of image and matching, lens calibration, image shape convert, light compensates and melts
Close etc..In the defect detecting system of Fig. 1, since optical crystal moves horizontally, detection CCD is remained with optical crystal surface
Plumbness, light-source brightness is constant, therefore there is no deformation and distortion between the image of every acquisition, the process of image mosaic can
It finds and matches and two steps of image co-registration, specific implementation step are as follows to be reduced to characteristic point:
(one by one), locally the characteristic point of acquisition image is found and is matched, and characteristic point (angle point) refers to stabilization, energy in image
The point of apparent reflection both image change characteristics.With mathematical description, characteristic point all has apparent derivative in 2 orthogonal directions.This
In use the common feature point extraction of computer vision field and matching process --- SIFT algorithms.SIFT algorithms (Speed Up
Robust Features) characteristic point is extracted in the case where geometry deformation and illumination change and is stablized, it is suitble to the extraction of localized target, together
When it also have high speed extraction characteristic.
Hessian matrixes are the cores of SURF algorithm, it is assumed that there are function f (x, y), then corresponding Hessian matrixes by
Function and its local derviation composition:
The discriminate of Hessian matrixes is its characteristic value, differentiates whether the point is extreme point by positive and negative.To a width figure
Picture replaces f (x, y) with the pixel value I (x, y) of different location, selects second order standard gaussian function as filter, pass through spy
Fixed internuclear convolutional calculation second-order partial differential coefficient, to obtain the Hessian matrixes for the image that scale is σ:
In formula, u=(x, y)T, L (u, σ)=G (σ) * I (u), internuclear function is the second order local derviation of Gaussian function:
Or
Value by calculating the discriminate of H (u, σ) can be with judging characteristic point, in order to balance the mistake between exact value and approximation
Difference, using following discriminate:
det(Happrox)=DxxDyy-(0.9Dxy)2 (4)
After finding characteristic point, SURF algorithm builds a pyramidal image space:By the size for changing filtering core
Multi-layer image is obtained, then each layer of precise positioning feature point in image space, finally also needs to calculate each characteristic point
Haar small echos respond, the principal direction for determining characteristic point.
The fusion of the fusion of (one or two), part acquisition image, image is according to the calculated two width figure of matching result in (1)
Overlapping region as between carries out the superposition of two images.In crystal scan image, due to illumination, the stability of geometrical condition,
The overlapping region of image can be approximately rectangle.By taking the image of left and right overlapping as an example, three parts can be divided into fused image:
Left and right exclusive region and public domain, the exclusive region of middle left and right retain respective original image respective pixel value, and public domain is pressed
Overlap proportion is superimposed the respective pixel value of left images.
Library OpenCV (Open Source Computer Vision are handled by cross-platform computer visual image of increasing income
Library image characteristic point extraction, matching and image co-registration process) are realized, chooses three optical crystal scanned pictures here, first
Splice two-by-two, then spliced again, the results are shown in Figure 3.
In source images, more significantly two defect points can be found, after completing splicing two-by-two, according to spliced figure
As pixel can calculate actual overlapping area, therefore it is capable of determining that the coordinate system conversion from single image to multiple images
Relationship.The image mosaic, it can be achieved that one piece optical plane of crystal is recycled with this.
(2), the splicing of acquisition image is realized based on coordinate system translation transformation method:
(2 1) determine scanning range, i.e. crystal coordinates system range scale according to acquisition picture number (quantity).In order to just
The splicing function of picture is locally acquired in fast verification optical crystal surface, it is ensured that relative position relation is accurate between every image
Really, it is 90mm × 90mm by the actual scanning size reduction of optical crystal in this example, and suitably increases the overlapping between adjacent image
Area size.For this purpose, microscope magnifications are down to 1.5X from 2.25X, while Scanning step is reduced into Δ x=Δs y=
3.0mm needs 31 × 31=961 of acquisition images, about occupies 5GB hard drive spaces under this condition.Image is wide by a height of 2456 ×
2058 pixels acquire overlaying graphics portion size Δ m=1152, Δ n=754 and thus calculate coordinate range X at this timescale=
30 × 2456-29 × 1152=40272, Yscale=30 × 2058-29 × 754=39874.Although theoretically X, Y-direction scanning
Range is equal, but in the outmost turns of actual scanning range, and field of view is more than Δ x and Δ y, the coordinate range of image is caused to be more than
Scanning range.
