CN110264445B - Battery silk-screen quality detection method combining block template matching with morphological processing - Google Patents
Battery silk-screen quality detection method combining block template matching with morphological processing Download PDFInfo
- Publication number
- CN110264445B CN110264445B CN201910460803.8A CN201910460803A CN110264445B CN 110264445 B CN110264445 B CN 110264445B CN 201910460803 A CN201910460803 A CN 201910460803A CN 110264445 B CN110264445 B CN 110264445B
- Authority
- CN
- China
- Prior art keywords
- template
- difference
- region
- battery
- area
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 100
- 230000000877 morphologic effect Effects 0.000 title claims abstract description 35
- 238000012545 processing Methods 0.000 title claims abstract description 21
- 230000007547 defect Effects 0.000 claims abstract description 79
- 238000000034 method Methods 0.000 claims abstract description 61
- 238000000605 extraction Methods 0.000 claims abstract description 21
- 230000011218 segmentation Effects 0.000 claims abstract description 18
- 238000013519 translation Methods 0.000 claims abstract description 14
- 230000009466 transformation Effects 0.000 claims abstract description 13
- 238000012937 correction Methods 0.000 claims abstract description 11
- 230000007797 corrosion Effects 0.000 claims abstract description 9
- 238000005260 corrosion Methods 0.000 claims abstract description 9
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims abstract description 8
- 238000005215 recombination Methods 0.000 claims abstract description 4
- 230000006798 recombination Effects 0.000 claims abstract description 4
- 238000010845 search algorithm Methods 0.000 claims abstract description 4
- 238000003780 insertion Methods 0.000 claims abstract description 3
- 230000037431 insertion Effects 0.000 claims abstract description 3
- 230000002441 reversible effect Effects 0.000 claims abstract description 3
- 238000009825 accumulation Methods 0.000 claims description 13
- 238000004458 analytical method Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000005520 cutting process Methods 0.000 claims description 8
- 238000007639 printing Methods 0.000 claims description 6
- 230000000903 blocking effect Effects 0.000 claims description 5
- 238000012216 screening Methods 0.000 claims description 5
- 238000003745 diagnosis Methods 0.000 claims description 4
- 230000003252 repetitive effect Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000005530 etching Methods 0.000 claims description 2
- 238000000638 solvent extraction Methods 0.000 claims 3
- 230000000694 effects Effects 0.000 abstract description 11
- 238000010586 diagram Methods 0.000 description 9
- 238000004519 manufacturing process Methods 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 238000007650 screen-printing Methods 0.000 description 7
- 230000010339 dilation Effects 0.000 description 6
- 238000004891 communication Methods 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 3
- 230000002950 deficient Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 239000002184 metal Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 230000008521 reorganization Effects 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 101150098958 CMD1 gene Proteins 0.000 description 1
- 101100382321 Caenorhabditis elegans cal-1 gene Proteins 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000008451 emotion Effects 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
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
- G06T7/0006—Industrial image inspection using a design-rule based approach
-
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30144—Printing quality
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a battery silk-screen quality detection method combining block template matching with morphological processing, which is characterized in that a text part template is established based on a morphological template matching method; extracting a battery region by applying threshold segmentation and morphological corrosion; determining the inclination angle by solving the minimum circumscribed rectangle of the battery area, searching the minimum circumscribed rectangle by adopting an equi-interval rotation search algorithm as a final result, and performing reverse rotation on the deflection angle by utilizing affine transformation to correct and cut the battery; identifying the defect detection of the part of the picture-inserting by adopting a difference shadow method; then, the picture insertion, the character segmentation, the character extraction and correction, the character recombination and the extraction of the character skeleton to be detected are sequentially shielded, the defects of the character part are identified by adopting a difference shadow method, and the defects are detected by adopting the translation of a suspicious region for difference detection and the micro-rotation of the suspicious region for difference detection. The method has the advantages of high accuracy, low cost and certain promotion effect on the field of automatic detection of the silk-screen quality of the domestic battery.
Description
Technical Field
The invention belongs to the technical field of machine vision automatic surface detection, and particularly relates to a battery silk-screen quality detection method combining block template matching and morphological processing.
Background
The silk-screen content of the battery comprises various complex patterns such as picture-in-picture, Chinese, English, Korean, numbers, bar codes and the like, and covers a large amount of important information including production places, models, specifications, warnings, use contraindications and the like, and the content is generally printed on the surface of the battery through a silk-screen printing mode, so the silk-screen quality is extremely important. The defects of silk-screen printing can be caused due to the influence of factors such as equipment, process, personnel and the like, and the main defects are defect, deflection, blur, double image, dirt, color difference, abnormal position and the like. The traditional appearance detection method mainly depends on naked eyes and a magnifier to carry out manual detection, and is inevitably influenced by factors such as human emotion, surrounding environment noise, work concentration degree and the like, so that the accuracy and the real-time performance of a detection result are difficult to ensure, and therefore, the development of a set of detection algorithm and an automatic detection system is urgent.
In order to realize automatic detection of the silk-screen quality, scholars at home and abroad carry out a great deal of research, and a plurality of classical methods such as a global template matching method, a pixel-by-pixel hierarchical detection method, a neural network algorithm, a wavelet transformation detection method, a Gabor transformation algorithm, a feature extraction method and the like are developed. However, the above method mainly has two problems:
1) the algorithm is too complex, the detection consumes long time, and the method is not suitable for being applied to the detection on a production line in a factory;
2) the characteristics of the application object are single, and the universality of the method and the detection capability of the method on the complex object need to be enhanced.
The battery silk-screen printing types detected by the invention are numerous, and the applicable method of each pattern or character is different. The characteristics of the picture inserting part and the character part in silk screen printing are obviously different, the lines of the picture inserting part are simple, wider and thicker, and the lines of the character part are fine and complex and need higher processing precision. The traditional classical method can not obtain good effect.
Disclosure of Invention
The invention aims to solve the technical problem that the defects in the prior art are overcome, and the battery silk-screen quality detection method combining the block template matching with morphological processing is provided, so that the traditional global template matching mode is broken through, the battery silk-screen is divided into a character part and an illustration part, two types of sub-methods are respectively designed according to the characteristics of the two parts, and finally, the two types of sub-methods are organically combined. The method can be used for rapidly and accurately detecting the quality of the silk screen.
