CN108346141B - Method for extracting defects of single-side light-entering type light guide plate - Google Patents

Method for extracting defects of single-side light-entering type light guide plate Download PDF

Info

Publication number
CN108346141B
CN108346141B CN201810025157.8A CN201810025157A CN108346141B CN 108346141 B CN108346141 B CN 108346141B CN 201810025157 A CN201810025157 A CN 201810025157A CN 108346141 B CN108346141 B CN 108346141B
Authority
CN
China
Prior art keywords
light guide
image
guide plate
defects
executing
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
Application number
CN201810025157.8A
Other languages
Chinese (zh)
Other versions
CN108346141A (en
Inventor
李俊峰
李明睿
朱文维
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Hengdian Technology Co.,Ltd.
Original Assignee
Hangzhou Hengdian Technology Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hangzhou Hengdian Technology Co ltd filed Critical Hangzhou Hengdian Technology Co ltd
Priority to CN201810025157.8A priority Critical patent/CN108346141B/en
Publication of CN108346141A publication Critical patent/CN108346141A/en
Application granted granted Critical
Publication of CN108346141B publication Critical patent/CN108346141B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G06T7/0008—Industrial image inspection checking presence/absence
    • G—PHYSICS
    • 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
    • G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/155—Segmentation; Edge detection involving morphological operators
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10004—Still image; Photographic image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Biochemistry (AREA)
  • Signal Processing (AREA)
  • Geometry (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Chemical & Material Sciences (AREA)
  • Analytical Chemistry (AREA)
  • Quality & Reliability (AREA)
  • General Health & Medical Sciences (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
  • Image Processing (AREA)
  • Image Analysis (AREA)

Abstract

The invention provides a method for extracting defects of a unilateral side light-incoming type light guide plate, which comprises the following steps: acquiring a light guide plate image, extracting a light guide plate body image, acquiring a width M and a height N, quickly detecting defects, quickly detecting whether the defects exist, removing noise interference, carrying out gray level conversion, carrying out OTSU threshold segmentation, automatically carrying out partition detection, traversing the gray level range of pixels in a light guide point area, judging whether bright and dark points exist in the light guide plate, carrying out morphological processing, judging whether pressure injury or foreign matters exist, carrying out image segmentation, judging whether scratch guide defects exist and extracting defective areas of unqualified products; the invention provides a self-adaptive automatic partitioning algorithm for developing a light guide plate, automatically divides different detection areas according to the density of light guide holes on the surface, and automatically adjusts the detection algorithm to realize defect extraction; the algorithm has high operation efficiency, high accuracy, high stability and strong robustness, can identify common defects, and has higher detection capability on unusual tiny defects.