(two or two) database information being written in automatic defects detection is read, by picture number sequence, every image is lacked
Trapping spot position carries out coordinate transformation.In defect autotest, 961 images are established in the database with image
The defect point information table of order name, has recorded the details of each defect point in this image.Data table numbering N is compiled with Image Name
Number " X-mx-Y-ny" correspondence be N=31mx+ny.Image split-joint method is converted according to coordinate, it is assumed that in image " X-p-Y-
Existing defects point is set to (x in q.bmp "i,xj) (pixel), it is complete after conversion if coordinate origin is located at sweep starting point
Office's coordinate is (xi+p·(W-Δm),xj+q·(H-Δn));Practical that origin is set as germ nucleus, then coordinate is (xi+p·
(W-Δm)-19565,xj+q·(H-Δn)-19565).In order to be recorded in all defect point information under global coordinate system, again
Establish a tables of data " tbAllDefectsInfo ", the setting of field is identical as Image Name in table;Often complete a defect point
Coordinate conversion, into the tables of data be written a line respective field value.The above process is recycled until traversing entire database, is read
The data information of every image, as shown in Figure 7.
(two or three) background image, mark all defect point position are created.In practical operation, due to the figure of full scan range
As coordinate range is too big, in new image due to Pixel Dimensions have exceeded can storage allocation size, can not new image it is (practical
The maximum image that can be created is no more than 15000 × 15000 pixels).Here by coordinate range scaled down, in changing coordinates model
It encloses and is contracted to 1/5 under size cases, then picture size is 8054 × 7974 pixels.At this point, defect point coordinates is also contracted to 1/5,
And defect area is contracted to 1/25, when area is less than 1, then indicates the defect point with 1 pixel.Finally formed stitching image is such as
Shown in Fig. 8.
In fig. 8, use white background as plane of crystal, the dot of black indicates the defect point of plane of crystal, dot
Diameter represents defect area.It is practical to detect 394 defect points altogether, in general, the defect point distribution on crystal
Position has certain randomness.Since image is preserved using jpg formats, practical 90mm × 90mm scanning ranges splice
Image size there was only 1.02MB;In terms of efficiency, taken from data base read-write to image mosaic overall process is finally completed
Within 1min, very efficiently.Since the process of coordinate conversion splicing is that digital operation is completed, it is substantially not present error.Thus
Demonstrate the technological feasibility that image mosaic is carried out using coordinate transformation method.
Consider from the efficiency and precision aspect of image mosaic, is turned based on the splicing of Image Feature Matching algorithm and image coordinate system
Change two methods of the feasibility comparison of splicing:
For the joining method based on Image Feature Matching algorithm, 1 pixel is can be as accurate as, and can be to a certain extent
It eliminates due to error caused by detection CCD wide, high direction and crystal movement axis direction nonparallelism.However, the image mosaic side
There are serious efficiencies for the image mosaic on large-aperture optical surface for method.Image pixel is not being compressed, is being put down
Primary splicing about takes 10s.If carrying out the complete of gamut to the large-aperture optical plane of crystal of 410mm × 410mm to sweep
It retouches, theoretically shares 118 × 360/3=14160 images.This connecting method is infeasible from efficiency.In addition, the image is spelled
It is to carry out operation to complete image to connect algorithm, and image is progress operation in memory, every bmp format in program operation process
Image size is 4.82MB.As stitching image is increasing, the image mosaic of follow-up large scale optical surface cannot achieve.