The invention adopts the following technical scheme:
the battery silk-screen quality detection method combining block template matching and morphological processing comprises the following steps:
s1, establishing an illustration part template by adopting a block template matching mode and establishing a character part template based on a morphological template matching method by taking each illustration as a unit;
s2, extracting a battery region by applying threshold segmentation and morphological corrosion;
s3, determining the inclination angle of the battery by solving the minimum external rectangle in the battery area, adopting an equal-interval rotation search algorithm, rotating at equal intervals from the horizontal direction by setting a rotation step length, searching the minimum external rectangle at each angle, wherein the minimum external rectangle is the final result, the corresponding rotation angle is the deflection angle of the battery, and performing reverse rotation on the deflection angle by affine transformation to correct and cut the battery;
s4, identifying the silk-screen version, cutting, extracting and correcting the picture to be inserted, and identifying the defect detection of the part of the picture to be inserted by adopting a difference shadow method; then, the picture insertion, the character segmentation, the character extraction and correction, the character recombination and the extraction of the character skeleton to be detected are sequentially shielded, the defects of the character part are identified by adopting a difference shadow method, and the defects are detected by adopting the translation of a suspicious region for difference detection and the micro-rotation of the suspicious region for difference detection.
Specifically, in step S1, storing the blocking information of the picture-inserting portion in each type of silk screen in the database in the form of a data file, where each picture-inserting corresponds to a sub-template ID, cutting the image according to the geometric size information of the sub-template obtained by the test, and then establishing sub-template data, where the establishment of the picture-inserting portion template specifically is:
s1011, assuming that S is m × n search graph, T is P × Q template graph, Si,jIs a sub-graph in the search graph, defining the absolute error:
wherein ε (i, j, s, T) is the absolute error definitional formula, T (s, T) is the pixel value of the template map at the (s, T) position,
and
respectively representing the mean values of the subgraph and the template, E is the mean value, the absolute error is the absolute value of the difference between corresponding positions after the subgraph and the template are respectively removed from the mean values, Si,jAnd (s, t) is the pixel value of the sub-image with the coordinate (i, j) at the upper left corner at the coordinate (i + s, j + t) position.
S1012, setting an error threshold Th for comparing with the error accumulated value of S1013;
s1013, randomly selecting non-repetitive pixel points in the template graph, calculating the absolute error with the current sub-graph, accumulating the errors, recording the accumulation times H when the error accumulation value exceeds Th, and counting the accumulation times H of all sub-graphs by using a table R.
Further, the number of times R (i, j) of the sub-graph S (i, j) with the coordinate (i, j) at the upper left corner is:
wherein, R (i, j) is the accumulated times of the subgraph S (i, j) with the coordinate of the upper left corner as (i, j).
Specifically, the establishment of the text part template specifically comprises the following steps:
s1021, shielding all the insets in the silk screen by using a rectangular shielding window according to the shape matching result of the part of the insets, enabling the gray value of an original area to be the same as the background gray value, eliminating redundant interference of the insets, positioning and decoding the bar codes in the original image by using a decoding algorithm, returning position and dimension information of the bar codes after decoding, and shielding the bar codes by using the rectangular shielding window;
s1022, segmenting the characters according to the columns by using the expansion operation in the mathematical morphology;
s1023, the rectangles obtained in the S1022 are sorted, after the sorting, the characters in each line can be obtained by sequentially making a difference between the rectangular area and the template image, and then the characters in each line are extracted by using a self-adaptive threshold segmentation method;
s1024, translating the template character area to a specified position, and recording translation data of each row or each column.
Specifically, step S3 specifically includes:
s301, setting an initial angle alpha 00 degree, the step length of the rotation angle is theta, and the Area array Area]；
S302, calculating the current angle alpha0Acquiring coordinate information of all points of the contour as the minimum circumscribed rectangle of the contour region under 0 degrees, screening the farthest point and the nearest point by taking the parallel direction of the current angle as a standard, and solving the distance d between the two points1The farthest point and the closest point are screened again by taking the vertical direction of the current angle as a standard, and the distance d between the two points is obtained2，d1And d2I.e. the length and width of the external rectangle, the screened 4 characteristic points can form four sides of the external rectangle, and the pair d1、d2And Area [0 ]]Recording;
s303, rotating the outline region for i times in sequence, wherein the angle alpha is formedi＝α0+ i × θ, the minimum bounding rectangle after rotation is obtained according to S302, and the length, width and Area [ i ] of the rotation are recorded in sequence]；
S304, when alpha is 90 degrees, stopping the rotation, and obtaining the Area array Area]Middle minimum element value Area [ j]And its corresponding angle alphajThe angle is the direction of the smallest circumscribed rectangle, α + j × θ.
Specifically, in step S4, the step of identifying the defects of the illustration part by using the difference shadow method specifically includes:
the definition operation form is:
C＝A-B
wherein, C is a difference region after difference, which can be represented by a set symbol as C ═ a-a ≈ B, a represents a target set, and B represents a structural element; assuming that the template area is M and the area to be detected is T, corroding the area M to obtain M1Swelling to obtain M2(ii) a Etching the region T to obtain T1Expansion to obtain T2The missing part of the silk screen to be detected is represented as M1-T2The multi-printing part of the silk screen to be detected is represented as T1-M2The difference result can eliminate the false defect of the contour.
Specifically, in step S4, the extraction of the text skeleton to be detected specifically includes:
and adopting a skeleton extraction algorithm based on a maximum disc, setting A to represent a target set and B to represent a structural element, wherein the morphological skeleton calculation can be given by the following formula:
wherein, the skeleton S (A) of the set A is composed of a skeleton subset Sk(A) Is formed by a skeleton subset Sk(A) The calculation formula is defined on the basis of the combination form of corrosion and open operation:
Sk(A)＝(AokB)-(AokB)oB
where K is 0, 1., K, AokB denotes the successive K etches of the set a by the structural element B, expressed as:
AokB＝(Ao(k-1)B)oB＝(...((AoB)oB)o...)oB
and K represents the maximum iteration number before the structural element B corrodes the set A into an empty set, and the structural element B corrodes the set A into the empty set when the iteration number exceeds K iterations.
