CN113256571A - Vector graphic feature point extraction method based on direction feature and local uniqueness - Google Patents

Vector graphic feature point extraction method based on direction feature and local uniqueness Download PDF

Info

Publication number
CN113256571A
CN113256571A CN202110513993.2A CN202110513993A CN113256571A CN 113256571 A CN113256571 A CN 113256571A CN 202110513993 A CN202110513993 A CN 202110513993A CN 113256571 A CN113256571 A CN 113256571A
Authority
CN
China
Prior art keywords
angle
arc
obtaining
length
cell
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202110513993.2A
Other languages
Chinese (zh)
Other versions
CN113256571B (en
Inventor
钟靖
张方德
贺兴志
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.)
Zehjiang Ovi Technology Co ltd
Original Assignee
Zehjiang Ovi 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 Zehjiang Ovi Technology Co ltd filed Critical Zehjiang Ovi Technology Co ltd
Priority to CN202110513993.2A priority Critical patent/CN113256571B/en
Publication of CN113256571A publication Critical patent/CN113256571A/en
Application granted granted Critical
Publication of CN113256571B publication Critical patent/CN113256571B/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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • 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
    • 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
    • G06T7/66—Analysis of geometric attributes of image moments or centre of gravity
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30141—Printed circuit board [PCB]

Abstract

The invention discloses a vector graphic feature point extraction method based on direction features and local uniqueness, which comprises two parts, namely direction feature value calculation and local uniqueness screening, wherein edge features are calculated in the direction feature value calculation part through a sliding window; and calculating the uniqueness among the regions in the local uniqueness screening part, scoring each window region by combining the directional features and the uniqueness thereof, and sequencing in a descending order to obtain the first n optimal feature points. A plurality of alignment feature points can be quickly extracted from the vector diagram, the feature points have good uniqueness, and the algorithm is mature and applied to automatic alignment equipment at present.