Relative to the joining method based on Image Feature Matching algorithm, it is more efficiently straight that image coordinate system converts joining method
It sees, the image mosaic of large-aperture optical plane of crystal microdefect detection is more feasible.Even if the image split-joint method meeting
It is limited by sweep speed and causes certain error, but can be moved by improving crystal in numerical control program in practical operation
The frequency acquisition of position signal reduces stitching error, to realizing that big area optical plane of crystal defects detection is the height of image
Precision is spliced.
Claims (4)
1. a kind of image split-joint method of large-aperture optical plane of crystal microdefect detection is to be based on large-aperture KDP crystal table
What face microdefect fast searching and micro- milling prosthetic device were realized, based on large-aperture KDP crystal surface microdefect fast searching with
Marble platform is equipped with below the air supporting frame of micro- milling prosthetic device, in marble platform, position immediately below corresponding air supporting frame
It sets and is provided with reparation window, microscope mobile platform holder (121) is equipped with microscope and detection CCD (881);
It is characterized by comprising the following steps:
Step 1 is scanned heavy caliber crystal element surface region to be measured, and using detection microscope and detection CCD to sweeping
It retouches region and carries out real time image collection, obtain part, individual scan image of heavy caliber crystal element batch;
The detection CCD is charge coupling device imaging sensor;
During acquiring image, detect the field range of CCD X, Y-axis moving direction on overlapping dimension be respectively Δ m
With Δ n;
Step 2, the size range for determining single picture and overlapping region size:
Mobile heavy caliber crystal element makes crystal element marginal position be located at and repairs in window, is observed in guarantee detection CCD
The intact trapping spot of plane of crystal;
Then micro- milling cutter is made to start to rotate, be lifted cutter three-shaft linkage platform, until in detecting CCD it is observed that cutter front end
Profile stops tool feeding;Cutter is located at certain position P in detecting CCD at this time1, acquisition image I1;Next mobile cutter is in X
Direction is moved, displacement distance x, in-position P2, acquisition image I2;It continues to move to x and reaches P3, acquisition image I3;Then in Y
Twice, displacement distance x acquires image I to direction continuous moving respectively4、I5;
The image of five positions is handled, using Image Feature Matching algorithm, extraction characteristic point calculates cutter in image and moves
Dynamic pixel distance,
Using the characteristic point for concentrating on tool nose position, according to characteristic point movement sequence, image is matched two-by-two successively;Point
Not Ji Suan the pixel distance that is moved in X, Y-direction of characteristic point, and calculate the mean deviation amount of characteristic point in the x, y direction, i.e.,
Characteristic point mean pixel displacement distance Δ P;
Thus the practical enlargement ratio K of image is obtained:
K=(α/x) Δs P
Wherein α is Pixel Dimensions;
Calculate the size range of single picture by the practical enlargement ratio K of image, W, H be respectively width per pictures and
Highly;And determine the overlapping dimension of each acquisition picture;
Step 3 realizes that the splicing of acquisition image and the coordinate of defect point are converted based on coordinate system translation transformation method:
Each image is with " X-m in single image acquisitionx-Y-ny" mode name, wherein mx、nyIndicate that X, Y-direction are walked respectively
The Scanning step number crossed, that is, represent the real time position residing for captured images, has identical mxOr nyThe image of number has equally
X or Y-direction coordinate position;Assuming that there is n images in the Y direction, X-direction has m scan images, in this m × n scan images
In the global coordinate system of composition, it is assumed that still using the upper left corner as origin;Use Ii,jNumber is respectively i, the figure of j on expression X, Y-direction
Picture, i, j=0,1,2..., then Ii,jCoordinate origin Oi,jPosition under global coordinate system is O'i,j, each image is in the overall situation
Coordinate value (i (W- Δs m), j (H- Δs n)) under coordinate system;Oi,jIndicate that the coordinate per pictures correspondence image coordinate system is former
Point;
And then determine each position of the defect point under global coordinate system in each image, and establish defect database.