Specifically, in step S4, the difference image method specifically identifies the text defect as follows:
respectively extracting a region T and a region M to be detected, extracting a region T framework T _ skelteon, expanding the framework to obtain T _ skeltration, and expanding the region T to obtain T _ diagnosis; extracting a framework M _ skelteon of the region M, expanding the framework to obtain M _ skeltration, and expanding the region M to obtain M _ diagnosis; and then, carrying out image difference, counting the number of the defects N1, and if N1 is not 0, carrying out suspicious region translation difference detection.
Further, the suspicious region translation difference detection specifically comprises:
translating the area T to be detected in four directions, namely, up, down, left and right, to obtain four translated areas T _01, T _02, T _03 and T _04, sequentially subtracting the four areas from the text template M to obtain four difference areas, and performing intersection operation on the four difference areas and the difference areas obtained by identifying defects by a difference shadow method to obtain an intersection area I1; and then screening the I1 by using the same defect analysis method to obtain a defect region, counting the number of the defects N2, and if N2 is not 0, carrying out slight rotation difference detection on the suspicious region.
Furthermore, the differential detection of the micro-rotation of the suspicious region specifically comprises:
and carrying out micro-rotation transformation on the rows or columns corresponding to the obtained defect regions to obtain two regions Rotate _01 and Rotate _02, sequentially carrying out difference on the two regions and the character template M, carrying out intersection operation on the obtained difference region and I1 again to obtain a final intersection region I2, carrying out defect analysis on the region to obtain a series of defect regions, and recording the number N3 and N3 of the defect regions as the final number of defects.
Compared with the prior art, the invention has at least the following beneficial effects:
the battery silk-screen quality detection method combining the block template matching with morphological processing can effectively detect silk-screen contents of different types, and has high accuracy and good real-time performance.
Furthermore, the purpose of establishing the block template is to accurately match each picture, so that the problem of insufficient registration precision of global matching is solved; and the sequential similarity matching detection has low calculation complexity and high matching precision.
Furthermore, the character template has the advantages of realizing the segmentation and recombination of single-line characters and improving the registration precision of the template characters and the characters to be detected.
Furthermore, the deflection angle of the battery is skillfully obtained by utilizing a minimum external rectangle algorithm, so that the battery is corrected and convenient for subsequent detection.
Furthermore, the interpolation defect identification by the differential photography method has the advantages that the interpolation defect identification method can be suitable for various complex patterns; the preprocessing of the template area and the illustration area can eliminate the interference of the false defects of the contour.
Furthermore, the character topological structure can be obtained by extracting the character skeleton, so that the character defect detection is not interfered by the uneven thickness of silk-screen lines.
Furthermore, in the step of identifying the character defects by the difference image method, the character defects can be accurately identified without generating false alarm by performing morphological processing on the skeleton.
Furthermore, the suspicious region translation error detection can eliminate the interference of process errors and prevent the false alarm caused by insufficient registration precision.
Furthermore, the slight rotation difference of the suspicious region can make up the problem of insufficient affine transformation precision during character inclination correction, and further eliminates false alarm.
In conclusion, the method can well complete the detection of the complex silk-screen printing of the battery, meets the production requirements of factories in real time, has high accuracy and lower cost, and has a certain promotion effect on the field of the automatic detection of the silk-screen printing quality of the battery in China.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
FIG. 1 is a raw image of a battery;
FIG. 2 is a diagram of a silk-screen segmented template;
FIG. 3 is a schematic diagram of matching;
FIG. 4 is a flow chart of the SSDA algorithm;
FIG. 5 is a pictorial shielding diagram;
FIG. 6 is a schematic diagram of character segmentation, wherein (a) is a schematic diagram of character binarization, and (b) is a schematic diagram of shape conversion;
FIG. 7 is a text reconfiguration diagram;
FIG. 8 is an original image;
FIG. 9 is a schematic drawing showing differentiation;
FIG. 10 is a pictorial detection flow diagram;
FIG. 11 is a flow chart of text detection;
FIG. 12 is a schematic view of subtraction preparation;
FIG. 13 is a difference image;
FIG. 14 is a schematic view of a modified process;
FIG. 15 is a hardware schematic;
FIG. 16 is a detection flow chart;
fig. 17 is a system architecture diagram.
Detailed Description
The invention provides a battery silk-screen quality detection method combining block templates with morphological processing, which divides the battery silk-screen into a character part and an illustration part, respectively designs two types of methods aiming at the characteristics of the two parts, and finally organically combines the two types of methods.
The invention discloses a battery silk-screen quality detection method combining block template matching and morphological processing, which comprises the following steps of:
s1, establishing template data and corresponding parameter configuration file
S101, creating a picture-inserting part template
And for the picture-inserting part, a mode of matching the block templates is adopted, so that the template is established by taking each picture-inserting part as a unit. As the types of the battery silk-screen which needs to be detected are numerous, the blocking information of the picture inserting part in each type of silk-screen needs to be stored in a database in a data file form, and each picture inserting corresponds to one sub-template ID.
Referring to fig. 2, the image as the template is an image with good printing quality, clear handwriting, and no defects. And cutting the image according to the geometric dimension information of the sub-template obtained by the experiment, and then establishing sub-template data. The template matching method adopted by the invention is a Sequential similarity Algorithm (SSDA) based on gray scale information.
The algorithm principle is as follows:
suppose S is an m × n search graph, T is a P × Q template graph, Si,jIs a sub-graph in the search graph (the initial position at the top left corner is (i, j)), obviously, i is more than or equal to 1 and less than or equal to m-P-1, and j is more than or equal to 1 and less than or equal to n-Q-1, as shown in FIG. 3.
S1011, defining an absolute error:
wherein ε (i, j, s, T) is the absolute error definitional equation, T (s, T) is the pixel value of the template map at position (s, T),
and
respectively representing the mean values of the subgraph and the template, E is the mean value, the absolute error is the absolute value of the difference between corresponding positions after the subgraph and the template are respectively removed from the mean values, Si,jAnd (s, t) is the pixel value of the sub-image with the coordinate (i, j) at the upper left corner at the coordinate (i + s, j + t) position.
S1012, setting an error threshold Th for comparing with the third step error accumulation value. The threshold value selects an optimal value according to the multiple matching effect;
s1013, randomly selecting non-repetitive pixel points in the template graph, calculating the absolute error with the current sub-graph, accumulating the errors, recording the accumulation times H when the error accumulation value exceeds Th, and counting the accumulation times H of all sub-graphs by using a table R.