Description

Method for extracting defects of single-side light-entering type light guide plate
Technical Field
The invention relates to a defect detection algorithm, in particular to a method for extracting defects of a single-side light-incoming type light guide plate.
Background
The Light Guide Plate (LGP) is made of optical acrylic/PC board, and then high-tech material with high reflectivity and no Light absorption is used, and Light Guide points are printed on the bottom surface of the optical acrylic board by laser engraving, V-shaped cross grid engraving and UV screen printing technology. The LGP has the bright characteristics of ultrathin property, ultrabrightness, uniform light guide, energy conservation, environmental protection, no dark space, durability, difficult yellowing, simple and quick installation and maintenance and the like, and is widely applied to the occasions of backlight sources of liquid crystal displays, ultrathin advertising lamp boxes, medical X-ray viewers, flat-plate lamp decoration illumination, light efficiency application of lighting engineering, luminous nameplates and the like. In the production and manufacturing processes of silk-screen printing, chemical etching, laser processing, point-hitting processing and the like of the light guide plate, due to the influence of factors such as raw material components, equipment use conditions, processing technology, worker operation and the like, processing defects such as bright spots, missing points, black spots, screen printing ink, line scratches, mirror surface point damages, shadows and the like inevitably occur on the surface of the light guide plate. If a defective light guide plate is used in a backlight of a liquid crystal display, a medical X-ray viewer, or the like, uniform light emission is affected, the efficiency of light use is reduced, and even more seriously, the eyesight of a person is impaired. Further, with the development of high performance and high precision of the device, the requirements for the material characteristics, surface quality, structural shape, reliability, and the like of the light guide plate are increasing. Therefore, the quality of the light guide plate must be checked before the light guide plate is shipped, and the light guide plate with defects must be removed.
Because the defects of the light guide plate have the characteristics of various types, different expression forms, fuzzy or no obvious edge, low contrast ratio and the like, domestic light guide plate production enterprises mainly use detection personnel who are trained professionally to highlight at various angles for manual detection. Manual detection has many limitations and problems: (1) the work environment of the staff is poor, the eyesight of the staff is easy to be deteriorated by the highlight work, and the staff can suffer from occupational diseases for a long time; (2) because defects can be found under the condition of multi-angle strong light, the requirements on skills and experience of staff are high, and the staff cannot easily master the working skills; (3) the product defects are not easy to intercept completely, and the product quality fluctuation is caused under the influence of staff experience, fluctuation change of working mood, reduction of concentration degree and the like; (4) the dependence degree of a factory on quality management personnel is high, the mobility of operating personnel is high, and training degree difference and subjective identification difference exist among different personnel, so that the detection accuracy and consistency cannot be guaranteed; (5) it is difficult for the auxiliary metrology tools to establish quantifiable quality standards using human eye recognition judgment. The digital image processing technology has the advantages of large information content, intuitive expression form, convenient transmission and storage and the like, along with the development of electronics, computers and communication technologies, the surface defect detection based on machine vision becomes possible, and domestic and foreign scholars develop extensive research on the surface defect detection, and some research achievements are successfully applied to the surface defect detection of products such as steel balls, rails and the like.
The detection algorithm is finally applied to online detection of the surface defects of the light guide plate. Because the requirement on the manufacturing accuracy of the light guide plate is high, the defect of the light guide plate is generally very small, a high-resolution linear array camera is required to be used for imaging in order to detect the defect of the light guide plate, and the image of one light guide plate is nearly 500MB, which puts a high requirement on the online detection efficiency of the defect. The light guide plate manufacturing enterprises generally require that the detection speed of each light guide plate is within 5 seconds, so the detection algorithm needs to have high surface defect correct identification rate and high operation efficiency. By adopting Curvelet transformation, nonsubsampled Contourlet transformation, shearlet transformation and wavelet transformation stool multi-scale analysis technologies, the operation efficiency of the algorithm cannot meet the requirements and the algorithm is difficult to realize in an embedded system; the correct recognition rate of some surface defects is lower, and the precision requirement cannot be met. In addition, the current light guide plate defect detection algorithm can only detect the light guide plate with uniformly arranged light guide points, but cannot detect the light guide plate with single-side incident light and non-uniform light guide points.
Accordingly, there is a need for improvements in the art.
Disclosure of Invention
The invention aims to provide a method for extracting defects of a unilateral side light-incoming type light guide plate based on a region self-adaptive technology.
In order to solve the above technical problem, the present invention provides a method for extracting defects of a single-side light-entering light guide plate, comprising the following steps:
(1) acquiring a light guide plate image F, and executing the step 2;
(2) extracting a light guide plate body image H (x, y) according to the light guide plate image F, and executing the step 3;
(3) acquiring the width M and the height N of the light guide plate body image H (x, y), and executing the step 4;
(4) performing defect rapid detection on the light guide plate body image H (x, y) by adopting a sparse representation method to obtain an SR value of the light guide plate body image H (x, y), and executing the step 5;
(5) if the SR value of the light guide plate body image H (x, y) is greater than the threshold TH and is therefore a defective image, execute step 6; otherwise, if the image is a defect-free image, the light guide plate is judged to be a qualified product;
(6) removing noise interference in the light guide plate body image H (x, y) by using mean value filtering to obtain a new image J (x, y); executing the step 7;
(7) carrying out gray level transformation on the new image J (x, y) to obtain an enhanced image K (x, y), and executing the step 8;
(8) performing OTSU threshold segmentation on the image K (x, y) obtained in the step 7, and segmenting the light guide point and the background to obtain a foreground image and a background image; executing the step 9;
(9) according to the density degree of the light guide points, automatic partition detection of the image K (x, y) obtained in the step 7 is realized; executing the step 10;
(10) traversing gray scale range of all light guide point region pixels of the foreground image, and utilizing a formula
Figure GDA0001598722570000031
Calculating the average value G of the gray value of each light guide pointaveIn the formula NiThe total number of pixels of the ith light guide point, GiThe gray sum of all pixels of the ith light guide point is obtained; executing the step 11;
(11) set the maximum evaluation value GmaxAnd a minimum evaluation value GminIf G isave＞GmaxIf the light guide plate has bright spots, the light guide plate is directly judged to be an unqualified product; if G isave＜GminIf the light guide plate has dark spots, the light guide plate is directly judged to be an unqualified product; if G ismin≤Gave≤GmaxIf no bright or dark spot exists, executing step 12;
(12) performing corrosion operation and expansion operation on the image partitioned in the step 9 to obtain a morphologically processed image, and then executing a step 13;
(13) analyzing all connected domain features of the morphologically processed image obtained in step 12, calculating the area S of each connected domain, and setting a determination value Smax(ii) a If S is presenti＞SmaxIf so, directly judging the light guide plate to be an unqualified product if the light guide plate has pressure damage or foreign matters; if all S are presenti≤SmaxThen go to step 14;
(14) segmenting the light guide plate body image H (x, y) to obtain a segmented image g (x, y); step 15 is executed;
(15) traversing all the segmented region lengths Li(i is 0,1,2,3.. N), and a discrimination length criterion L is setmaxIf L isi＞LmaxIf the light guide plate has scratch defects, judging the light guide plate to be an unqualified product; if L isi≤LmaxJudging the product as a qualified product;
(16) and extracting all defect areas of the light guide plate judged as the unqualified product, finally representing the defect areas by means of the minimum external rectangle, and calculating the mathematical characteristics of the defect areas.
As an improvement of the method for extracting the defects of the single-side light-incoming type light guide plate, the step (4) comprises the following steps:
inputting image blocks s divided by a light guide plate body image H (x, y), constructing an obtained dictionary D', and selecting an atomic number k;
outputting a sparse representation coefficient X;
s1, partitioning the image to be detected into blocks by the same method, wherein S is one of the blocks; initial residual r0The index set ^ Θ, J ═ Θ, and the iteration time t ═ 1;
s2 calculating residual rtSearching k maximum values from the relation number u of the atoms in the dictionary D', and adding indexes corresponding to the coordinates of the maximum values into an index set J;
