CN108760766B - Image splicing method for detecting surface micro-defects of large-caliber optical crystal - Google Patents
Image splicing method for detecting surface micro-defects of large-caliber optical crystal Download PDFInfo
- Publication number
- CN108760766B CN108760766B CN201810517270.8A CN201810517270A CN108760766B CN 108760766 B CN108760766 B CN 108760766B CN 201810517270 A CN201810517270 A CN 201810517270A CN 108760766 B CN108760766 B CN 108760766B
- Authority
- CN
- China
- Prior art keywords
- image
- images
- defect
- size
- caliber
- 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
Images
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
An image splicing method for detecting micro defects on the surface of a large-caliber optical crystal relates to an image splicing method for detecting micro defects. The invention aims to solve the problem that the image acquisition and defect identification links of large-caliber crystals in the current defect detection consume longer time. The invention firstly scans the surface area of the large-caliber crystal element to be detected, and uses a detection microscope and a detection CCD to acquire real-time images of the scanning area, and determines the size range and the size of the overlapping area of a single picture: and then realizing the splicing of the collected images and the coordinate conversion of the defect points based on a coordinate system translation transformation method, determining the position of each defect point in each image under a global coordinate system, and establishing a defect database. The method is suitable for image splicing of optical crystal surface micro-defect detection.
Description
Technical Field
The invention belongs to the field of optical engineering, and particularly relates to an image splicing method for micro-defect detection.
Background
In order to relieve the problems of shortage of fossil fuels and serious environmental pollution faced by human beings, the research and development work of laser-driven inertial confinement nuclear fusion devices is carried out in various countries in the world so as to obtain controllable clean fusion energy. The nonlinear optical crystal material represented by KDP is used for manufacturing a photoelectric switch and a frequency multiplier due to unique optical performance, and becomes an irreplaceable core element in the current laser nuclear fusion engineering. However, both water-soluble growth and ultra-precision processing of large-aperture optical crystals are extremely difficult (conventional production cycles for monolithic crystals are up to two years). In addition, micrometer-scale defect points are introduced on the surface of the optical crystal element through mechanical processing and laser pretreatment, the defect points are very easy to induce laser damage in a high-energy laser use environment and expand rapidly in a subsequent strong laser targeting process, so that the whole crystal element is scrapped, and the optical performance and the service life of the large-caliber crystal element are finally limited seriously. At the present stage, the problem of laser damage caused by the micro defects on the surface of the optical crystal becomes a technical bottleneck for limiting the output energy improvement of the laser nuclear fusion device, and the control and removal technology of the micro defects on the surface of the large-caliber optical crystal is developed, so that the method has great significance for realizing the expected target of laser nuclear fusion ignition.
At present, the micro-machining technology is adopted to carry out precise repair on the micro-defects on the surface of the optical crystal, which is a most promising strategy for relieving the laser damage of the element caused by the defects and prolonging the service life of the large-caliber optical crystal element, and the strategy can realize the recycling of the expensive optical crystal with high precision, high quality and large caliber, thereby ensuring the stable operation of the high-energy load of the laser nuclear fusion device. The rapid and accurate detection of the micro-defects on the surface of the optical crystal is the key to realize the precise repair of the optical crystal. Firstly, the size and shape information of the micro-defects on the surface of the optical crystal directly determine the formulation of the subsequent micro-mechanical repair strategy and the selection of process parameters. In addition, the detection precision of the micro-defect point on the surface of the optical crystal can seriously affect the determination of the relative position of the tool and the defect point to be repaired in the repairing process, and finally can affect whether the defect point is successfully repaired or not. Particularly, the target density of the laser nuclear fusion device requires that the replacement, detection, repair and reinstallation processes of the large-caliber crystal element must be completed within 4 hours, namely, the defect detection must have the characteristic of high efficiency. However, the surface micro defects of the large-caliber optical crystal have small size, various shapes and uneven distribution, and only in the links of image acquisition and defect identification of the large-caliber crystal in defect detection, the time is usually consumed for several hours. In addition, only the defect information of a large number of local areas of the element can be obtained through the efficient detection of the microdefect, the position coordinates of the defect point are only the pixel coordinates relative to a single picture, and the position coordinates of the defect in the whole optical crystal surface global coordinate system are the most valuable information for repairing in the precision repairing process of the microdefect. Therefore, it is necessary to develop an image stitching method for detecting micro defects on the surface of a large-aperture optical crystal, which can efficiently and precisely stitch the batch and local defect images obtained by detection to obtain defect position distribution within the surface range of the full-aperture optical crystal, thereby providing important parameter information for the effective repair of the large-aperture optical crystal element.