The number of times R (i, j) of the sub-graph S (i, j) with the coordinate (i, j) at the upper left corner is as follows:
wherein, R (i, j) is the accumulated times of the subgraph S (i, j) with the coordinate of the upper left corner as (i, j). The analysis definition shows that the more deviated the sub-graph of the template, the faster the error accumulation speed is, the smaller the accumulation number H, the slower the threshold Th is reached, and vice versa.
Referring to fig. 4, in the calculation process, after the sum of the accumulated errors of the random points exceeds the threshold (the number of recorded accumulation H), the current sub-graph is discarded and the next sub-graph is calculated. And (5) after traversing all the sub-images, selecting the (i, j) sub-image corresponding to the maximum R value as a matching image. In a very special case, if R has a plurality of maximum values, the matching image with the smallest accumulated error is taken.
By applying the algorithm, the template matching function can be accurately completed, and the effect is good.
S102, establishing a character part template
The template matching method based on morphology is provided, a large amount of operations are not needed, the registration of complex and fine patterns can be completed, and the template establishing process is as follows:
s1021, shielding picture
Referring to fig. 5, according to the result of matching the shapes of the parts of the insets, all the insets in the silk screen are shielded by using a rectangular shielding window, so that the gray level of the original area is the same as the background gray level, and the redundant interference of the insets is eliminated. Positioning and decoding the bar code in the original image by using a decoding algorithm, returning the position and size information of the bar code after decoding, shielding the bar code by using a rectangular window by using the information, and obtaining the next image which is the image after shielding the inter-cut picture and the bar code;
s1022 character segmentation
Characters are generally square characters, and the characters are divided according to the columns by observing the silk-screen layout, so that the space separation with a larger distance is kept between the character columns and the columns. This can be achieved by using a dilation operation in mathematical morphology.
Dilation is a fundamental operation of mathematical morphological image processing. Three basic operations of dilation, erosion, and skeleton extraction are used herein. The principle of morphological dilation is simply the local maximum of the image, which expands the target boundary to the outside, and can be used to fill in some holes in the target area and eliminate the small particle noise contained in the target area.
According to the morphological expansion principle, the expansion effect is best by using structural elements similar to characters in shape, so rectangular structural elements are selected. If the whole column extraction is desired, the vertical direction needs to be expanded to a greater degree so as to enable the characters to be adhered into a column; the horizontal direction should avoid the adhesion of characters in different rows to cause the failure of extraction, so the horizontal direction is slightly expanded. By combining the consideration and the experimental test result, the best segmentation effect can be achieved by adopting the rectangular elements with the width of 4 pixels and the length of 50 pixels. Similarly, for the characters moving in the horizontal direction, the rectangular expansion effect with the width of 50 pixels and the length of 4 pixels is the best.
At this point each row or column of text has stuck to become a band region. Then, a morphological shape conversion algorithm is adopted to convert the shape of the strip area into a regular rectangle, so that each generated rectangular area contains the position information of the corresponding row or column characters, and the character segmentation is realized, as shown in fig. 6.
The irregular area is converted into a regular area, such as a rectangle, an ellipse, a circumscribed circle, etc., by using a shape conversion algorithm. In the present invention, a rectangle transformation algorithm is used, that is, the minimum bounding rectangle of each stripe region in the vertical direction is obtained, as shown in fig. 6 (b).
S1023, character extraction and correction
And (3) sequencing the rectangles obtained in the previous step according to a certain rule, sequentially subtracting the rectangular region from the template image after sequencing to obtain each line of characters, and then extracting each line of characters by using a self-adaptive threshold segmentation method. Because the text part is complicated and fine, and a defect false report may be caused by slight deviation, accurate registration of the template text and the text to be detected is required. Therefore, the extracted characters are subjected to inclination correction by affine transformation, so that the characters are in a strictly horizontal (0 degree) or vertical (90 degrees) position, and subsequent detection is facilitated.
S1024, character translation reorganization
The corrected template characters are still basically positioned at the original positions, the character positions are unstable due to the existence of printing errors, the direct comparison detection effect is not ideal, the text translates the template character areas, translates the template character areas to the designated positions, and records the translation data of each row or each column, so that the subsequent defect marking is facilitated. The text template is also completed by this time, as shown in fig. 7.
S2, extracting battery area
Referring to fig. 8, the original image view field content includes two interference regions, namely a background stage and a metal electrode plate, in the actual detection process, the battery is grabbed by a mechanical arm, and the position and the angle of the battery placed in each detection are different due to the influence of factors such as control precision and vibration, wherein the black background is the stage, the red circle mark part is the metal electrode plate, and the two parts both belong to regions unrelated to the detection, and the extraction of the battery region can be realized by applying threshold segmentation, morphological corrosion and the like.
S3, battery correction and cutting
The battery correction refers to a battery area, and thus the current picture tilt angle needs to be obtained. Because the shape of the battery cell area is similar to a rectangle, the inclination angle of the battery cell area is determined by solving the minimum external rectangle of the battery area. The method adopts an equispaced rotation search algorithm, the rotation is performed at equal intervals from the horizontal direction by setting the rotation step length, the minimum external rectangle is searched under each angle, the minimum area of the external rectangles is the final result, and the rotation angle is the deflection angle of the battery.
The method comprises the following specific steps:
s301, setting an initial angle alpha 00 degree, the step length of the rotation angle is theta, and the Area array Area]；
S302, calculating the current angle alpha0The minimum circumscribed rectangle of the outline region under 0 degrees is specifically: obtaining coordinate information of all points of the contour, screening the farthest point and the closest point (from the original point) by taking the parallel direction of the current angle as a standard, and solving the distance d between the two points1The farthest point and the closest point are screened again by taking the vertical direction of the current angle as a standard, and the distance d between the two points is obtained2，d1And d2Namely the length and the width of the external rectangle, and the four sides of the external rectangle can be formed by the screened 4 characteristic points. To d1、d2And Area [0 ]]Recording;
s303, rotating the outline region for i times in sequence, wherein the angle alpha is formedi＝α0+ i × θ, calculating the minimum bounding rectangle after rotation according to the second step method, and recording the length, width and Area [ i ] of the rotation]；
S304, after alpha is equal to 90 DEGAnd the rotation is terminated. Obtaining Area array Area]Middle minimum element value Area [ j]And its corresponding angle alphajThe angle is the direction of the smallest circumscribed rectangle, α + j × θ.