s3 calculates a local constraint weighting factor ω ═ dist (S, D '), where ω represents the similarity between residuals S and D', dist (S, D ') [ dist (S, D')1),...,dist(s,dm)]TRepresenting residual s and dictionary di(i ═ 1.., m) in euclidean distance; then weighting the k maximum correlation coefficients u, and satisfying the condition of 2| u | ≧ | umaxAdding the atom of | to the index set Λ to update the support set;
s4 adopts least square method to approximate signal Xt＝arg min‖s-D'Λxt－1‖2Calculating residual r under the new support sett＝s-D'Λxt；
S5 if | rt‖2＜10-1Or the number num (Λ) of sparse coefficients is more than or equal to 6, the step S6 is skipped; otherwise, returning to repeat the step S2 until the stop condition is met;
s6 outputs a sparse representation coefficient X;
s7 use sparse table to find out l of coefficient X0The ratio of the norm to the size of the original image is used as an evaluation function of sparsity of the light guide plate body image H (x, y):
SR＝||X||0/mn。
as a further improvement of the method for extracting the defects of the single-side light-entering light guide plate, the step (14) comprises the following steps:
14.1, performing fast Fourier transform on the light guide plate body image H (x, y) according to the size M N of the image H (x, y); step 14.2 is executed;
14.2, generating an ideal low-pass filter; step 14.3 is executed;
Figure GDA0001598722570000041
wherein D is0The radius of the pass band is represented, and the calculation mode of D (u, v) is the distance between two points;
Figure GDA0001598722570000042
14.3, performing convolution operation on the light guide plate body image H (x, y) in a frequency domain by using an ideal low-pass filter; setting an arbitrary point f (x, y) in the light guide plate body image H (x, y), wherein the point after convolution is g (x, y); step 14.4 is executed;
Figure GDA0001598722570000043
14.4, performing inverse fast Fourier transform on the light guide plate body image H (x, y) after convolution in the step (14.3), and transforming the defects into a space domain; step 14.5 is executed;
14.5, adopting a fixed threshold value TH, carrying out image segmentation on the image obtained in the step (14.4) by using the following formula, and executing a step 15;
Figure GDA0001598722570000051
where g (x, y) is the image after segmentation and TH is the segmentation threshold.
As a further improvement of the method for extracting defects of the single-side light-entering light guide plate, the step (9) comprises the following steps:
according to the light guide point from sparse to dense, an image K (x, y) is averagely divided into N (i is 1,2, the area of each region is S); obtaining the Nth threshold value according to the threshold value result in the step (8)iThe number of regional light guide points is MiFrom the formula
Figure GDA0001598722570000052
Obtaining the density P of the Ni regioni(ii) a Setting a discrimination density value Qj(j ═ 1,2,3.. 9), for all regions of image K (x, y), if (Q)j＜Pi＜Qj+1)U(Qj＜Pi+1＜Qj+1) Then N isi+Ni+1，NiRegion and Ni+1And merging the areas.
As a further improvement of the method for extracting defects of the single-side light-entering light guide plate of the present invention, the step (16) includes:
taking pixel points of the corresponding light guide plate body image H (x, y) as random variable values f (x, y) and p + q orders of the interested region TMoment of being
Figure GDA0001598722570000053
The centroid coordinate of the region of interest T is (x)1,y1):
Figure GDA0001598722570000054
Moving the center of mass to the origin position of the reference coordinate system to obtain the center distance:
Figure GDA0001598722570000055
similarly, find u00、u20、u02Etc., then the length and width of the minimum bounding rectangle of the region of interest T:
Figure GDA0001598722570000056
an image containing the region of interest T is obtained.
The technical advantages of the method for extracting the defects of the unilateral side light-incoming light guide plate are as follows:
the invention provides a self-adaptive automatic partitioning algorithm for developing a light guide plate, which can automatically partition different detection areas according to the density of light guide holes on the surface, automatically adjust the detection algorithm and realize defect extraction. The online operation experiment result shows that the algorithm has high operation efficiency, high accuracy and strong stability and robustness, can identify common defects and has higher detection capability on unusual tiny defects;
the invention has strong adaptability to illumination change and light guide plate type change;
the invention only needs to adjust a plurality of control parameters during production and installation, and then all automatic detection is carried out without manual guard;
the algorithm is stable, and the system is convenient to overhaul and maintain;
the invention can also detect the line scratch defect.
Drawings
The following describes embodiments of the present invention in further detail with reference to the accompanying drawings.
FIG. 1 is a schematic flow chart of a method for extracting defects of a single-side light-incident light guide plate according to the present invention;
FIG. 2 is a schematic view of a process for rapid defect detection of an image of a light guide plate;
FIG. 3 is an image taken at step 1 of the present invention;
FIG. 4 is the mean filtered image obtained in step 6 of the present invention;
FIG. 5 is the contrast adjusted image obtained in step 7 of the present invention;
FIG. 6 is an automatic zoning image of the light guide plate obtained in step 9 of the present invention;
FIG. 7 is a view of the step 10 of the present invention of obtaining a traversal gray-level image;
FIG. 8 is a dilated image resulting from step 12 of the present invention;
FIG. 9 is a post-etch image obtained at step 12 of the present invention;
FIG. 10 is a connected component image obtained in step 13 of the present invention;
FIG. 11 is the fast Fourier transformed image obtained at step 14.1 of the present invention;
FIG. 12 is a low pass filtered image of the invention containing scratch defects obtained at step 14.3;
FIG. 13 is a diagram showing the result of testing a light guide plate for scratches, wherein the inside of the frame is a defect area;
FIG. 14 is an original drawing of a light guide plate with a bright spot defect;
FIG. 15 is a diagram showing the result of detecting a light guide plate having a bright spot defect, wherein the inside of the frame is a defective area;
FIG. 16 is an original drawing of a light guide plate with a crush defect;
FIG. 17 is a diagram showing the results of testing a light guide plate for defects due to pressure damage, wherein the inside of the frame is a defect area;
FIG. 18 is an original drawing of a light guide plate with scratch defects;
FIG. 19 is a diagram showing the results of testing a light guide plate for scratches, wherein the inside of the frame is a defect area.
Detailed Description
The invention will be further described with reference to specific examples, but the scope of the invention is not limited thereto.
Example 1, a method for extracting defects from a single-sided light-incident light guide plate, as shown in fig. 1-19,
(1) acquiring a light guide plate image F with high-precision gray scale by adopting a 16k line scanning camera of a Dalsa company; executing the step 2;
through observation, the light guide plate can uniformly emit light by various light guide points with different densities and sizes. The manufacturing precision requirement of the light guide plate is high, the defects of the light guide plate are generally very small, and a high-resolution linear array camera is required to be used for imaging in order to detect the defects of the light guide plate.
(2) Removing the interference of the edge of the detection workbench and extracting the body part of the light guide plate; executing the step 3;
first, the light guide plate image F (X, Y) is denoised by a Gaussian filter, and the smoothed image G (X, Y) is as follows
G(X,Y)＝H(X,Y,σ)×F(X,Y)
Wherein a Gaussian function
Figure GDA0001598722570000071
(e is a natural constant, e is approximately equal to 2.7182), sigma is the standard deviation of a Gaussian filter function, the smoothness degree is controlled, and sigma is 0.2 through a large number of experimental comparisons.
The gradient magnitude T and the direction angle θ of the image G (X, Y) are calculated.
Figure GDA0001598722570000072
θ[X，Y]＝arctan(GX(X，Y)/GY(X，Y))
Wherein G isXAnd GYRespectively X, Y partial derivatives.
GXAnd GYThe following 2 × 2 first order difference approximation calculation can be used:
Gx＝[F(X+1，Y)-F(X，Y)+F(X+1，Y+1)-F(X，Y+1)]/2
Gy＝[F(X，Y+1)-F(X，Y)+F(X+1，Y+1)-F(X+1，Y)]/2
the gradient magnitude T (X, Y) is non-maximally suppressed using interpolation (here, interpolation of the samples). Selecting double thresholds and connecting image edges, wherein high and low threshold parameters in the traditional Canny algorithm are manually selected, in order to enhance the applicability and the sensitivity of the algorithm, the threshold of the patent adopts an adaptive threshold, the largest pixel value in the passing process is searched and recorded as L, an image edge gradient histogram is constructed, and the accumulation of the total number of pixels which are not 0 in the image after the non-maximum value suppression (NMS) is counted and recorded as Hist.
Assuming that the gray-scale value at C × Hist (0 < C < 1) is L, the high and low thresholds THH、THLIs calculated as follows
THH＝L+1
THL＝0.3*THH
And obtaining a light guide plate edge image Q (X, Y) by the self-adaptive Canny edge detection algorithm, and extracting a light guide plate body image H (X, Y) according to the light guide plate edge image Q (X, Y) subtracted by the light guide plate image F (X, Y) to obtain the light guide plate body image H (X, Y).
(3) Acquiring the width M and the height N of the light guide plate body image H (x, y); executing the step 4;
(4) performing defect rapid detection on the light guide plate body image H (x, y) by adopting a sparse representation method; executing the step 5;
the invention provides a sparsity evaluation function for representing the sparsity of linear expression coefficients of a sample to be detected in a dictionary D', and judging whether an image to be detected is a non-defective image or a defective image by using the size of the sparsity expression coefficients so as to achieve the purpose of detecting defects of a light guide plate.