Disclosure of Invention
The invention aims to solve the problem that the image acquisition and defect identification links of large-caliber crystals in the current defect detection consume longer time.
An image splicing method for detecting micro defects on the surface of a large-caliber optical crystal is realized based on a large-caliber KDP crystal surface micro defect quick searching and micro milling repairing device, a marble platform is arranged below an air floatation frame of the large-caliber KDP crystal surface micro defect quick searching and micro milling repairing device, a repairing window is arranged on the marble platform and right below the air floatation frame correspondingly, and a microscope and a detection CCD (881) are arranged on a microscope moving platform support (121);
an image splicing method for detecting micro-defects on the surface of a large-caliber optical crystal comprises the following steps:
the detection CCD is a charge-coupled device image sensor;
in the process of collecting images, the overlapping sizes of the visual field range of the CCD in the moving direction of the X, Y axis are detected to be respectively delta m and delta n;
moving the large-caliber crystal element to enable the edge position of the crystal element to be positioned in the repair window, and ensuring that no defect point is observed on the crystal surface in the CCD;
then, the micro milling cutter starts to rotate, the three-axis linkage platform of the cutter is lifted until the front end profile of the cutter can be observed in the detection CCD, and the cutter feeding is stopped; at this time, the tool is located at a certain position P in the CCD1Acquiring an image I1(ii) a Then the moving tool moves in the X direction by a distance X to reach the position P2Acquiring an image I2(ii) a Continue moving x to P3Acquiring an image I3(ii) a Then continuously moving twice in the Y direction by a moving distance x, and respectively collecting images I4、I5；
Processing the images at the five positions, extracting feature points by adopting an image feature matching algorithm to calculate the pixel distance of the cutter movement in the images,
matching every two images in sequence according to the moving sequence of the characteristic points by using the characteristic points concentrated at the tool nose position of the tool; respectively calculating the pixel distance of the characteristic point moving in the X, Y direction, and calculating the average offset of the characteristic point in the X, Y direction, namely the average pixel moving distance delta P of the characteristic point;
the actual magnification K of the image is thus:
K＝(α/x)·ΔP
wherein α is the pixel size;
calculating the size range of a single picture according to the actual magnification K of the image, wherein W, H is the width and the height of each picture respectively; determining the overlapping size of each collected picture;
each image in a single image acquisition is represented by' X-mx-Y-ny"wherein m isx、nyRespectively representing the scanning steps of the X, Y directional walkingThe distance number, i.e. representing the real-time position at which the captured image is located, has the same mxOr nyThe numbered images have the same X or Y coordinate position; assuming that n images exist in the Y direction and m scanned images exist in the X direction, and in a global coordinate system formed by the m multiplied by n scanned images, assuming that the upper left corner is still used as an origin; by means of Ii,jDenotes X, Y images numbered I, j, I, j being 0,1,2i,jOrigin of coordinates Oi,jThe position under the global coordinate system is O'i,jCoordinate values (i (W- Δ m), j (H- Δ n)) of each image in the global coordinate system; o isi,jRepresenting the coordinate origin of the image coordinate system corresponding to each picture;
and further determining the position of each defect point in each image under the global coordinate system, and establishing a defect database.
Further, the image splicing method for detecting the surface micro-defects of the large-caliber optical crystal further comprises the following steps:
and 4, image splicing is carried out by adopting a coordinate system translation transformation method:
firstly, establishing a blank canvas according to the number and arrangement condition of the images, wherein the size of the canvas is determined by the size of the images and the size of the mutually overlapped parts;
then, coordinate conversion is carried out on the positions of all the defective points in each image through the position information of the defective points in the defect database, and a fitting ellipse indicating the shape and the size of each defective point is drawn at the corresponding position on the 'canvas'; and completing image splicing for detecting the microdefects only containing defect information.
Further, it is necessary to measure X, Y axis positioning error using a laser interferometer and perform follow-up error compensation before determining the size range and the size of the overlapping area of a single picture.
Further, x is 500 μm.
The invention has the following beneficial effects:
(1) by comparing the efficiency and the precision based on the image feature matching algorithm and the invention, the invention takes no more than 1min to finish the whole process of image splicing for the acquired images of 90mm multiplied by 90 mm.