The obtained external rectangular direction is the deflection angle of the battery, and the battery is reversely rotated by-alpha by affine transformationjThus, battery calibration is achieved. Irrelevant areas still exist around the battery silk-screen area, so the battery silk-screen area is cut down and extracted by setting appropriate geometric parameters, and redundancy is reduced.
S4, detecting defects, namely firstly inserting pictures and then writing
S401, picture interpolation detection
Referring to fig. 10, the method includes the following steps:
s4011, identifying silk-screen version
The method has the following specific steps that the specific method is that the partitioned templates are used for matching, patterns identical to the template patterns can be found in the image to be detected, and a group of data is obtained at the same time: the center coordinates of the matching pattern, the deflection angle, and the matching score. Determining whether the template is matched with the silk-screen version to be detected or not according to the matching scores, wherein the correct version is determined when the score is high;
s4012, cutting, extracting and correcting the picture
And after template matching, obtaining the central coordinates and deflection angles of each picture on the image to be detected, and cutting out the picture by using a rectangular frame by using the data. And then extracting the picture-inserting region by a threshold segmentation method. Performing inclination correction on the interpolation region by affine transformation according to the deflection angle to obtain an interpolation region with a standard angle, wherein the interpolation region can be used for further detection;
s4013, defect identification by difference image method
The core of the difference image method is that the image is subjected to difference operation, and the difference operation is applied to a binary image or a contour region.
Referring to fig. 9, the operation is defined as:
C＝A-B
wherein, C is a difference region after difference, which can be represented by a set symbol as C ═ a-a ≈ B, a represents a target set, and B represents a structural element;
and if the template image is M and the image to be detected is T, the silk-screen missing part can be expressed as M-T, and the silk-screen multi-printing part can be expressed as T-M.
Analyzing the difference result to determine whether the defect exists or not; if the difference is directly made, a plurality of false defect areas are generated, which are called contour false defects due to the fact that screen printing is not uniform and inaccurate and line thicknesses are different, and therefore, when multi-print detection (T-M) is carried out, the T area is corroded, and the M area is expanded; during skip printing detection (M-T), the M area is corroded, and the T area is expanded. And then carrying out region communication on the difference region subjected to difference by using a region communication algorithm, setting a reasonable area threshold (taking the condition that no visual obvious defect is caused as a standard), and taking the region exceeding the threshold as a defect region. The region extraction adopts an adaptive threshold segmentation method, has certain adaptability to different illumination conditions, and can ensure the detection accuracy.
S402, character defect detection
Please refer to fig. 11, which includes the following steps:
s4021 and shielding picture
Shielding the matched picture with a rectangular shielding window to eliminate interference;
s4022, character segmentation
The method is the same as the template manufacturing method;
s4023, character extraction and correction
The method is the same as the template manufacturing method;
s4024, character reorganization
The method is the same as the template manufacturing method;
s4025, extracting the character skeleton to be detected
Because the text content is complex, the details are more, the strokes are different in thickness and the geometric dimension is small, the complete font is easy to damage by using common morphological corrosion, and the defect is formed to cause misjudgment, the skeleton outline of the text is extracted by using a skeleton extraction algorithm. Not only ensures the complete outline of the font, but also ensures that the detection is not influenced by the thickness of the font, and greatly improves the detection accuracy.
Skeleton means that the target is represented by a single-pixel wide thin line without changing the target topology. The skeleton of the target has the same number of connected components and holes as the target itself, in short, the skeleton maintains the euler number of the target. At present, a plurality of different skeleton definitions and skeleton extraction algorithms exist, and the invention adopts a skeleton extraction algorithm based on a maximum disc. Assuming that a represents the target set and B represents the structural element, this morphological skeleton calculation can be given by:
the above formula shows that the skeleton S (A) of the set A is composed of a skeleton subset Sk(A) The union of (a). Skeleton subset Sk(A) The calculation formula is defined on the basis of the combination form of corrosion and open operation:
Sk(A)＝(AokB)-(AokB)oB
where K is 0, 1., K, AokB denotes the successive K etches of the set a by the structural element B, which can be expressed as:
AokB＝(Ao(k-1)B)oB＝(...((AoB)oB)o...)oB
k is the calculation times of the skeleton subset, and the mathematical expression is as follows:
and K represents the maximum iteration number before the structural element B corrodes the set A into an empty set, and the structural element B corrodes the set A into the empty set when the iteration number exceeds K iterations.
By applying the algorithm, the morphological framework of the character to be detected can be accurately extracted.
S4026, identifying defects by difference shadow method
At this time, a series of preprocessing operations are required to obtain two partial regions to be differenced, and referring to fig. 12, for a flow chart of a preparation stage of difference image detection, the regions to be detected T and M are respectively extracted, and four regions required for image differencing are obtained: m _ Skeldilation, M _ dilation, T _ Skeldilation, T _ dilation; the reason for expanding the framework is that the framework regions are all single-pixel wide, and the small area is difficult to detect; then, performing image difference; referring to FIG. 13, the defect analysis method is the same as that of the inset portion and will not be repeated here. The number of defects N1 is counted, and if N1 is not 0, S4027 is performed.
S4027, suspicious region translation difference detection
Due to silk-screen process errors, rows or columns of suspected defects of S4026 need to be subjected to key detection, and a large number of experiments show that position errors possibly exist in character areas after registration, namely deviations of about 1-3 pixels, so that an algorithm needs to be designed to solve the problem in order to prevent false alarm.
And translating the to-be-detected region T in four directions, namely up, down, left and right, to obtain four translated regions T _01, T _02, T _03 and T _04, wherein the translation distance is recommended to be about two pixels, and the specific numerical value is determined by the experimental effect. And (4) sequentially carrying out difference on the four areas and the character template M to obtain four difference areas, and carrying out intersection operation on the four difference areas and the difference area obtained in the step 6 to obtain an intersection area I1. Then, a defect region is obtained by screening from I1 by using the same defect analysis method, and the defect number N2 is counted. As shown in fig. 14, if N2 is not 0, S4028 is performed.