2. a kind of image split-joint method of large-aperture optical plane of crystal microdefect detection according to claim 1,
It is characterized in that, it is further comprising the steps of:
Step 4 carries out image mosaic using coordinate system translation transformation method:
It first according to amount of images and arrangement situation, establishes " painting canvas " of a blank, the size of painting canvas is by image and mutually
Lap size determines;
Then, by the location information of defect point in defect database, coordinate is carried out to all defect point position in every image
Conversion, the corresponding position on " painting canvas " draw according to defect point size and illustrate its shape, the fitted ellipse of size;It completes only
The image mosaic of microdefect detection containing defect information.
3. a kind of image split-joint method of large-aperture optical plane of crystal microdefect detection according to claim 1 or 2,
It is characterized in that, determining that the size range of single picture and overlapping region size need to utilize laser interferometer measurement X, Y before
The position error of axis, and carry out tracking error compensation.
4. a kind of image split-joint method of large-aperture optical plane of crystal microdefect detection according to claim 3,
It is characterized in that, the x is 500 μm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810517270.8A CN108760766B (en) | 2018-05-25 | 2018-05-25 | Image splicing method for detecting surface micro-defects of large-caliber optical crystal |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810517270.8A CN108760766B (en) | 2018-05-25 | 2018-05-25 | Image splicing method for detecting surface micro-defects of large-caliber optical crystal |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108760766A true CN108760766A (en) | 2018-11-06 |
CN108760766B CN108760766B (en) | 2020-12-01 |
Family
ID=64006094
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810517270.8A Active CN108760766B (en) | 2018-05-25 | 2018-05-25 | Image splicing method for detecting surface micro-defects of large-caliber optical crystal |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108760766B (en) |
Cited By (17)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109916914A (en) * | 2019-04-10 | 2019-06-21 | 清华大学深圳研究生院 | A kind of product defects detection method and device |
CN110376475A (en) * | 2019-06-20 | 2019-10-25 | 浙江四点灵机器人股份有限公司 | Glass surface line defct device for fast detecting and method |
CN110411346A (en) * | 2019-08-12 | 2019-11-05 | 哈尔滨工业大学 | A kind of aspherical fused quartz element surface microdefect method for rapidly positioning |
CN110501344A (en) * | 2019-08-30 | 2019-11-26 | 无锡先导智能装备股份有限公司 | Battery material online test method |
CN110806412A (en) * | 2019-11-15 | 2020-02-18 | 中国工程物理研究院激光聚变研究中心 | Defect size detection method and system based on optical element |
CN111122594A (en) * | 2020-02-11 | 2020-05-08 | 中国工程物理研究院激光聚变研究中心 | Method for detecting subsurface defects of optical element by fluorescence enhancement method |
CN111292241A (en) * | 2020-02-28 | 2020-06-16 | 浙江大学 | Large-diameter optical element regional scanning splicing method |
CN111553897A (en) * | 2020-04-27 | 2020-08-18 | 上海果纳半导体技术有限公司 | Wafer defect detection equipment |
CN112991334A (en) * | 2020-05-08 | 2021-06-18 | 视睿(杭州)信息科技有限公司 | Global positioning method and device for industrial particle continuous product |
CN113029919A (en) * | 2021-03-13 | 2021-06-25 | 长春长光辰英生物科学仪器有限公司 | Cell enrichment and fluorescence counting detection device and detection and counting method |