One light guide plate image (light guide plate body image H (x, y)) obtained by a 16K line scan camera is about 500MB, which puts high demands on the efficiency of defect online detection. The defective rate of the light guide plates is generally low, and if each light guide plate is subjected to detailed detection and the defect type is identified, the online detection requirement is certainly difficult to meet. The invention designs a rapid detection method for only detecting whether the light guide plate has defects or not, and then carries out detailed detection on the images of the light guide plate with the defects.
The purpose of signal sparse representation is to represent signals by using as few atoms as possible in a given overcomplete dictionary, and a more concise representation mode of the signals can be obtained, so that information contained in the signals can be obtained more easily, and the signals can be processed more conveniently. Researches show that after signals are subjected to sparse representation, the more sparse the signals are, the higher the accuracy of the signals after reconstruction is, and the sparse representation can adaptively select a proper overcomplete dictionary according to the characteristics of the signals. The purpose of sparse representation of the signal is to find an adaptive dictionary to make the representation of the signal sparsest. According to the characteristic of sparse representation, the invention adopts sparse representation to rapidly detect whether the light guide plate has defects.
And partitioning the defect-free light guide plate image, taking the image block as an atom, extracting image features, and constructing an over-complete redundant dictionary. Setting the size of the image block as K multiplied by K, constructing an alternative atom library D of the defect-free light guide plate image, wherein each atom (each column) in the alternative atom library D corresponds to one image block which is K2X 1 column vector, noted
Figure GDA0001598722570000081
The alternative atom library is redundant, the column atoms have great correlation, and the alternative atom library needs to be trained and optimized further, so that each column atom in the alternative atom library D can be linearly represented by the rest column atoms in the alternative atom library D, and the represented coefficients are sparse. The dictionary training problem can be modeled as:
Figure GDA0001598722570000082
as described above
Figure GDA0001598722570000083
The solution of norm is an NP difficult problem, and the above l can be used according to the compressed sensing theory0Minimization problem conversion to l1Norm minimization problem:
Figure GDA0001598722570000084
solving the solution Z of the above formula by using Orthogonal Matching Pursuit (OMP) algorithm1,z2,…,zN}∈RN×NAnd counting all row coordinates which are not all zero in Z, and selecting the column atoms in D corresponding to the row coordinates as the column atoms of D', namely completing the optimization of the dictionary.
A defect-free image can be sparsely reconstructed with the preferred D 'of the complete dictionary, while for a defect image, it cannot be sparsely reconstructed with the preferred D' of the complete dictionary. For the image I to be detected, only the linear expressed coefficient of the image I under the optimized D' of the finished dictionary is needed to be solved, and whether the image contains defects can be judged according to the sparsity of the expressed coefficient.
It has been found that the orthogonal matching pursuit algorithm (OMP) does not necessarily yield its most sparse solution when used for sparse representation. Therefore, the present invention solves for the sparse representation coefficient X using the improved ROMP algorithm as follows:
and inputting an image block s to be reconstructed, a dictionary D', and the selected atomic number k.
And outputting a sparse representation coefficient X.
Step 1 is initialized. And partitioning the light guide plate body image H (x, y), wherein s is one block. Initial residual r0The method comprises the following steps of (i) setting an index set ^ theta, J ^ theta and iteration number t as 1;
step 2, 1 st atom selection. Calculating the residual error rtSearching k maximum values from the relation number u of the atoms in the dictionary D', and adding indexes corresponding to the coordinates of the maximum values into an index set J;
step 3, 2 nd atom selection (regularization). Calculating a local constraint weighting coefficient ω ═ dist (s, D '), where ω represents the similarity between residuals s and D', and dist (s, D ') [ dist (s, D') ]1),...,dist(s,dm)]TRepresenting residual s and dictionary diAn euclidean distance between (i ═ 1.., m). Then, the k maximum correlation coefficients u are weighted and the correlation is performedThe number satisfies the condition 2| u | ≧ u |maxAdding the atom of | to the index set Λ to update the support set;
step 4 updates the residual. Signal approximation by least squares method xt＝arg min‖s-D'Λxt－1‖2Calculating residual r under the new support sett＝s-D'Λxt；
Step 5 stop conditions. If | rt‖2＜10-1Or the number num (lambada) of the sparse coefficients is more than or equal to 6, skipping to the step (6); otherwise, returning to repeat the step (2) until the stop condition is met;
and 6, outputting the sparse representation coefficient X. The improved ROMP algorithm improves the regularization of atom selection by performing local constraint weighting on the correlation coefficients, thereby obtaining accurate X.
Finally, the coefficient X is expressed by a sparse table0The ratio of the norm to the original image size (called sparsity) is used as an evaluation function of Sparsity (SR) of the light guide plate body image H (x, y):
SR＝||X||0/mn
(5) judging whether the light guide plate has defects; executing the step 6;
if the SR value of the light guide plate body image H (X, y) is larger than a certain threshold value TH, the linear expression coefficient X of the image to be detected is not sparse enough, the image to be detected cannot be well reconstructed from the dictionary D', and therefore the image with the defects is obtained, the step 6 is executed; otherwise, if the image is a defect-free image, the light guide plate is judged to be a qualified product.
(6) Removing noise interference in the light guide plate body image H (x, y) by using mean value filtering to obtain a new image J (x, y); executing the step 7;
due to interference of electrical noise and the like, the light guide plate image inevitably has noise, and the invention adopts mean value filtering to avoid the influence of image noise on detection.
For each pixel point P in a given image H (x, y), its neighborhood S is taken. And setting that the S contains M pixels, and taking the weighted average value of the M pixels as the gray value of the image pixel P obtained after the processing. Using weighting of the grey level of each pixel in the neighborhood of a pixelThe method of replacing the original gray scale of the pixel with the average value is a neighborhood weighted average technology. The general shape of the neighborhood S is square, rectangle, cross, or the like. Let the noise n (i, j) be additive noise, and each point is uncorrelated, and it is expected to be 0, and the variance is σ2G (x, y) is an image that is not contaminated by noise, and the image H (x, y) containing noise is mean filtered as follows:
Figure GDA0001598722570000101
thus, the noise variance of the filtered image is as follows:
Figure GDA0001598722570000102
(7) because the color of the light guide point and the background are not distinguished strongly, a new image J (x, y) needs to be subjected to gray level conversion, the contrast is enlarged, the detection accuracy is improved, and an enhanced image K (x, y) is obtained; executing the step 8;
each point in the new image J (x, y) is processed with the following formula to obtain an image K (x, y):
K(x,y)＝a×j(x,y)+b
a and b are fixed values obtained by experiments, wherein a is 1.3, and b is 10.
(8) Performing OTSU threshold segmentation on the image K (x, y) obtained in the step 7, and segmenting the light guide point from the background; executing the step 9;
the OTSU algorithm is also called a maximum inter-class difference method, and divides an image into a background part and a foreground part according to the gray characteristic of the image. The larger the inter-class variance between the background and the foreground is, the larger the difference between the two parts constituting the image is, and the smaller the difference between the two parts is caused when part of the foreground is mistaken for the background or part of the background is mistaken for the foreground. Thus, a segmentation that maximizes the inter-class variance means that the probability of false positives is minimized. If the gray level of the gray image is L, the gray range is [0, L-1], and the optimal threshold of the image is calculated by using the OTSU algorithm as follows:
t＝Max[w0(t)×(u0(t)-u)2+w1(t)×(u1(t)-u)2]
wherein: when the threshold value of the division is t, w0As background proportion, u0As background mean value, w1As a foreground proportion, u1Is the foreground mean, and u is the mean of the light guide plate image. T with the maximum value of the expression is the optimal threshold value for segmenting the image.
And (3) segmenting the image K (x, y) into a background image and a foreground image by adopting an optimal threshold value, and concentrating the light guide points on the foreground image.
(9) According to the density degree of the light guide points, automatic partition detection of the image K (x, y) obtained in the step 7 is realized; executing the step 10;
according to the structure of the single-side light-incoming type LGP, the light guide plate can uniformly emit light through various light guide points with different densities and sizes. The light guide points of the single-side incident light type light guide plate are unevenly distributed, and the farther away from the light source side, the denser the light guide points are distributed. In order to solve the problem of false detection caused by different densities, partition processing is required.
According to the light guide point from sparse to dense, an image K (x, y) is averagely divided into N (i is 1,2, N) areas, and the area of each area is S. Obtaining the Nth threshold value according to the threshold value result in the step (8)iThe number of regional light guide points is MiFrom the formula