(2) The image coordinate system conversion splicing method can rapidly realize the splicing and defect coordinate conversion of the local images in batches through coordinate system translation transformation according to the size of the overlapping part between the adjacent collected images, and extract and summarize the information of the defect point in each collected image in the global coordinate system of the surface of the whole crystal;
(3) the invention calibrates the magnification of the defect detection microscope, determines the size of the field of view of the microscope and the detection CCD and the size of the overlapped part of a single acquired picture, and provides necessary parameter information for image splicing;
drawings
FIG. 1 is a top view of a detection microscope and a detection CCD of a large-caliber KDP crystal surface micro-defect rapid searching and micro-milling repairing device;
FIG. 2 is a schematic view of the overlapping area between the scanned images of a large aperture optical crystal;
FIG. 3 is a mosaic of defect scan images based on image feature matching;
FIG. 4 is a schematic view of a magnification calibration process;
FIG. 5 is a schematic diagram of feature point matching;
FIG. 6 is a schematic diagram of coordinate transformation splicing effect;
FIG. 7 is a cross-sectional view of data information reading each image through the database;
FIG. 8 is a schematic diagram of a final stitched image formed using the present invention in an example.
Detailed Description
The first embodiment is as follows:
an image splicing method for detecting micro-defects on the surface of a large-caliber optical crystal is realized based on a large-caliber KDP crystal surface micro-defect quick searching and micro-milling repairing device (application number: 201310744691.1), the large-caliber KDP crystal surface micro-defect quick searching and micro-milling repairing device is shown in figure 1, a marble platform is arranged below an air floatation frame of the large-caliber KDP crystal surface micro-defect quick searching and micro-milling repairing device, a repairing window is arranged on the marble platform and in a position corresponding to the position right below the air floatation frame, and a microscope and a detection CCD881 are arranged on a microscope moving platform support 121;
an image splicing method for detecting micro-defects on the surface of a large-caliber optical crystal comprises the following steps:
the detection CCD (charge Coupled device) is a charge Coupled device image sensor;
in the process of collecting images, the overlapping sizes of the visual field range of the CCD in the moving direction of the X, Y axis are detected to be respectively delta m and delta n;
the 'continuous motion acquisition' grating type scanning means that the optical crystal makes continuous motion along a grating type scanning path, and simultaneously, the upper detection CCD acquires an image at a certain time interval (namely, each time the crystal moves by one scanning step); during the crystal scanning process, a certain margin (respectively expressed by deltam and deltan) is left between the scanning step distance in the direction X, Y and the distance of the visual field range in the corresponding direction of the images, and the margin is used for determining the relative position relationship between each image of the continuous scanning and providing a necessary common overlapping area for splicing the images, as shown in fig. 2;
moving the large-caliber crystal element to enable the edge position of the crystal element to be positioned in the repair window, and ensuring that no obvious defect point exists on the crystal surface observed in the CCD;
then, the micro milling cutter starts to rotate, the three-axis linkage platform of the cutter is lifted until the front end profile of the cutter can be observed clearly in the detection CCD, the cutter feeding is stopped, and the cutter cannot be in contact with the lower surface of the upper crystal; at this time, the tool is located at a certain position P in the CCD1Acquiring an image I1(ii) a Then the cutter is moved in the X direction by a motor, and the moving distance isLeaving x, to position P2Acquiring an image I2(ii) a Continue moving x to P3Acquiring an image I3(ii) a Then continuously moving twice in the Y direction by a moving distance x, and respectively collecting images I4、I5As shown in fig. 4;
because the movement precision of the cutter moving motor is very high, the error can be effectively controlled by taking the distance as the standard length; although the imaging unit for detecting the CCD is theoretically a standard square, certain errors may exist, and imaging distortion of the CCD can be inhibited through simultaneous calibration in X, Y two directions;
processing the images at the five positions, extracting the feature points by adopting an image feature matching algorithm, and calculating the pixel distance of the cutter movement in the images, so that the error caused by artificially selecting the reference points can be avoided, and the accuracy is higher;