CN114113115A (en) * | 2021-11-29 | 2022-03-01 | 哈尔滨工业大学 | High-precision automatic positioning method for micro-defects on surface of large-diameter element |
CN114280083A (en) * | 2021-12-16 | 2022-04-05 | 重庆日联科技有限公司 | Detection method for realizing industrial X-ray nondestructive testing of large-size flat casting based on linear array camera automatic CNC programming |
CN114549697A (en) * | 2022-02-18 | 2022-05-27 | 北京百度网讯科技有限公司 | Image processing method, device, equipment and storage medium |
CN114594038A (en) * | 2022-05-09 | 2022-06-07 | 天津立中车轮有限公司 | Scanning splicing application method for accurately measuring and calculating porosity |
CN116542857A (en) * | 2023-06-28 | 2023-08-04 | 南京凯视迈科技有限公司 | Multi-image self-adaptive splicing method based on large similarity |
CN116643393A (en) * | 2023-07-27 | 2023-08-25 | 南京木木西里科技有限公司 | Microscopic image deflection-based processing method and system |
WO2023245741A1 (en) * | 2022-06-24 | 2023-12-28 | 长鑫存储技术有限公司 | Method and apparatus for positioning defect in semiconductor device, and storage medium |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2399056A1 (en) * | 1977-07-29 | 1979-02-23 | Sagem | IMPROVEMENTS IN NUMERICALLY CONTROLLED MACHINE TOOLS |
WO2007093976A1 (en) * | 2006-02-13 | 2007-08-23 | Fraudhalt Limited | A method and apparatus for determining if an optical disk originated from a valid source |
CN101614677A (en) * | 2008-06-25 | 2009-12-30 | 沈阳利泰自控技术有限责任公司 | The high-precision laser holographic automatic detection system of wing-splint honeycomb structure |
CN102661956A (en) * | 2012-04-24 | 2012-09-12 | 浙江大学 | Super-smooth surface defect detection system and distortion correction method thereof |
CN103692561A (en) * | 2013-12-30 | 2014-04-02 | 哈尔滨工业大学 | Surface microdefect fast search and micro-milling repair device for large-diameter KDP crystal elements |
CN105444699A (en) * | 2015-11-11 | 2016-03-30 | 苏州大学附属儿童医院 | Coordinate and displacement error detection and compensation method for microscope operating system |
CN105737748A (en) * | 2016-02-22 | 2016-07-06 | 中国核电工程有限公司 | Fuel pellet boundary dimension and apparent defect detection device and method |
CN105931188A (en) * | 2016-05-06 | 2016-09-07 | 安徽伟合电子科技有限公司 | Method for image stitching based on mean value duplication removal |
CN106290395A (en) * | 2016-10-12 | 2017-01-04 | 湖北器长光电股份有限公司 | A kind of wafer detecting apparatus based on image procossing and method |
CN106525865A (en) * | 2016-11-30 | 2017-03-22 | 南京理工大学 | Wafer image analysis device and method based on image processing |
CN106645203A (en) * | 2017-02-13 | 2017-05-10 | 广州视源电子科技股份有限公司 | Image capturing method and device |
CN107356608A (en) * | 2017-07-21 | 2017-11-17 | 中国工程物理研究院激光聚变研究中心 | The quick dark field detection method of heavy caliber fused quartz optical component surface microdefect |
-
2018
- 2018-05-25 CN CN201810517270.8A patent/CN108760766B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
FR2399056A1 (en) * | 1977-07-29 | 1979-02-23 | Sagem | IMPROVEMENTS IN NUMERICALLY CONTROLLED MACHINE TOOLS |
WO2007093976A1 (en) * | 2006-02-13 | 2007-08-23 | Fraudhalt Limited | A method and apparatus for determining if an optical disk originated from a valid source |
CN101614677A (en) * | 2008-06-25 | 2009-12-30 | 沈阳利泰自控技术有限责任公司 | The high-precision laser holographic automatic detection system of wing-splint honeycomb structure |
CN102661956A (en) * | 2012-04-24 | 2012-09-12 | 浙江大学 | Super-smooth surface defect detection system and distortion correction method thereof |
CN103692561A (en) * | 2013-12-30 | 2014-04-02 | 哈尔滨工业大学 | Surface microdefect fast search and micro-milling repair device for large-diameter KDP crystal elements |