Figure GDA0001598722570000111
To obtain NiDensity P of the regioni. Setting a discrimination density value Qj(j ═ 1,2,3.. 9), for all regions, if (Q)j＜Pi＜Qj+1)U(Qj＜Pi+1＜Qj+1) Then N isi+Ni+1，NiRegion and Ni+1And merging the areas.
(10) Traversing the gray scale range of all the pixels of the light guide point region of the foreground image according to the light guide point region extracted in the step 8, and utilizing a formula
Figure GDA0001598722570000112
Calculating the average value G of the gray value of each light guide pointaveIn the formula NiThe total number of pixels of the ith light guide point, GiThe gray sum of all pixels of the ith light guide point is obtained; executing the step 11;
(11) set the maximum evaluation value GmaxAnd a minimum evaluation value GminIf G isave＞GmaxIf the light guide plate has bright spots, the light guide plate is directly judged to be an unqualified product; if G isave＜GminIf the light guide plate has dark spots, the light guide plate is directly judged to be an unqualified product; if G ismin≤Gave≤GmaxIf there is no bright or dark spot, step 12 is performed. And judging that the light guide plate is unqualified due to the existence of bright spots and dark spots.
(12) Because the light guide points in the high-density area are very close to each other, even some light guide points in the image partitioned in the step 9 are connected together, which causes great interference to detection, and therefore, the area needs to be corroded.
The invention adopts a round structural element for corrosion, X is an image to be processed (an image partitioned in step 9), B1Is a structural element, R1Is a processed image. Structural element B1The mathematical expression for the erosion of the image X is as follows:
Figure GDA0001598722570000121
where Θ represents the operator of corrosion and x refers to the translation of the set.
The basic process of corrosion operation is: moving structural element B in image X to be processed2If moved to and structural element B2When the sub-image is identical, the structural element B is added to the sub-image2The pixel corresponding to the origin of (A) is marked, and so on, so that the structural element B is2Moving in the image X to be processed, and finally the set of all marked pixels is the structural element B2As a result after the etching, an image after the etching is formed.
And (3) because the distance between the light guide points in the low-density area is relatively large, and defects such as pressure damage, foreign matters and the like are easy to miss detection, the expansion operation is carried out on the image partitioned in the step 9.
Let Y be the image to be processed (image partitioned in step 9), B2Is a structural element, BvIs B2Reflection of (2), R2For the processed image, the mathematical expression for the dilation operation is as follows:
Figure GDA0001598722570000122
where x is the amount of translation of the set,
Figure GDA0001598722570000123
representing a inflation operator.
The expansion operation begins with the structural element B2Performing reflection operation with the origin as the center to obtain BvThen shift B in the image Y to be processedvWhen B is presentvWhen there is at least one non-zero point of intersection with the image to be processed, then B isvUntil a complete image to be processed is processed, the set of the marked pixels and the original image to be processed is an expanded image (i.e. an image after morphological processing).
Then step 13 is executed;
(13) judging whether the light guide plate has no damage or foreign matters; analyzing step 12 all connected component features of the morphologically processed image, wherein the area S of each connected component is calculatedi。
Setting a determination value Smax. If S isi＞SmaxIf so, directly judging the light guide plate to be an unqualified product if the light guide plate has pressure damage or foreign matters; si≤SmaxThen step 14 is performed.
(14) And (5) judging the scratch defects.
And 14.1, performing fast Fourier transform on the light guide plate body image H (x, y). Step 14.2 is executed;
from step (3), it can be obtained that the size of the image H (x, y) is M × N, and the image H (x, y) is a periodic M × N discrete signal, and the fourier transform type is 2-DFT, which is expressed as follows:
Figure GDA0001598722570000131
wherein u-0, 1, 2.., M-1; v is 0,1,2,.., N-1, and u, v are frequency values. x, y are frequency values in the spatial domain, and M and N are the size of the digital image.
14.2, an ideal low-pass filter is generated. Step 14.3 is executed;
in the frequency domain, the high-frequency part represents the detail and texture information of the image; the low frequency part represents contour information of the image. If a low pass filter is used for a fine image, only low frequency signals will pass, and the result of the filtering will only leave contours. Therefore, fine light guide points in the light guide plate image can be filtered out by the low frequency filter operation so as to highlight the scratch defect.
Figure GDA0001598722570000132
Wherein D is0Representing the radius of the passband. D (u, v) is calculated in a manner that is the distance between two points;
Figure GDA0001598722570000133
wherein the size of the image is M N;
and 14.3, performing convolution operation on the light guide plate body image H (x, y) in a frequency domain by using the filter. Step 14.4 is executed;
setting an arbitrary point f (x, y) in the light guide plate body image H (x, y), wherein the point after convolution is g (x, y);
Figure GDA0001598722570000134
14.4, inverse fast Fourier transform, transforming the defect to the spatial domain. Step 14.5 is executed;
Figure GDA0001598722570000135
wherein u-0, 1, 2.., M-1; v is 0,1,2,.., N-1, and u, v are frequency values. x, y are frequency values in the spatial domain, and M and N are the size of the digital image.
14.5, finally, operating the image in the 14.4 by using a fixed segmentation threshold TH according to the following formula, segmenting the image, and executing the step 15;
Figure GDA0001598722570000136
where g (x, y) is the image after segmentation and TH is the segmentation threshold. Repeated algorithmic validation of the collected samples determined the optimal threshold TH-38 for the present experimental conditions.
(15) Traverse all region lengths L of the segmented image g (x, y)i(i is 0,1,2,3.. N), and a discrimination length criterion L is setmaxIf L isi＞LmaxIf the light guide plate has scratch defects, judging the light guide plate to be an unqualified product; if L isi≤LmaxAnd judging the product as a qualified product.
(16) Through the steps, all the defect areas of the light guide plate can be extracted, the defect areas are finally represented by means of the minimum external rectangle, and the mathematical characteristics of the defect areas are calculated.
For the light guide plate judged as unqualified product, the pixel point of the corresponding light guide plate body image H (x, y) is taken as the random variable value f (x, y), and the p + q moment of the interested region T (namely ROI) is
Figure GDA0001598722570000141
The centroid coordinate of the target area is (x)1,y1):
Figure GDA0001598722570000142
Moving the center of mass of the target to the origin position of the reference coordinate system to obtain the center distance:
Figure GDA0001598722570000143
similarly, find u00、u20、u02And the like. Then the length and width of the minimum bounding rectangle of the ROI:
Figure GDA0001598722570000144
an image containing the region of interest T is obtained.
Experiment one:
(1) using a line scan camera, a grayscale image as shown in fig. 3 is obtained;
(2) performing 7 × 7 mean filtering on the grayscale image to obtain a mean filtered image as shown in fig. 6 (the other steps between step 1 and step 2 adopt the steps and formulas of the above embodiment);
(3) carrying out gray scale transformation on the mean value filtering image according to the formula 1, and further expanding the contrast ratio by Pout(x,y)＝a×Pin(x, y) + b, wherein a is 1.3 and b is 10 (obtained by experiment);
(4) according to the density degree of the light guide points, the light guide plate image is automatically partitioned, as shown in fig. 7;
let N be 10 and Q be 3, the image is automatically divided into 4 regions (4 regions are obtained by experiment, and the other steps between step 4 and step 5 adopt the steps and formulas of the above embodiment)
(5) Performing automatic threshold segmentation on the partitioned image, and determining a segmentation threshold t by using the following formula;
t＝Max[w0(t)×(u0(t)-u)2+w1(t)×(u1(t)-u)2](obtained by experiment)
(6) The high-density region adopts corrosion operation, and the structural element B1A 2-pixel radius circle is adopted, and the following formula is used for corrosion to obtain a graph 9 (obtained by an experiment);
Figure GDA0001598722570000151
(7) the low-density region adopts expansion operation and structural element B2Using a 3 pixel radius circle, the expansion is performed using the following formula to obtain fig. 8 (obtained experimentally);
Figure GDA0001598722570000152
(8) analyzing the characteristics of the connected domain, and calculating the area S of all the regionsiIs provided with SmaxIf S is 90i＞SmaxPressing damage or foreign body defect;
(9) performing fast Fourier transform on the image according to the following formula;
Figure GDA0001598722570000153
(10) generating an ideal low-pass filter H (u, v);
Figure GDA0001598722570000154
(11) inverse fast Fourier transform, transforming the defect to the spatial domain;
Figure GDA0001598722570000155
(12) the fixed threshold TH is 38, and the scratch defects are segmented by using the following formula;
Figure GDA0001598722570000156
(13) minimum bounding rectangle M of extracted defect areapqDisplaying the defects;
Figure GDA0001598722570000157
finally, it is also noted that the above-mentioned lists merely illustrate a few specific embodiments of the invention. It is obvious that the invention is not limited to the above embodiments, but that many variations are possible. All modifications which can be derived or suggested by a person skilled in the art from the disclosure of the present invention are to be considered within the scope of the invention.