because the surface of the crystal does not have defect points with obvious characteristics, the characteristic points concentrated at the tool tip position are utilized to match two images in sequence according to the movement sequence of the characteristic points, and the effect is shown in figure 5; performing four matching operations on the 5 images, respectively calculating the pixel distance of the characteristic point moving in the X, Y direction, and calculating the average offset of the characteristic point in the X, Y direction, namely the average pixel moving distance delta P of the characteristic point;
the time of image matching operation is about 13s each time, and the efficiency is acceptable because the magnification calibration process is only needed to be carried out once before the optical crystal is scanned; the actual magnification K of the image is thus:
K＝(α/x)·ΔP＝(3.45/500)·ΔP
wherein α is the pixel size;
calculating the size range of a single picture according to the actual magnification K of the image, wherein W, H is the width and the height of each picture respectively; determining the overlapping size of each collected picture; taking an example of 2.25X with accurate magnification, the image width W is 3.766mm, Δ m is 0.266mm (corresponding to 196pixels), the height H is 3.156mm, Δ n is 0.156mm (corresponding to 102 pixels);
the stitching principle based on the coordinate system translation transformation method is shown in FIG. 6, in which each image in a single image acquisition is represented by "X-mx-Y-ny"wherein m isx、nyRespectively representing the number of scanning steps passed by X, Y, i.e. representing the real-time position of the captured image, having the same mxOr nyThe numbered images have the same X or Y coordinate position; assuming that n images exist in the Y direction and m scanned images exist in the X direction, and in a global coordinate system formed by the m multiplied by n scanned images, assuming that the upper left corner is still used as an origin; by means of Ii,jDenotes X, Y images numbered I, j, I, j being 0,1,2i,jOrigin of coordinates Oi,jThe position under the global coordinate system is O'i,jCoordinate values (i (W- Δ m), j (H- Δ n)) of each image in the global coordinate system, as shown in FIG. 6; o isi,jRepresenting the coordinate origin of the image coordinate system corresponding to each picture;
and further determining the position of each defect point in each image under the global coordinate system, and establishing a defect database.
The second embodiment is as follows:
the image stitching method for detecting the surface micro-defects of the large-aperture optical crystal is characterized by further comprising the following steps of:
and 4, image splicing is carried out by adopting a coordinate system translation transformation method:
the crystal scanning image is essentially a carrier of information, most of the information contained in the image is negligible, and only the parameter information of the defect point obtained by detection is really important information; therefore, as can be seen from fig. 6, the essence of image stitching through coordinate system transformation is to extract and summarize a limited amount of defect point information in each image, and omit a complete image containing a large amount of useless information;
firstly, establishing a blank canvas according to the number and arrangement condition of the images, wherein the size of the canvas is determined by the size of the images and the size of the mutually overlapped parts;
then, coordinate conversion is carried out on the positions of all the defective points in each image through the position information of the defective points in the defect database, and a fitting ellipse indicating the shape and the size of each defective point is drawn at the corresponding position on the 'canvas'; therefore, the information of all defect points in the whole scanning range can be intuitively read from the canvas; and completing image splicing for detecting the microdefects only containing defect information.
Other steps and parameters are the same as in the first embodiment
The third concrete implementation mode:
in this embodiment, before determining the size range and the size of the overlapping region of a single picture, it is necessary to measure the positioning error of the X, Y axis using a laser interferometer and perform the following error compensation.
In order to ensure the applicability and the accuracy of the invention, the motion characteristics of the X-axis and Y-axis linear units of the crystal scanning motion need to be improved, and the positioning error is measured and compensated:
the premise of conversion splicing based on an image coordinate system is to ensure that the scanning movement position precision of the optical crystal is very high, and the size of an overlapping part between images acquired each time can be determined;
a grating ruler is arranged in a servo motor and a linear motor of the optical crystal moving shaft to realize position signal feedback, and the motion stable state characteristic and the dynamic characteristic of the X, Y shaft are improved by adjusting the PID parameter of a motor closed-loop control system;
measuring X, Y axis positioning error by using a laser interferometer, and compensating following error, thereby realizing that the positioning error of the two axes is controlled within 3 μm (300mm stroke); at this time, the positional accuracy of the optical crystal scanning motion can be satisfied.