Description

Vector graphic feature point extraction method based on direction feature and local uniqueness
Technical Field
The invention belongs to the field of image processing, and particularly relates to a vector graphic feature point extraction method based on directional features and local uniqueness.
Background
When a product manufactured according to two-dimensional planar CAD/CAM is subjected to Automated Optical Inspection (AOI), the detection and positioning of the product must rely on the key technology of feature identification. Through feature recognition, the type and the position (rotation and offset) of a product can be obtained, even the deformation amount (stretching and distortion) can be calculated, and the application is concentrated in the fields of CAD drawing digital storage, PCB AOI, film AOI and the like. With the continuous development of scientific technology, the image processing technology is advanced and innovated continuously. However, the mainstream research is still focused on the fields of natural images and medical images, the research on the industrial field is less, and the development of the image processing technology in the industrial field is slow. The target graphics extracted by the feature points are CAM vector graphics, and in the CAM vector graphics, all the graphics are standard geometric patterns, such as polygons consisting of straight lines and circular arcs, and can be easily converted into dot matrix images; however, unlike the bitmap, the CAM pattern decile standard causes a large number of shapes with the same structure to exist in the CAM pattern decile standard, so that the found feature points have no uniqueness in a certain area, which causes a misalignment condition to occur when the features are aligned in a later stage. The problem cannot be well solved by using the traditional feature extraction algorithm, such as SIFT, SUFT, Opencv Fast algorithm and the like.
Disclosure of Invention
Aiming at the defects in the prior art, the invention aims to provide a vector graphic feature point extraction method based on direction features and local uniqueness.
In order to achieve the purpose, the invention provides the following technical scheme:
a vector graphic feature point extraction method based on direction features and local uniqueness comprises the following steps:
the method comprises the following steps of firstly, preprocessing a vector diagram to obtain a graph outline, cutting a line segment and an arc of the graph outline, calculating the total area, and equally dividing a frame-wrapping area of each outline into a plurality of cells for the obtained graph outline, the line segment and the arc, wherein the cells form a table grid for recording the positions of the cells;
step two, obtaining angle classification of the line segments, obtaining corresponding angles and characteristic lengths of the line segments, and storing the corresponding angles and characteristic lengths into table grids;
step three, obtaining the angle classification of the circular arc, obtaining the central angle and the corresponding arc length of the circular arc, and storing the central angle and the corresponding arc length into a table grids;
step four, converting the table grids into window windows, traversing all the areas in the table grids by a set step length, and merging the features in the window range into the eight-direction feature vector VdirIn (1),
Figure BDA0003060579880000021
w is the length of a single window, and the score calculation is carried out on all the windows to obtain the total score
Figure BDA0003060579880000022
liIs the total characteristic length of i direction in eight directions, ljIs the total length of the features in the j direction, αiI direction corresponds to an angle, αjThe direction of the j is corresponding to the angle,
finally, the total score is adjusted according to the orthogonal logarithm to obtain
Figure BDA0003060579880000023
ortho is an orthogonal logarithm, and Z is an adjustment value of no orthogonal time synchronization in a window;
step five, taking the original window as the center, expanding the adjacent area with the length lambda outwards to form a unique area, and obtaining the difference diff (m, n) of the characteristic vector in each window,
Figure BDA0003060579880000024
vmfeature vector, v, representing the mth windownRepresenting the direction feature vector of the nth window;
and step six, obtaining the final total score (score) diff (m, N) according to the total score obtained in the step four and the difference diff (m, N) obtained in the step five, scoring all windows, and then sorting the windows in a descending order to obtain the first N optimal feature points.
In the second step, the corresponding vertical axis offset when the unit length of the x axis is 1 is obtained through the slopey，
Figure BDA0003060579880000025
Offset according to vertical axisyDetermining the number of the line segments crossing the cell; h iscellIs the side length of the cell, (x)start,ystart) Represents the coordinates of the start point of the line segment, (x)end,yend) Representing the coordinates of the line segment end point; simultaneously, the line segment angle is obtained according to the slope
Figure BDA0003060579880000031
And classifies the acquired line segment angle into one of eight directions,
Figure BDA0003060579880000032
if the line span is greater than one cell and the middle line is located in the same cell, the sum of the lengths of the line segments in the adjacent cells is a fixed value l,
Figure BDA0003060579880000033
obtaining the cross axis index of the cell where the starting point of the line segment is
Figure BDA0003060579880000034
Obtaining cell longitudinal axis index of line segment starting point
Figure BDA0003060579880000035
Obtaining the percentage P of the length of the cross shaft of the first segment line segment to the side length of the cellstart，
Figure BDA0003060579880000036
xroiThe abscissa representing the left boundary of the roi,
then, the characteristic length l of the first segment line segment is obtainedstart，lstart＝hcell*PstartDpi, dpi represents pixel density, and the characteristic length l of the obtained first segment line segmentstartAnd storing the corresponding angle into table grids;
obtaining cell cross axis index of line segment terminal point
Figure BDA0003060579880000037
Obtaining cell longitudinal axis index of line segment terminal point
Figure BDA0003060579880000038
Acquiring the percentage P of the length of the transverse shaft of the tail segment line segment to the side length of the cellend，
Figure BDA0003060579880000039
Then obtaining the characteristic length l of the tail segment line segmentend，lend＝hcell*Pend*dpi，
And obtaining the characteristic length l of the tail segment line segmentendAnd the corresponding angle is stored in the table grids.
If the middle line segment is not located in the same cell, cutting the middle line segment to form a first segment and a tail segment,
obtaining the percentage of the cut tail segment in the cell
Figure BDA00030605798800000310
Respectively obtaining the characteristic length l of the first section and the tail sectiondownAnd lup，