Claims (4)

1. The method for extracting the defects of the unilateral side light-incoming type light guide plate is characterized by comprising the following steps of:
(1) acquiring a light guide plate image F, and executing the step 2;
(2) extracting a light guide plate body image H (x, y) according to the light guide plate image F, and executing the step 3;
(3) acquiring the width M and the height N of the light guide plate body image H (x, y), and executing the step 4;
(4) performing defect rapid detection on the light guide plate body image H (x, y) by adopting a sparse representation method to obtain an SR value of the light guide plate body image H (x, y), and executing the step 5;
(5) if the SR value of the light guide plate body image H (x, y) is greater than the threshold TH and is therefore a defective image, execute step 6; otherwise, if the image is a defect-free image, the light guide plate is judged to be a qualified product;
(6) removing noise interference in the light guide plate body image H (x, y) by using mean value filtering to obtain a new image J (x, y); executing the step 7;
(7) carrying out gray level transformation on the new image J (x, y) to obtain an enhanced image K (x, y), and executing the step 8;
(8) performing OTSU threshold segmentation on the image K (x, y) obtained in the step 7, and segmenting the light guide point and the background to obtain a foreground image and a background image; executing the step 9;
(9) according to the density degree of the light guide points, automatic partition detection of the image K (x, y) obtained in the step 7 is realized; executing the step 10;
(10) traversing gray scale range of all light guide point region pixels of the foreground image, and utilizing a formula
Figure FDA0003201429910000011
Calculating the average value G of the gray value of each light guide pointaveIn the formula NiThe total number of pixels of the ith light guide point, GiThe gray sum of all pixels of the ith light guide point is obtained; executing the step 11;
(11) set the maximum evaluation value GmaxAnd a minimum evaluation value GminIf G isave＞GmaxIf the light guide plate has bright spots, the light guide plate is directly judged to be an unqualified product; if G isave＜GminIf the light guide plate has dark spots, the light guide plate is directly judged to be an unqualified product; if G ismin≤Gave≤GmaxIf no bright or dark spot exists, executing step 12;
(12) performing corrosion operation and expansion operation on the image partitioned in the step 9 to obtain a morphologically processed image, and then executing a step 13;
(13) analyzing all connected domain features of the morphologically processed image obtained in step 12, calculating the area S of each connected domain, and setting a determination value Smax(ii) a If S is presenti＞SmaxIf so, directly judging the light guide plate to be an unqualified product if the light guide plate has pressure damage or foreign matters; if all S are presenti≤SmaxThen go to step 14;
(14) segmenting the light guide plate body image H (x, y) to obtain a segmented image g (x, y); step 15 is executed;
(15) traversing all the segmented region lengths Li(i is 0,1,2,3.. N), and a discrimination length criterion L is setmaxIf L isi＞LmaxIf the light guide plate has scratch defects, judging the light guide plate to be an unqualified product; if L isi≤LmaxJudging the product as a qualified product;
(16) and extracting all defect areas of the light guide plate judged as the unqualified product, finally representing the defect areas by means of the minimum external rectangle, and calculating the mathematical characteristics of the defect areas.
2. The method for extracting defects of a single-sided side-entry light guide plate according to claim 1, wherein the step (4) comprises the steps of:
inputting image blocks s divided by a light guide plate body image H (x, y), constructing an obtained dictionary D', and selecting an atomic number k;
outputting a sparse representation coefficient X;
s1, partitioning the image to be detected into blocks by the same method, wherein S is one of the blocks; initial residual r0The index set ^ theta, J ═ theta and the iteration number t ═ 1;
s2 calculating residual rtSearching k maximum values from the relation number u of the atoms in the dictionary D', and adding indexes corresponding to the coordinates of the maximum values into an index set J;
s3 calculates a local constraint weighting factor ω ═ dist (S, D '), where ω represents the similarity between residuals S and D', dist (S, D ') [ dist (S, D')1),...,dist(s,dm)]TRepresenting residual s and dictionary di(i ═ 1.., m) in euclidean distance; then weighting the k maximum correlation coefficients u, and satisfying the condition of 2| u | ≧ | umaxAdding the atom of | to the index set Λ to update the support set;
s4 adopts least square method to approximate signal Xt＝arg min‖s-D'Λxt－1‖2Calculating residual r under the new support sett＝s-D'Λxt；
S5 if | rt‖2＜10-1Or the number num (Λ) of sparse coefficients is more than or equal to 6, the step S6 is skipped; otherwise, returning to repeat the step S2 until the stop condition is met;
s6 outputs a sparse representation coefficient X;
s7 sparse representation of l of coefficient X0The ratio of the norm to the size of the original image is used as an evaluation function of sparsity of the light guide plate body image H (x, y):
SR＝||X||0/M*N。
3. the method for extracting defects of a single-sided side-entry light guide plate according to claim 2, wherein the step (14) comprises the steps of:
14.1, performing fast Fourier transform on the light guide plate body image H (x, y) according to the size M N of the image H (x, y); step 14.2 is executed;
14.2, generating an ideal low-pass filter; step 14.3 is executed;
Figure FDA0003201429910000031
wherein D is0The radius of the pass band is represented, and the calculation mode of D (u, v) is the distance between two points;
Figure FDA0003201429910000032
14.3, performing convolution operation on the light guide plate body image H (x, y) in a frequency domain by using an ideal low-pass filter; setting an arbitrary point f (x, y) in the light guide plate body image H (x, y), wherein the point after convolution is g (x, y); step 14.4 is executed;
Figure FDA0003201429910000033
14.4, performing inverse fast Fourier transform on the light guide plate body image H (x, y) after convolution in the step (14.3), and transforming the defects into a space domain; step 14.5 is executed;
14.5, adopting a fixed threshold value TH, carrying out image segmentation on the image obtained in the step (14.4) by using the following formula, and executing a step 15;
Figure FDA0003201429910000034
where g (x, y) is the image after segmentation and TH is the segmentation threshold.
4. The method for extracting defects of a single-sided side-entry light guide plate according to claim 3, wherein the step (9) comprises the steps of:
according to the direction from sparse to dense of the light guide points, drawingDividing the image K (x, y) into N (i ═ 1, 2., N) regions on average, each region having an area S; obtaining the Nth threshold value according to the threshold value result in the step (8)iThe number of regional light guide points is MiFrom the formula
Figure FDA0003201429910000035
Obtaining the density P of the Ni regioni(ii) a Setting a discrimination density value Qj(j ═ 1,2,3.. 9), for all regions of image K (x, y), if (Q)j＜Pi＜Qj+1)U(Qj＜Pi+1＜Qj+1) Then N isi+Ni+1，NiRegion and Ni+1And merging the areas.
CN201810025157.8A 2018-01-11 2018-01-11 Method for extracting defects of single-side light-entering type light guide plate Active CN108346141B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810025157.8A CN108346141B (en) 2018-01-11 2018-01-11 Method for extracting defects of single-side light-entering type light guide plate