Other steps and parameters are the same as in one or both of the embodiments
Examples
The image splicing and the invention comparison are directly carried out through an image feature matching algorithm, and the following are explained:
an optical crystal element to be detected is installed on a large-caliber KDP crystal surface micro-defect rapid searching and micro-milling repairing device (application number: 201310744691.1);
scanning the surface area of the large-caliber crystal element to be detected based on a grating type scanning scheme of 'continuous motion acquisition' of the large-caliber crystal element, and acquiring real-time images of the scanning area by using a detection microscope and a detection CCD (charge coupled device) to obtain batch local single scanning images of the large-caliber crystal element;
the detection CCD (charge Coupled device) is a charge Coupled device image sensor;
then, image splicing is carried out by adopting a splicing method based on an image feature matching algorithm and the image splicing is carried out by the invention:
adopting a splicing method based on an image feature matching algorithm:
the image feature matching algorithm splicing method is a comprehensive complex implementation process, needs to call a large number of image algorithm supports to implement, and mainly comprises image feature searching and matching, lens calibration, image shape transformation, light compensation, fusion and the like. In the defect detection system shown in fig. 1, because the optical crystal moves horizontally, the detection CCD and the surface of the optical crystal are always kept in a vertical state, and the light source brightness is constant, there is no deformation and distortion between each collected image, and the process of image stitching can be simplified into two steps of feature point finding and matching and image fusion, and the specific implementation steps are as follows:
and (one by one), searching and matching the characteristic points of the locally acquired image, wherein the characteristic points (angular points) are stable points in the image and can obviously reflect the change characteristics of the image. Mathematically, the characteristic points have significant derivatives in both orthogonal directions. The SIFT algorithm, which is a common feature point extraction and matching method in the field of computer vision, is adopted. The SIFT algorithm (Speed Up Robust Features) is stable in extracting feature points under the conditions of geometric deformation and illumination change, is suitable for extracting local targets, and has the characteristic of high-Speed extraction.
The Hessian matrix is the core of the SURF algorithm, and assuming the existence of the function f (x, y), the corresponding Hessian matrix consists of the functions and their partial derivatives:
the discriminant of the Hessian matrix is the eigenvalue, and whether the point is an extreme point is determined by the positive and negative. For an image, pixel values I (x, y) at different positions are used for replacing f (x, y), a second-order standard Gaussian function is selected as a filter, and a second-order partial derivative is calculated through convolution among specific kernels, so that a Hessian matrix of the image with the scale sigma is obtained:
wherein u is (x, y)TL (u, σ) ═ G (σ) × i (u), and the internuclear function is the second-order partial derivative of the gaussian function:
The feature points can be discriminated by calculating the value of the discriminant of H (u, σ), and in order to balance the error between the accurate value and the approximate value, the following discriminant is used:
det(Happrox)＝DxxDyy-(0.9Dxy)2 (4)
after finding the feature points, the SURF algorithm constructs a pyramid-shaped image space: obtaining a multi-layer image by changing the size of a filtering kernel, then accurately positioning feature points on each layer in an image space, and finally calculating the Haar wavelet response of each feature point to be used for determining the main direction of the feature points.
And (2) fusing the locally acquired images, wherein the fusing of the images is to superpose the two images according to the overlapping area between the two images calculated according to the matching result in the step (1). In a crystal scan image, the overlapping area of the image may be approximated as a rectangle due to illumination, stability of geometry. Taking the left and right overlapped images as an example, the fused image can be divided into three parts: the left unique area and the right unique area keep corresponding pixel values of respective original images, and the common area superposes the corresponding pixel values of the left image and the right image according to the overlapping proportion.
The image feature point extraction, matching and image fusion processes are realized through a cross-platform open Source Computer Vision image processing library OpenCV (open Source Computer Vision library), wherein three optical crystal scanning images are selected, spliced two by two and then spliced again, and the result is shown in FIG. 3.
Two obvious defect points can be found in a source image, after splicing is completed in pairs, the actual overlapping area can be calculated according to the spliced image pixels, and therefore the coordinate system conversion relation from a single image to multiple images can be determined. By the circulation, image splicing of the whole surface of the optical crystal can be realized.
And (II) realizing the splicing of the collected images based on a coordinate system translation transformation method:
and (II) determining a scanning range, namely a crystal coordinate system scale range according to the number (quantity) of the acquired images. In order to facilitate the quick verification of the splicing function of the locally acquired pictures on the surface of the optical crystal and ensure the accurate relative position relationship between each image, the actual scanning size of the optical crystal is reduced to 90mm multiplied by 90mm in the embodiment, and the size of the overlapping area between the adjacent images is properly increased. For this reason, the microscope magnification is reduced from 2.25X to 1.5X, and the scanning step is reduced to Δ X ═ Δ y ═ 3.0mm, under which condition 31 × 31 ═ 961 images are acquired, which occupies about 5GB of hard disk space. The image width and height are 2456 × 2058 pixels, and the image overlap size Δ m is 1152 and Δ n is 754, so that the coordinate range X is calculatedscale＝30×2456-29×1152＝40272，Yscale30 × 2058-29 × 754 ═ 39874. Theoretically, although the X, Y directions scan the equal range, at the outermost circle of the actual scan range, the image field of view is larger than Δ x and Δ y, resulting in the image coordinate range being larger than the scan range.