ldown＝hcell*Pdown*dpi
lup＝hcell*(1-Pdown)*dpi。
In the third step, obtaining the radius r of the arc fitting circle and the angle of the circle center corresponding to the vertex through circle fitting;
acquiring the horizontal and vertical intersection points of the whole circle in the given arc minimum frame, storing the acquired intersection points, judging whether the arc section is a closed arc, if not, adding a starting point and an end point in the set, and sorting according to the angle descending order; if the arc is the arc, removing points which do not belong to the arc, and directly screening redundant points according to the angle range of the arc;
and after all the intersection points are obtained, sequencing according to the number of the formed central angles, sequentially traversing adjacent end points, and calculating the arc characteristics.
When the characteristics of two points on the arc are calculated, after the forward and reverse directions of the statistical direction are determined, the angles of the two points of the known line segment are normalized, and the angles are calculated according to the defined forward and reverse directions for classification.
If the direction is in the clockwise direction, the direction is,
obtaining angle classifications of starting point angles
Figure BDA0003060579880000041
γstartIs the angle of the circle center corresponding to the starting point of the arc line,
angle classification to obtain endpoint angle
Figure BDA0003060579880000042
γendIs the angle of the circle center corresponding to the end point of the arc line,
if the two angles are classified to be consistent, the arc length of the section is obtained
Figure BDA0003060579880000048
Relocating angle classifications to 8 evenly classified classifications
Figure BDA0003060579880000043
Storing the angle classification and the total arc length into grids,
Figure BDA0003060579880000044
is the average cell abscissa of the arc starting and ending points,
Figure BDA0003060579880000045
the mean cell ordinate of the starting point and the end point of the arc line;
if the two angle classifications are not consistent, then
Obtaining the first arc length
Figure BDA0003060579880000046
Obtaining the arc length of the tail section
Figure BDA0003060579880000047
Relocating the two-angle classification to an eight-directional angle classification,
first segment angle classification
Figure BDA0003060579880000051
mod8，
Tail section angle classification
Figure BDA0003060579880000052
mod8，
Classifying the arc lengths and angles of the first section and the tail section and storing the arc lengths and angles into grids;
if the two angle classifications differ by no less than 2,
sorting by angle
Figure BDA0003060579880000053
As a starting point, the method comprises the following steps of,
Figure BDA0003060579880000054
for the traversal of the end point, the angle and the arc length along the way
Figure BDA0003060579880000055
And storing in grids.
If in the counterclockwise direction, the first and second,
obtaining starting point angle classifications
Figure BDA0003060579880000056
Obtaining endpoint angle classifications
Figure BDA0003060579880000057
If the two angle classifications are consistent, the total arc length is obtained
Figure BDA0003060579880000058
Relocating angle classifications to 8 evenly classified classifications
Figure BDA0003060579880000059
Storing the angle classification and the total arc length into grids,
Figure BDA00030605798800000510
if the two angle classifications are not consistent, then
Obtaining the first arc length
Figure BDA00030605798800000511
Obtaining the arc length of the tail section
Figure BDA00030605798800000512
Relocating the two-angle classification to an eight-directional angle classification,
classifying the arc lengths and angles of the first section and the tail section and storing the arc lengths and angles into grids;
if the two angle classifications differ by not less than 2, to
Figure BDA00030605798800000513
As a starting point, the method comprises the following steps of,
Figure BDA00030605798800000514
traversal is performed for the endpoint.
In the fifth step, all windows are traversed, the difference degree of each window and all windows in the domain is calculated once, and the feature vectors v of any two windowsm、vnAnd through
Figure BDA00030605798800000515
And obtaining the difference, calculating the difference between the central point and other windows in the unique region, obtaining the difference, comparing the difference with the current difference of the two windows, and replacing the difference with the minimum value.
A window-sized clean circular profile is created, and the feature is subjected to a benchmark score calculation,
after the benchmark score is obtained, the window sets which are sorted according to the descending order of the total score are screened,
if the window score is larger than the reference score, the window is the area with the highest score in the area, the neighborhood window in the unique area is directly cleaned,
and if the window score is smaller than the reference score, searching all surrounding windows by eight neighborhoods and replacing the windows with the maximum difference values.
The invention has the beneficial effects that: the invention can quickly extract a plurality of Alignment feature points from the vector diagram, and the feature points have good uniqueness, and can be mature applied to industrial equipment such as an automatic Alignment Exposure machine (automatic Alignment Exposure), a Direct Imaging machine (Direct Imaging), automatic Optical Inspection (automatic Optical Inspection) and the like.
Drawings
FIG. 1 is a graphical illustration of a change of direction.
FIG. 2 is a partial vector graph.
FIG. 3 is a schematic diagram of a cell and window.
FIG. 4 is a schematic diagram showing the length of the horizontal axis exceeding the length of the vertical axis.
Fig. 5 is a schematic diagram of fig. 4 after the horizontal and vertical axis coordinates are exchanged.
FIG. 6 is a schematic diagram of the distribution of middle line segments and cells.
Fig. 7 is a schematic diagram of the middle line segment of the acell distribution.
FIG. 8 is a schematic diagram of the horizontal and vertical intersection points of the fitted circle within the minimum bounding box.
Fig. 9 is a schematic drawing of a 16-part of a fitted circle.
Fig. 10 is a schematic diagram of the division of the circular arc in the clockwise direction.
Fig. 11 is a schematic view of two angles in the same region in the counterclockwise direction.
FIG. 12 is a schematic view of the stepping of windows in grids.
FIG. 13 is a schematic diagram of a unique comparison area.
Fig. 14 is a partial outline fill of a PCB.
Fig. 15 is a schematic diagram of characteristic point extraction of a PCB portion.
Fig. 16 is an enlarged schematic view of a feature area at any position of the PCB.
Detailed Description
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 only a part of the embodiments of the present invention, and not all of the embodiments. 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.