S4028, detecting the difference of the suspicious region by micro rotation.
As can be seen from a large number of experimental results, there are still few false alarms after the error analysis. The analysis reason is that a micro angle rotation error exists between the region to be detected and the template region, and the error is caused by the fact that the affine transformation correction precision is insufficient. In order to enable the detection result to reach the most accurate degree, the rows or the columns corresponding to the defect regions obtained in the S4027 are subjected to micro rotation transformation, for example, the angle is 0.6 degrees, the specific numerical value is determined by the silk-screen experiment effect of the type, and the direction is clockwise or counterclockwise.
Two regions, Rotate _01 and Rotate _02, are obtained, and are sequentially differed from the character template M, and the obtained difference region is subjected to intersection operation with I1 again to obtain a final intersection region I2, as shown in fig. 14. The defect analysis of the area can obtain a series of defect areas, and the number of the defect areas is recorded as N3, N3 is the final number of defects.
The detection result reaches the optimal state, and the alarm is not missed or mistakenly reported. The method only corrects the suspicious region for multiple times by using the idea of grading detection, thereby not only improving the accuracy of the algorithm, but also ensuring the real-time property. The multi-print defect detection is taken as an example to illustrate, and the missing print defect detection only needs to follow the same method.
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 15, the platform hardware includes an industrial personal computer, a PLC device, a manipulator, a feeding and sorting device, a light source controller, camera equipment, and a communication line; the co-control machine is connected with the light source through the light source controller and connected with the camera equipment, the industrial personal computer is respectively connected with the feeding belt, the manipulator, the non-defective product conveying belt and the defective product conveying belt through the PLC device, and the manipulator is used for detecting the battery.
Referring to fig. 16, the battery silk-screen quality detection method combining block template matching and morphological processing of the present invention is applied to factory assembly line production, and specifically includes the following steps:
step 1: start detection system
Step 2: initialization parameter configuration
Initializing parameters of a detection system according to the type of the battery, wherein the parameters comprise a template ID, a threshold parameter, a detection algorithm parameter and other information;
and 3, step 3: grabbing and feeding
The detection software sends a feeding command Cmd1, the PLC receives the command, grabs the battery to be detected from the feeding area and moves to the area to be detected, and the detection is waited;
and 4, step 4: defect detection
Sequentially detecting the battery picture and characters according to an algorithm, marking defects, and returning a detection result to a detection software database for storage;
and 5, step 5: battery sorting
And the detection software determines whether the battery is good according to the detection data. If the good products are good products, sending a good product sorting command Cmd2 to the PLC; if the goods are not good, a defective sorting command Cmd3 is sent to the PLC, and a sorting completion command is returned to the host.
And 6, step 6: PLC reset
After receiving the sorting completion command, the host computer sends out a PLC reset command Cmd4, commands the gripping device to return to the initial position of the feeding area, and returns a reset completion command to the host computer;
and 7, step 7: next cycle detection
And after receiving the reset completion command, the host sends a feeding command to the PLC device, the equipment starts the next period of detection, and the steps 3 to 7 are repeated.
Referring to fig. 17, the detection system according to the present invention includes four parts, which are respectively:
1) user information management
The content comprises common user login information, administrator login information, a user database and a project history database;
2) core detection algorithm
The detection algorithm detailed in section 3 is integrated and embedded into software, and comprises an illustration detection algorithm and a character detection algorithm;
3) system parameter configuration
The parameters comprise a plurality of key parameters such as a detection mode (off-line detection or on-line detection), a communication mode, detection parameters, template parameters, image acquisition equipment parameters and the like;
4) result display and output
The part of contents mainly comprise defect statistics, defect visual display, detection result data storage, generation of related real-time data on a production line and the like, and are used for visually displaying defect positions, defect quantity, detection progress and the like on a software interface, so that a worker can conveniently check the defect positions, the defect quantity, the detection progress and the like.
The above-mentioned contents are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereby, and any modification made on the basis of the technical idea of the present invention falls within the protection scope of the claims of the present invention.
Claims (8)
1. The battery silk-screen quality detection method combining block template matching and morphological processing is characterized by comprising the following steps of:
s1, with each picture-inserting as a unit, establishing a picture-inserting part template in a partitioning template matching mode, establishing a text part template based on a morphological template matching method, storing partitioning information of a picture-inserting part in each type of silk screen in a database in the form of a data file, enabling each picture-inserting to correspond to a sub-template ID, cutting the image according to the geometric size information of the sub-template obtained by a test, then establishing sub-template data, wherein the establishment of the picture-inserting part template specifically comprises the following steps:
s1011, assuming that S is m × n search graph, T is P × Q template graph, Si,jIs a sub-graph in the search graph, defining the absolute error:
wherein ε (i, j, s, T) is the absolute error definitional formula, T (s, T) is the pixel value of the template map at the (s, T) position,
and
respectively representing the mean values of the subgraph and the template, E is the mean value, the absolute error is the absolute value of the difference between corresponding positions after the subgraph and the template are respectively removed from the mean values, Si,j(s, t) is the pixel value of the sub-image with the coordinate (i, j) at the upper left corner at the coordinate (i + s, j + t) position;
s1012, setting an error threshold Th for comparing with the error accumulated value of S1013;
s1013, randomly selecting non-repetitive pixel points in the template graph, calculating the absolute error with the current sub-graph, accumulating the errors, recording the accumulation times H when the error accumulation value exceeds Th, and counting the accumulation times H of all sub-graphs by using a table R;
the number of times R (i, j) of the sub-graph S (i, j) with the coordinate (i, j) at the upper left corner is as follows:
wherein Th is a threshold parameter;
s2, extracting a battery region by applying threshold segmentation and morphological corrosion;
s3, determining the inclination angle of the battery by solving the minimum external rectangle in the battery area, adopting an equal-interval rotation search algorithm, rotating at equal intervals from the horizontal direction by setting a rotation step length, searching the minimum external rectangle at each angle, wherein the minimum external rectangle is the final result, the corresponding rotation angle is the deflection angle of the battery, and performing reverse rotation on the deflection angle by affine transformation to correct and cut the battery;
s4, identifying the silk-screen version, cutting, extracting and correcting the picture to be inserted, and identifying the defect detection of the part of the picture to be inserted by adopting a difference shadow method; then, the picture insertion, the character segmentation, the character extraction and correction, the character recombination and the extraction of the character skeleton to be detected are sequentially shielded, the defects of the character part are identified by adopting a difference shadow method, and the defects are detected by adopting the translation of a suspicious region for difference detection and the micro-rotation of the suspicious region for difference detection.