(II) reading database information written in automatic defect detection, and setting the position of the defect point of each image according to the image numbering sequenceAnd (4) performing standard transformation. In the automatic defect detection program, defect point information tables named in image order are established in a database for 961 images, and details of each defect point in the image are recorded. Data sheet number N and image name number "X-mx-Y-ny"is such that N is 31mx+ny. According to the coordinate transformation image splicing method, the position of a defect point in an image X-p-Y-q.bmp is assumed to be (X)i,xj) (pixel), if the origin of coordinates is set at the scanning start point, the converted global coordinates are (x)i+p·(W-Δm),xj+ q- (H- Δ n)); when the origin is actually set as the center of the crystal, the coordinate is (x)i+p·(W-Δm)-19565,xj+ q. (H-. DELTA.n) -19565). In order to record all defect point information under the global coordinate system, a data table "tbAllDefectsInfo" is reestablished, and the setting of fields in the table is the same as the image name; each time the coordinate conversion of a defective point is completed, a corresponding field value of a row is written into the data table. The above process is circulated until the whole database is traversed, and the data information of each image is read, as shown in fig. 7.
And (II) creating a background image and marking the positions of all the defect points. In actual operation, since the image coordinate range of the full scan range is too large, an image cannot be newly created because the pixel size exceeds the allocable memory size when the image is newly created (the maximum image that can be actually newly created does not exceed 15000 × 15000 pixels). Here the coordinate range is scaled down to 1/5 in the case of the current coordinate range size, the image size is 8054 × 7974 pixels. At this time, the coordinates of the defect point are also reduced to 1/5, the area of the defect is reduced to 1/25, and when the area is less than 1, the defect point is represented by 1 pixel. The resulting stitched image is shown in fig. 8.
In fig. 8, a white background is used as the crystal surface, black dots represent defect points on the crystal surface, and the size of the diameter of the dots represents the defect area. In practice, a total of 394 defect points are detected, and the distribution positions of the defect points on the crystal have certain randomness in general. Because the image is stored in a jpg format, the size of the image spliced in the actual scanning range of 90mm multiplied by 90mm is only 1.02 MB; in the aspect of efficiency, the time consumed in the whole process from the reading and writing of the database to the final completion of the image splicing is within 1min, and the method is very efficient. Because the coordinate conversion splicing process is completed by digital operation, errors basically do not exist. Therefore, the process feasibility of image splicing by using a coordinate transformation method is verified.
Considering from the aspects of efficiency and precision of image splicing, the feasibility comparison of two methods of image feature matching algorithm splicing and image coordinate system conversion splicing is carried out:
the splicing method based on the image feature matching algorithm can be accurate to 1 pixel, and errors caused by the fact that the width direction and the height direction of the CCD and the direction of the crystal motion axis are not parallel can be eliminated to a certain extent. However, the image stitching method has a serious efficiency problem for image stitching of a large-aperture optical surface. Without compressing the image pixels, on average one stitching takes about 10 s. If a full range of complete scans are made of the surface of a large-aperture optical crystal of 410mm × 410mm, theoretically, 118 × 360/3 images (14160 images) are obtained. This splicing approach is not feasible in terms of efficiency. In addition, the image splicing algorithm is used for operating the complete image, the image is operated in a memory in the program running process, and the size of each bmp format image is 4.82 MB. With the increase of the spliced image, the image splicing of the subsequent large-size optical surface cannot be realized.
Compared with the splicing method based on the image feature matching algorithm, the image coordinate system conversion splicing method is more efficient and intuitive, and is more feasible for image splicing of large-aperture optical crystal surface micro-defect detection. Even if the image splicing method is limited by the scanning speed to cause certain errors, in actual operation, the splicing errors can be reduced by improving the acquisition frequency of crystal motion position signals in a numerical control program, so that the large-area optical crystal surface defect detection is high-precision image splicing.