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810025157.8A CN108346141B (en) 2018-01-11 2018-01-11 Method for extracting defects of single-side light-entering type light guide plate

Publications (2)

Publication Number Publication Date
CN108346141A CN108346141A (en) 2018-07-31
CN108346141B true CN108346141B (en) 2021-10-22

Family

ID=62960478

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810025157.8A Active CN108346141B (en) 2018-01-11 2018-01-11 Method for extracting defects of single-side light-entering type light guide plate

Country Status (1)

Country Link
CN (1) CN108346141B (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109272525B (en) * 2018-09-17 2020-12-22 浙江理工大学 Light guide plate white point defect detection method
CN109886914B (en) * 2018-12-19 2020-01-17 浙江企银印务科技有限公司 Paper defect detection method based on local brightness invariance prior
CN110095474B (en) * 2019-04-12 2021-11-16 浙江大学台州研究院 Bandage winding device and method based on machine vision and defect screening
CN110530894B (en) * 2019-06-14 2022-05-13 杭州舜浩科技有限公司 Light guide plate bright spot defect detection method
CN110298825B (en) * 2019-06-17 2021-07-02 杭州舜浩科技有限公司 Method for detecting black spot defect of light guide plate
CN113340902A (en) * 2021-05-19 2021-09-03 沈阳风驰软件股份有限公司 Method and device for detecting surface defects of catalyst carrier, and electronic device
CN113706498A (en) * 2021-08-25 2021-11-26 广东奥普特科技股份有限公司 Visual detection device for lithium battery pole piece
CN113610852B (en) * 2021-10-10 2021-12-10 江苏祥顺布业有限公司 Yarn drafting quality monitoring method based on image processing
CN114240926B (en) * 2021-12-28 2022-12-13 湖南云箭智能科技有限公司 Board card defect type identification method, device and equipment and readable storage medium
CN114882026B (en) * 2022-07-08 2022-09-02 山东科尔自动化仪表股份有限公司 Sensor shell defect detection method based on artificial intelligence
CN114897906A (en) * 2022-07-14 2022-08-12 南通华烨塑料工业有限公司 Modified plastic appearance defect control method based on computer vision

Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TW200844429A (en) * 2007-05-15 2008-11-16 Chi-Hao Yeh An automatic optical inspection approach for detecting and classifying the surface defects on coating brightness enhancement film
CN102129573A (en) * 2011-03-10 2011-07-20 西安电子科技大学 SAR (Synthetic Aperture Radar) image segmentation method based on dictionary learning and sparse representation
CN102937599A (en) * 2012-10-25 2013-02-20 中国科学院自动化研究所 Non-destructive testing systems and method used for detecting a metal-containing object through X-ray detection
CN103499343A (en) * 2013-09-25 2014-01-08 武汉海达数云技术有限公司 Automatic classifying and identifying method for road line point cloud based on three-dimensional laser reflection strength
CN104574274A (en) * 2013-10-18 2015-04-29 财团法人资讯工业策进会 Image processing method and system applying same
CN104732215A (en) * 2015-03-25 2015-06-24 广西大学 Remote-sensing image coastline extracting method based on information vector machine
CN104931233A (en) * 2015-07-03 2015-09-23 东莞市彤光电子科技有限公司 Light guide plate detection apparatus
CN105069771A (en) * 2015-05-22 2015-11-18 同济大学 Active-contour-based SAR image after-calamity water body information extraction method
CN105787955A (en) * 2016-03-29 2016-07-20 河南森源电气股份有限公司 Sparse segmentation method and device of strip steel defect
CN105913419A (en) * 2016-04-07 2016-08-31 南京汇川图像视觉技术有限公司 TFT-LCD mura defect detection method based on ICA learning and multichannel fusion
CN106650770A (en) * 2016-09-29 2017-05-10 南京大学 Mura defect detection method based on sample learning and human visual characteristics
CN106733701A (en) * 2016-12-23 2017-05-31 郑州轻工业学院 Potato multi-stage combination dysnusia detecting system and detection method

Patent Citations (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
TW200844429A (en) * 2007-05-15 2008-11-16 Chi-Hao Yeh An automatic optical inspection approach for detecting and classifying the surface defects on coating brightness enhancement film
CN102129573A (en) * 2011-03-10 2011-07-20 西安电子科技大学 SAR (Synthetic Aperture Radar) image segmentation method based on dictionary learning and sparse representation
CN102937599A (en) * 2012-10-25 2013-02-20 中国科学院自动化研究所 Non-destructive testing systems and method used for detecting a metal-containing object through X-ray detection
CN103499343A (en) * 2013-09-25 2014-01-08 武汉海达数云技术有限公司 Automatic classifying and identifying method for road line point cloud based on three-dimensional laser reflection strength
CN104574274A (en) * 2013-10-18 2015-04-29 财团法人资讯工业策进会 Image processing method and system applying same
CN104732215A (en) * 2015-03-25 2015-06-24 广西大学 Remote-sensing image coastline extracting method based on information vector machine
CN105069771A (en) * 2015-05-22 2015-11-18 同济大学 Active-contour-based SAR image after-calamity water body information extraction method
CN104931233A (en) * 2015-07-03 2015-09-23 东莞市彤光电子科技有限公司 Light guide plate detection apparatus
CN105787955A (en) * 2016-03-29 2016-07-20 河南森源电气股份有限公司 Sparse segmentation method and device of strip steel defect
CN105913419A (en) * 2016-04-07 2016-08-31 南京汇川图像视觉技术有限公司 TFT-LCD mura defect detection method based on ICA learning and multichannel fusion
CN106650770A (en) * 2016-09-29 2017-05-10 南京大学 Mura defect detection method based on sample learning and human visual characteristics
CN106733701A (en) * 2016-12-23 2017-05-31 郑州轻工业学院 Potato multi-stage combination dysnusia detecting system and detection method

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
***国." TFT-LCD显示屏Mura缺陷自动光学检测算法研究".《中国优秀硕士学位论文全文数据库（电子期刊）信息科技辑》.2014, *

Also Published As

Publication number Publication date
CN108346141A (en) 2018-07-31

Similar Documents

Publication Publication Date Title
CN108346141B (en) Method for extracting defects of single-side light-entering type light guide plate
CN108416766B (en) Double-side light-entering type light guide plate defect visual detection method
CN112308832B (en) Bearing quality detection method based on machine vision
CN111693549B (en) Method for detecting and classifying defects of mobile phone cover plate glass
CN105894036B (en) A kind of characteristics of image template matching method applied to mobile phone screen defects detection
KR101635461B1 (en) System and Method for Automatically Detecting a Mura Defect using Mask Filtering in Wavelet Transform
CN113570605B (en) Defect detection method and system based on liquid crystal display panel
CN108876768B (en) Shadow defect detection method for light guide plate
Liu et al. Automatic inspection system of surface defects on optical IR-CUT filter based on machine vision
CN111612737B (en) Artificial board surface flaw detection device and detection method
CN109801286B (en) Surface defect detection method for LCD light guide plate
CN109102497B (en) High-resolution light guide plate image defect detection method
CN114511516B (en) Micro LED defect detection method based on unsupervised learning
WO2022038575A1 (en) Surface defect detection system
CN109064439B (en) Partition-based single-side light-entering type light guide plate shadow defect extraction method
Ma et al. An automatic detection method of Mura defects for liquid crystal display
CN107895371B (en) Textile flaw detection method based on peak coverage value and Gabor characteristics
CN112782179A (en) Method and system for detecting defects of light-reflecting surface of product
Dominguez-Nicolas et al. Indentation image analysis for Vickers hardness testing
CN109785290B (en) Steel plate defect detection method based on local illumination normalization
CN116337869A (en) Machine vision-based detection system for production of double-sided circuit board
Kamani et al. Automatic paint defect detection and classification of car body
CN110530894B (en) Light guide plate bright spot defect detection method
JP2007192688A (en) Flaw inspection method
CN113487569B (en) Complex background image defect detection method and system based on combination of frequency domain and space domain

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
TA01 Transfer of patent application right
TA01 Transfer of patent application right

Effective date of registration: 20210816

Address after: 311215 room 806-6, building 1, No. 371, Xingxing Road, Xiaoshan District, Hangzhou City, Zhejiang Province

Applicant after: Hangzhou Hengdian Technology Co.,Ltd.

Address before: 310018, No. 2, No. 5, Xiasha Higher Education Park, Hangzhou, Zhejiang

Applicant before: Zhejiang University of Technology

GR01 Patent grant
GR01 Patent grant