CN105444699A (en) * | 2015-11-11 | 2016-03-30 | 苏州大学附属儿童医院 | Coordinate and displacement error detection and compensation method for microscope operating system |
CN105737748A (en) * | 2016-02-22 | 2016-07-06 | 中国核电工程有限公司 | Fuel pellet boundary dimension and apparent defect detection device and method |
CN105931188A (en) * | 2016-05-06 | 2016-09-07 | 安徽伟合电子科技有限公司 | Method for image stitching based on mean value duplication removal |
CN106290395A (en) * | 2016-10-12 | 2017-01-04 | 湖北器长光电股份有限公司 | A kind of wafer detecting apparatus based on image procossing and method |
CN106525865A (en) * | 2016-11-30 | 2017-03-22 | 南京理工大学 | Wafer image analysis device and method based on image processing |
CN106645203A (en) * | 2017-02-13 | 2017-05-10 | 广州视源电子科技股份有限公司 | Image capturing method and device |
CN107356608A (en) * | 2017-07-21 | 2017-11-17 | 中国工程物理研究院激光聚变研究中心 | The quick dark field detection method of heavy caliber fused quartz optical component surface microdefect |
Non-Patent Citations (3)
Title |
---|
TIAN, QI-CHONG: "Histogram-Based Color Transfer for Image Stitching", 《JOURNAL OF IMAGING》 * |
王中任: "大尺寸工件在机视觉测量的图像拼接方法", 《测试与仪器》 * |
罗群明: "图像拼接方法综述", 《传感器与微系统》 * |
Cited By (23)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109916914A (en) * | 2019-04-10 | 2019-06-21 | 清华大学深圳研究生院 | A kind of product defects detection method and device |
CN110376475A (en) * | 2019-06-20 | 2019-10-25 | 浙江四点灵机器人股份有限公司 | Glass surface line defct device for fast detecting and method |
CN110411346A (en) * | 2019-08-12 | 2019-11-05 | 哈尔滨工业大学 | A kind of aspherical fused quartz element surface microdefect method for rapidly positioning |
CN110501344A (en) * | 2019-08-30 | 2019-11-26 | 无锡先导智能装备股份有限公司 | Battery material online test method |
CN110806412A (en) * | 2019-11-15 | 2020-02-18 | 中国工程物理研究院激光聚变研究中心 | Defect size detection method and system based on optical element |
CN111122594A (en) * | 2020-02-11 | 2020-05-08 | 中国工程物理研究院激光聚变研究中心 | Method for detecting subsurface defects of optical element by fluorescence enhancement method |
CN111292241A (en) * | 2020-02-28 | 2020-06-16 | 浙江大学 | Large-diameter optical element regional scanning splicing method |
CN111553897B (en) * | 2020-04-27 | 2024-03-12 | 上海果纳半导体技术有限公司 | Wafer defect detection equipment |
CN111553897A (en) * | 2020-04-27 | 2020-08-18 | 上海果纳半导体技术有限公司 | Wafer defect detection equipment |
CN112991334A (en) * | 2020-05-08 | 2021-06-18 | 视睿(杭州)信息科技有限公司 | Global positioning method and device for industrial particle continuous product |
CN113029919A (en) * | 2021-03-13 | 2021-06-25 | 长春长光辰英生物科学仪器有限公司 | Cell enrichment and fluorescence counting detection device and detection and counting method |
CN113029919B (en) * | 2021-03-13 | 2024-04-02 | 长春长光辰英生物科学仪器有限公司 | Cell enrichment and fluorescence counting detection device and detection and counting method |
CN114113115A (en) * | 2021-11-29 | 2022-03-01 | 哈尔滨工业大学 | High-precision automatic positioning method for micro-defects on surface of large-diameter element |
CN114113115B (en) * | 2021-11-29 | 2023-08-18 | 哈尔滨工业大学 | High-precision automatic positioning method for micro defects on surface of large-caliber element |
CN114280083B (en) * | 2021-12-16 | 2023-11-07 | 重庆日联科技有限公司 | Detection method for realizing industrial X-ray nondestructive detection of large-size flat castings based on automatic CNC programming of linear array camera |
CN114280083A (en) * | 2021-12-16 | 2022-04-05 | 重庆日联科技有限公司 | Detection method for realizing industrial X-ray nondestructive testing of large-size flat casting based on linear array camera automatic CNC programming |
CN114549697A (en) * | 2022-02-18 | 2022-05-27 | 北京百度网讯科技有限公司 | Image processing method, device, equipment and storage medium |