Claims (3)
1. An image splicing method for detecting micro defects on the surface of a large-caliber optical crystal is realized based on a large-caliber KDP crystal surface micro defect quick searching and micro milling repairing device, a marble platform is arranged below an air floatation frame of the large-caliber KDP crystal surface micro defect quick searching and micro milling repairing device, a repairing window is arranged on the marble platform and right below the air floatation frame correspondingly, and a microscope and a detection CCD (881) are arranged on a microscope moving platform support (121);
the method is characterized by comprising the following steps:
step 1, scanning the surface area of a large-caliber crystal element to be detected, and acquiring real-time images of the scanned area by using a detection microscope and a detection CCD (charge coupled device) to obtain batch local single-sheet scanned images of the large-caliber crystal element;
the detection CCD is a charge-coupled device image sensor;
in the process of collecting images, the overlapping sizes of the visual field range of the CCD in the moving direction of the X, Y axis are detected to be respectively delta m and delta n;
step 2, determining the size range and the size of an overlapping area of a single picture:
moving the large-caliber crystal element to enable the edge position of the crystal element to be positioned in the repair window, and ensuring that no defect point is observed on the crystal surface in the CCD;
then, the micro milling cutter starts to rotate, the three-axis linkage platform of the cutter is lifted until the front end profile of the cutter can be observed in the detection CCD, and the cutter feeding is stopped; at this time, the tool is located at a certain position P in the CCD1Acquiring an image I1(ii) a Then the moving tool moves in the X direction by a distance X to reach the position P2Acquiring an image I2(ii) a Continue moving x to P3Acquiring an image I3(ii) a Then continuously moving twice in the Y direction by a moving distance x, and respectively collecting images I4、I5；
Processing the images at the five positions, extracting feature points by adopting an image feature matching algorithm to calculate the pixel distance of the cutter movement in the images,
matching every two images in sequence according to the moving sequence of the characteristic points by using the characteristic points concentrated at the tool nose position of the tool; respectively calculating the pixel distance of the characteristic point moving in the X, Y direction, and calculating the average offset of the characteristic point in the X, Y direction, namely the average pixel moving distance delta P of the characteristic point;
the actual magnification K of the image is thus:
K＝(α/x)·ΔP
wherein α is the pixel size;
calculating the size range of a single picture according to the actual magnification K of the image, wherein W, H is the width and the height of each picture respectively; determining the overlapping size of each collected picture;
step 3, realizing the splicing of the collected images and the coordinate conversion of the defect points based on a coordinate system translation transformation method:
each image in a single image acquisition is represented by' X-mx-Y-ny"wherein m isx、nyRespectively representing the number of scanning steps passed by X, Y, i.e. representing the real-time position of the captured image, having the same mxOr nyThe numbered images have the same X or Y coordinate position; assuming that n images exist in the Y direction and m scanned images exist in the X direction, and in a global coordinate system formed by the m multiplied by n scanned images, assuming that the upper left corner is still used as an origin; by means of Ii,jDenotes X, Y images numbered I, j, I, j being 0,1,2i,jOrigin of coordinates Oi,jThe position under the global coordinate system is O'i,jCoordinate values (i (W- Δ m), j (H- Δ n)) of each image in the global coordinate system; o isi,jRepresenting the coordinate origin of the image coordinate system corresponding to each picture;
further determining the position of each defect point in each image under the global coordinate system, and establishing a defect database;
step 4, carrying out image splicing by adopting a coordinate system translation transformation method:
firstly, establishing a blank canvas according to the number and arrangement condition of the images, wherein the size of the canvas is determined by the size of the images and the size of the mutually overlapped parts;
then, coordinate conversion is carried out on the positions of all the defective points in each image through the position information of the defective points in the defect database, and a fitting ellipse indicating the shape and the size of each defective point is drawn at the corresponding position on the 'canvas'; and completing image splicing for detecting the microdefects only containing defect information.
2. The image stitching method for detecting the surface micro-defects of the large-aperture optical crystal as claimed in claim 1, wherein positioning errors of X, Y axes are measured by a laser interferometer and follow-up error compensation is performed before determining the size range and the size of the overlapped region of a single picture.