2. The battery silk-screen quality detection method combining block template matching and morphological processing as claimed in claim 1, wherein the text portion template establishment specifically comprises:
s1021, shielding all the insets in the silk screen by using a rectangular shielding window according to the shape matching result of the part of the insets, enabling the gray value of an original area to be the same as the background gray value, eliminating redundant interference of the insets, positioning and decoding the bar codes in the original image by using a decoding algorithm, returning position and dimension information of the bar codes after decoding, and shielding the bar codes by using the rectangular shielding window;
s1022, segmenting the characters according to the columns by using the expansion operation in the mathematical morphology;
s1023, the rectangles obtained in the S1022 are sorted, after the sorting, the characters in each line can be obtained by sequentially making a difference between the rectangular area and the template image, and then the characters in each line are extracted by using a self-adaptive threshold segmentation method;
s1024, translating the template character area to a specified position, and recording translation data of each row or each column.
3. The battery silk-screen quality detection method combining the blocking template matching and the morphological processing as claimed in claim 1, wherein step S3 specifically comprises:
s301, setting an initial angle alpha00 degree, the step length of the rotation angle is theta, and the Area array Area]；
S302, calculating the current angle alpha0Acquiring coordinate information of all points of the contour as the minimum circumscribed rectangle of the contour region under 0 degrees, and paralleling the minimum circumscribed rectangle with the current angleThe farthest point and the nearest point are screened out for the standard, and the distance d between the two points is obtained1The farthest point and the closest point are screened again by taking the vertical direction of the current angle as a standard, and the distance d between the two points is obtained2，d1And d2I.e. the length and width of the external rectangle, the screened 4 characteristic points can form four sides of the external rectangle, and the pair d1、d2And Area [0 ]]Recording;
s303, rotating the outline region for i times in sequence, wherein the angle alpha is formedi＝α0+ i × θ, the minimum bounding rectangle after rotation is obtained according to S302, and the length, width and Area [ i ] of the rotation are recorded in sequence]；
S304, when alpha is 90 degrees, stopping the rotation, and obtaining the Area array Area]Middle minimum element value Area [ j]And its corresponding angle alphajThe angle is the direction of the smallest circumscribed rectangle, α + j × θ.
4. The battery silk-screen quality detection method combining the blocking template matching and the morphological processing as claimed in claim 1, wherein in step S4, the defect detection for identifying the part of the illustration by using the difference shadow method specifically comprises:
the definition operation form is:
C＝A-B
wherein, C is the difference region after difference, and can be represented as C-a ═ B by the set symbol; assuming that the template area is M and the area to be detected is T, corroding the area M to obtain M1Swelling to obtain M2(ii) a Etching the region T to obtain T1Expansion to obtain T2The missing part of the silk screen to be detected is represented as M1-T2The multi-printing part of the silk screen to be detected is represented as T1-M2And performing difference to eliminate false contour defects, wherein A represents a target set and B represents a structural element.
5. The battery silk-screen quality detection method combining the block template matching with the morphological processing as claimed in claim 1, wherein in step S4, the extraction of the character skeleton to be detected specifically comprises:
and adopting a skeleton extraction algorithm based on a maximum disc, setting A to represent a target set and B to represent a structural element, wherein the morphological skeleton calculation can be given by the following formula:
wherein, the skeleton S (A) of the set A is composed of a skeleton subset Sk(A) Is formed by a skeleton subset Sk(A) The calculation formula is defined on the basis of the combination form of corrosion and open operation:
Sk(A)＝(AοkB)-(AοkB)οB
where K is 0, 1., K, a omicron kB denotes the successive K corrosions of the structural element B to the set a, expressed as:
AοkB＝(Aο(k-1)B)οB＝(...((AοB)οB)ο...)οB
and K represents the maximum iteration number before the structural element B corrodes the set A into an empty set, and the structural element B corrodes the set A into the empty set when the iteration number exceeds K iterations.
6. The battery silk-screen quality detection method combining the block template matching with the morphological processing as claimed in claim 1, wherein in step S4, the difference image method for identifying the text defects specifically comprises:
respectively extracting a region T and a region M to be detected, extracting a region T framework T _ skelteon, expanding the framework to obtain T _ skeltration, and expanding the region T to obtain T _ diagnosis; extracting a framework M _ skelteon of the region M, expanding the framework to obtain M _ skeltration, and expanding the region M to obtain M _ diagnosis; and then, carrying out image difference, counting the number of the defects N1, and if N1 is not 0, carrying out suspicious region translation difference detection.
7. The battery silk-screen quality detection method combining blocking template matching and morphological processing as claimed in claim 6, wherein the suspicious region translation difference detection specifically comprises:
translating the area T to be detected in four directions, namely, up, down, left and right, to obtain four translated areas T _01, T _02, T _03 and T _04, sequentially subtracting the four areas from the text template M to obtain four difference areas, and performing intersection operation on the four difference areas and the difference areas obtained by identifying defects by a difference shadow method to obtain an intersection area I1; and then screening the I1 by using the same defect analysis method to obtain a defect region, counting the number of the defects N2, and if N2 is not 0, carrying out slight rotation difference detection on the suspicious region.