CN114594038A (en) * | 2022-05-09 | 2022-06-07 | 天津立中车轮有限公司 | Scanning splicing application method for accurately measuring and calculating porosity |
WO2023245741A1 (en) * | 2022-06-24 | 2023-12-28 | 长鑫存储技术有限公司 | Method and apparatus for positioning defect in semiconductor device, and storage medium |
CN116542857A (en) * | 2023-06-28 | 2023-08-04 | 南京凯视迈科技有限公司 | Multi-image self-adaptive splicing method based on large similarity |
CN116542857B (en) * | 2023-06-28 | 2023-10-20 | 南京凯视迈科技有限公司 | Multi-image self-adaptive stitching method |
CN116643393A (en) * | 2023-07-27 | 2023-08-25 | 南京木木西里科技有限公司 | Microscopic image deflection-based processing method and system |
CN116643393B (en) * | 2023-07-27 | 2023-10-27 | 南京木木西里科技有限公司 | Microscopic image deflection-based processing method and system |
Also Published As
Publication number | Publication date |
---|---|
CN108760766B (en) | 2020-12-01 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108760766A (en) | A kind of image split-joint method of large-aperture optical plane of crystal microdefect detection | |
CN107356608B (en) | Rapid dark field detection method for surface microdefects of large-caliber fused quartz optical element | |
CN108645867B (en) | Rapid locating and batch detection method for micro defects on surface of large-caliber optical crystal | |
CN114113114B (en) | Automatic process method for detecting and repairing micro defects on surface of large-caliber element | |
CN109940603A (en) | A kind of crusing robot arrives point tolerance compensating control method | |
JP2008014940A (en) | Camera calibration method for camera measurement of planar subject and measuring device applying same | |
CN112381827B (en) | Rapid high-precision defect detection method based on visual image | |
CN112697112B (en) | Method and device for measuring horizontal plane inclination angle of camera | |
CN110411346A (en) | A kind of aspherical fused quartz element surface microdefect method for rapidly positioning | |
CN112365502B (en) | Calibration method based on visual image defect detection | |
CN109472778B (en) | Appearance detection method for towering structure based on unmanned aerial vehicle | |
CN114289332A (en) | Visual identification and positioning method and device for workpiece sorting and sorting system | |
CN111126381A (en) | Insulator inclined positioning and identifying method based on R-DFPN algorithm | |
CN114113116B (en) | Accurate detection process method for surface microdefect of large-caliber element | |
CN114434036B (en) | Three-dimensional vision system for gantry robot welding of large ship structural member and operation method | |
CN102592302B (en) | Digital cartoon intelligent dynamic detection system and dynamic detection method | |
CN113983934B (en) | Copper-clad plate online high-speed dimension measurement method and device based on double-line-array camera | |
CN113344782B (en) | Image stitching method and device, storage medium and electronic device | |
CN110977154A (en) | Positioning marking method suitable for large breadth | |
CN110238520A (en) | A kind of automatic precision method for laser welding based on CCD vision-based detection | |
CN105928464B (en) | Automatic measurement system based on image mosaic and measurement method | |
CN117074418A (en) | Method, system and storage medium for detecting semiconductor defect | |
JP2011141807A (en) | Method and system for management of mill ends | |
WO2021062696A1 (en) | Element reproduction method, device, and system | |
CN114113115B (en) | High-precision automatic positioning method for micro defects on surface of large-caliber element |
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 |