3. The image stitching method for detecting the surface micro-defects of the large-aperture optical crystal as claimed in claim 2, wherein the moving distance x in the step 2 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 CN108760766A (en) | 2018-11-06 |
| CN108760766B true 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) |
Families Citing this family (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109916914B (en) * | 2019-04-10 | 2021-07-02 | 清华大学深圳研究生院 | Product defect 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 |
| CN110411346B (en) * | 2019-08-12 | 2020-12-25 | 哈尔滨工业大学 | Method for quickly positioning surface micro-defects of aspheric fused quartz element |
| 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 |
| CN111292241B (en) * | 2020-02-28 | 2022-02-18 | 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 |
| CN111553911A (en) * | 2020-05-08 | 2020-08-18 | 视睿（杭州）信息科技有限公司 | Global positioning method and device for industrial particle continuous product |
| 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 |
| 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 |
| CN114549697B (en) * | 2022-02-18 | 2024-08-06 | 北京百度网讯科技有限公司 | Image processing method, device, equipment and storage medium |
| CN114594038B (en) * | 2022-05-09 | 2022-08-09 | 天津立中车轮有限公司 | Scanning splicing application method for accurately measuring and calculating porosity |
| CN117334589A (en) * | 2022-06-24 | 2024-01-02 | 长鑫存储技术有限公司 | Method, apparatus and storage medium for locating defect in semiconductor device |
| 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 |
| 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 |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2399056A1 (en) * | 1977-07-29 | 1979-02-23 | Sagem | IMPROVEMENTS IN NUMERICALLY CONTROLLED MACHINE TOOLS |
| 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 |
| CN106645203A (en) * | 2017-02-13 | 2017-05-10 | 广州视源电子科技股份有限公司 | Image acquisition 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 |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7693336B2 (en) * | 2004-06-15 | 2010-04-06 | Fraudhalt Limited | 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 |
| 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 |
-
2018
- 2018-05-25 CN CN201810517270.8A patent/CN108760766B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| FR2399056A1 (en) * | 1977-07-29 | 1979-02-23 | Sagem | IMPROVEMENTS IN NUMERICALLY CONTROLLED MACHINE TOOLS |
| 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 |
| CN106645203A (en) * | 2017-02-13 | 2017-05-10 | 广州视源电子科技股份有限公司 | Image acquisition 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 (1)
| Title |
|---|
| 大尺寸工件在机视觉测量的图像拼接方法;王中任;《测试与仪器》;20100426;第44卷(第1期);第95-98页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108760766A (en) | 2018-11-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108760766B (en) | Image splicing method for detecting surface micro-defects of large-caliber optical crystal | |
| CN108645867B (en) | Rapid locating and batch detection method for micro defects on surface of large-caliber optical crystal | |
| CN107356608B (en) | Rapid dark field detection method for surface microdefects of large-caliber fused quartz optical element | |
| CN102490523B (en) | Automatic laser engraving system based on computer vision and service method of automatic laser engraving system | |
| CN105234802B (en) | A kind of small bulb tooling order turntable Polishing machining device and cutter presetting cutter method | |
| CN110411346B (en) | Method for quickly positioning surface micro-defects of aspheric fused quartz element | |
| CN112381827B (en) | Rapid high-precision defect detection method based on visual image | |
| CN114113114B (en) | Automatic process method for detecting and repairing micro defects on surface of large-caliber element | |
| CN114113116B (en) | Accurate detection process method for surface microdefect of large-caliber element | |
| He et al. | A critical review for machining positioning based on computer vision | |
| CN110238520B (en) | Automatic precise laser welding method based on CCD visual detection | |
| CN112697112B (en) | Method and device for measuring horizontal plane inclination angle of camera | |
| CN112365502B (en) | Calibration method based on visual image defect detection | |
| CN110044262A (en) | Contactless precision measuring instrument and measurement method based on image super-resolution rebuilding | |
| CN115170484A (en) | Characterization and classification method for surface defects of laser additive manufacturing workpiece | |
| CN105928464B (en) | Automatic measurement system based on image mosaic and measurement method | |
| CN105928463B (en) | Automatic measurement system and measurement method based on Feature Points Matching | |
| CN117074418A (en) | Method, system and storage medium for detecting semiconductor defect | |
| CN108705689B (en) | Automatic tool setting process method for repairing surface microdefects of large-caliber precise optical crystal | |
| CN116735604A (en) | Nondestructive detection system and method for internal defects of large-size laser crystal | |
| CN111260561A (en) | Rapid multi-graph splicing method for mask defect detection | |
| CN114113112B (en) | Surface micro defect positioning and identifying method based on three-light-source microscopic system | |
| CN115984530A (en) | Large-size shaft part image acquisition device and acquisition method | |
| CN114758346A (en) | High-precision image global microscopic printing document identification system | |
| JP2001304819A (en) | Nozzle hole measuring method and apparatus |
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 |