8. The battery silk-screen quality detection method combining partitioning template matching and morphological processing as claimed in claim 7, wherein the suspicious region micro-rotation difference detection specifically comprises:
and carrying out micro-rotation transformation on the rows or columns corresponding to the obtained defect regions to obtain two regions Rotate _01 and Rotate _02, sequentially carrying out difference on the two regions and the character template M, carrying out intersection operation on the obtained difference region and I1 again to obtain a final intersection region I2, carrying out defect analysis on the region to obtain a series of defect regions, and recording the number N3 and N3 of the defect regions as the final number of defects.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910460803.8A CN110264445B (en) | 2019-05-30 | 2019-05-30 | Battery silk-screen quality detection method combining block template matching with morphological processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910460803.8A CN110264445B (en) | 2019-05-30 | 2019-05-30 | Battery silk-screen quality detection method combining block template matching with morphological processing |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110264445A CN110264445A (en) | 2019-09-20 |
| CN110264445B true CN110264445B (en) | 2021-09-07 |
Family
ID=67915972
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910460803.8A Active CN110264445B (en) | 2019-05-30 | 2019-05-30 | Battery silk-screen quality detection method combining block template matching with morphological processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110264445B (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111275697B (en) * | 2020-02-10 | 2022-04-22 | 西安交通大学 | Battery silk-screen quality detection method based on ORB feature matching and LK optical flow method |
| CN111986190B (en) * | 2020-08-28 | 2022-03-18 | 哈尔滨工业大学（深圳） | Printed matter defect detection method and device based on artifact elimination |
| CN112926438B (en) * | 2021-02-22 | 2024-04-05 | 深圳中科飞测科技股份有限公司 | Detection method and device, detection equipment and storage medium |
| CN113109348B (en) * | 2021-03-12 | 2022-03-29 | 华南理工大学 | Paddle image transfer printing defect identification method based on machine vision |
| CN113111868B (en) * | 2021-03-16 | 2023-08-18 | 广州大学 | Character defect detection method, system, device and storage medium |
| CN113189505B (en) * | 2021-03-26 | 2022-06-28 | 深圳市磐锋精密技术有限公司 | Mobile phone battery state monitoring system based on big data |
| CN113888487A (en) * | 2021-09-23 | 2022-01-04 | 广东工业大学 | Matching method of multi-surface flexible printing label design draft and printing head draft |
| CN114964388B (en) * | 2022-08-02 | 2022-10-21 | 托伦斯半导体设备启东有限公司 | Method for detecting defects of shell material of slurry pump based on optical technology |
| CN115063421B (en) * | 2022-08-16 | 2022-10-28 | 成都数联云算科技有限公司 | Pole piece region detection method, system and device, medium and defect detection method |
| CN115423783A (en) * | 2022-09-13 | 2022-12-02 | 优层智能科技(上海)有限公司 | Quality inspection method and device for photovoltaic module frame and junction box glue filling |
| CN116958137B (en) * | 2023-09-19 | 2023-12-12 | 易启科技(吉林省)有限公司 | New energy battery surface defect detection method and device |
| CN117095009B (en) * | 2023-10-20 | 2024-01-12 | 山东绿康装饰材料有限公司 | PVC decorative plate defect detection method based on image processing |
| CN118072319B (en) * | 2024-01-12 | 2024-08-23 | 人民教育出版社有限公司 | Image-text printing quality analysis system based on machine vision |
| CN118379279B (en) * | 2024-06-21 | 2024-11-08 | 杭州汇萃智能科技有限公司 | Print pattern defect detection method, system and readable storage medium |
| CN118429353A (en) * | 2024-07-05 | 2024-08-02 | 浙江赫宇包装科技有限公司 | Method and system for detecting defects of printed labels |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103363893B (en) * | 2012-03-26 | 2016-03-30 | 新疆农业大学 | A kind of red fuji apple size detection method |
| CN103617627A (en) * | 2013-12-11 | 2014-03-05 | 天津工业大学 | Detection method for defects of printing characters on relay based on image registration technology |
| CN108355987B (en) * | 2018-01-08 | 2019-10-11 | 西安交通大学 | A kind of screen printing of battery quality determining method based on piecemeal template matching |
-
2019
- 2019-05-30 CN CN201910460803.8A patent/CN110264445B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN110264445A (en) | 2019-09-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110264445B (en) | Battery silk-screen quality detection method combining block template matching with morphological processing | |
| CN108918536B (en) | Tire mold surface character defect detection method, device, equipment and storage medium | |
| US6438265B1 (en) | Method of binarization in an optical character recognition system | |
| CN113689429B (en) | Wood board defect detection method based on computer vision | |
| CN112364834A (en) | Form identification restoration method based on deep learning and image processing | |
| WO2021190155A1 (en) | Method and apparatus for identifying spaces in text lines, electronic device and storage medium | |
| JP2021527256A (en) | Systems and methods for detecting and classifying patterns in images with a vision system | |
| CN111626145A (en) | Simple and effective incomplete form identification and page-crossing splicing method | |
| CN111738252B (en) | Text line detection method, device and computer system in image | |
| CN113063802A (en) | Printed label defect detection method and device | |
| CN116245882A (en) | Circuit board electronic element detection method and device and computer equipment | |
| CN112560538A (en) | Method for quickly positioning damaged QR (quick response) code according to image redundant information | |
| CN109977714A (en) | A kind of integrated vision positioning method of the more QR codes of warehoused cargo | |
| CN108985294B (en) | Method, device and equipment for positioning tire mold picture and storage medium | |
| JP4492258B2 (en) | Character and figure recognition and inspection methods | |
| CN102073868A (en) | Digital image closed contour chain-based image area identification method | |
| CN106648171B (en) | A kind of interactive system and method based on lettering pen | |
| CN115880362A (en) | Code region positioning method and device, computer equipment and computer readable storage medium | |
| Villegas-Jaramillo et al. | An Inexactly Supervised Methodology Based on Multiple Instance Learning, Convolutional Neural Networks and Dissimilarities for Interpretable Defect Detection and Localization on Textured Surfaces | |
| CN112434543A (en) | Image reconstruction method and system for distorted color direct component marking | |
| CN114004786B (en) | Deep learning-based automobile central control panel appearance detection method | |
| CN111860103A (en) | Action identification method based on Tensorflow target detection | |
| CN117830199B (en) | Comprehensive detection method for defects of track elastic strip fastener by fusing image and point cloud | |
| Peng et al. | A corner-clustering method for detection of slab management numbers sprayed on steel slabs | |
| Sas et al. | Logical structure recognition of diagram images |
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 | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20240207 Address after: No. 2 Xiangsheng Road, Zone B, Dacheng Industrial Park, Anzhen Street, Xishan District, Wuxi City, Jiangsu Province, 214105 Patentee after: Wuxi Huansheng precision alloy material Co.,Ltd. Country or region after: China Address before: 710049 No. 28 West Xianning Road, Shaanxi, Xi'an Patentee before: XI'AN JIAOTONG University Country or region before: China |
|
| TR01 | Transfer of patent right |