It should be noted that all the directional indicators (such as up, down, left, right, front, and rear … …) in the embodiment of the present invention are only used to explain the relative position relationship between the components, the movement situation, etc. in a specific posture (as shown in the drawing), and if the specific posture is changed, the directional indicator is changed accordingly.
In the present invention, unless otherwise expressly stated or limited, the terms "connected," "secured," and the like are to be construed broadly, and for example, "secured" may be a fixed connection, a removable connection, or an integral part; can be mechanically connected or connected; they may be directly connected or indirectly connected through intervening media, or they may be connected internally or in any other suitable relationship, unless expressly stated otherwise. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
In addition, the descriptions related to "first", "second", etc. in the present invention are only for descriptive purposes and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one such feature. In addition, technical solutions between various embodiments may be combined with each other, but must be realized by a person skilled in the art, and when the technical solutions are contradictory or cannot be realized, such a combination should not be considered to exist, and is not within the protection scope of the present invention.
The method mainly comprises two parts of direction characteristic value calculation and local uniqueness screening, wherein the direction characteristic value calculation part calculates edge characteristics through a sliding window; and calculating the uniqueness among the regions in the local uniqueness screening part, scoring each window region by combining the directional features and the uniqueness thereof, and sequencing in a descending order to obtain the first n optimal feature points.
For the computation of the directional eigenvalue, the present invention divides the vector graphics with reference to the concept of Histogram of Oriented Gradient (HOG).
FIG. 1(b) is the result of the rotational transformation of FIG. 1(a) with a single edge as the calibration template, and FIG. 1(a) can be used to determine the x-axis offset; FIG. 1(b) can be used to determine the y-axis offset. Obviously, such edges can help to locate the offset and rotation angle of the overall image, while the combination of edges at multiple angles can help us to locate more accurately. It is obvious that there are some patterns suitable for optimally locating feature points in the vector pattern, as shown in fig. 2, which are all good feature points in view of the concept of the general feature search algorithm (SIFT, SUFT, Opencv Fast, etc.), the number of vectors in fig. 1(a) is equal, and the number of vectors in the four-corner direction and the number of vectors in the orthogonal direction in fig. 2(b) and fig. 1(c) are substantially close. According to the method, from the angle characteristics of the edge, the detected edge direction is expanded into eight directions of 0 degree, 22.5 degrees, 45 degrees, 67.5 degrees, 90 degrees, 112.5 degrees, 135 degrees and 157.5 degrees to detect the characteristics, and characteristic values corresponding to all the directions are counted for subsequent uniqueness analysis.
In addition, basic data structure elements cell and a sliding window designed by the invention are defined as follows:
the cell is a square with an extremely small side length, preferably, the default length of the cell is set to 50 mils, the cell is used for measuring the basic unit of the area characteristic value, and a plurality of cells form a characteristic window.
window size, i.e., feature size, the side length of the window is preferably defined as 4 times of the cell, i.e., 200 mils, and the window step size is 1 for sufficient feature information extraction.
A schematic diagram of the cell and the window is shown in fig. 3, wherein the dotted line part is a part expanded when measuring the feature, and is used for better considering the feature information of the edge of the area and preventing the edge from falling on the window boundary.
In consideration of huge graphic information, R-Tree developed from B-Tree is used for storing graphic outline data, and Sort-Tile-Recursive (STR) algorithm is introduced to load data.
Firstly, a vector diagram of a circuit board is obtained, all independent outlines in the vector diagram are extracted, and then an obvious repeated part, a part with a small area and a character part are filtered out. And splitting each contour into a plurality of combinations of line segments and circular arcs, cutting the line segments and the arc lines of all the filtered contours, and calculating the total area for subsequent characteristic value calculation.
And traversing the graph contour obtained by preprocessing and the line segment and the circular arc separated from the graph contour, dividing a frame-wrapping area (a minimum rectangle capable of covering the whole contour) of each contour into a plurality of cells, and forming a table grid by the plurality of cells for recording the positions of the cells.
When the contour is cut, and the angular range of the line segment is unified with the direction of the horizontal axis as a reference, if the length of the vertical axis occupied by the line segment is greater than that of the horizontal axis as shown in fig. 4, the horizontal and vertical axis coordinates need to be exchanged.
For straight line segments.
The side length of the cell is hcellThe vertical axis offset corresponding to the unit length of the x-axis of 1 is obtained from the slopey
Figure BDA0003060579880000091
In the formula (x)start,ystart) Represents the coordinates of the start point of the line segment, (x)end,yend) Coordinates representing line segment end point
Calculating the angle of the line segment according to the slope
Figure BDA0003060579880000092
Classifying the line segment angle to one of eight directions, locating to subscript Iangle
Figure BDA0003060579880000093
As shown in fig. 4, after cell cutting, the first segment and the last segment of the line segment may not occupy the cell, and the actual occupation ratio needs to be calculated, and if the actual occupation ratio is directly ignored, the accuracy is lost if the actual occupation ratio is reduced to one cell.
Obviously, when the span of the straight line segment is greater than one cell, segmentation processing is required, and when the span is not less than two cells, the sum of the lengths of the straight line segment in the adjacent cells is a fixed value l
Figure BDA0003060579880000094
Calculating the cross-axis index of the cell where the starting point is located
Figure BDA0003060579880000095
Figure BDA0003060579880000096
Calculating the vertical axis index of the cell where the starting point is located
Figure BDA0003060579880000097
Figure BDA0003060579880000098
Calculating the percentage P of the length of the transverse axis of the first segment, such as the segment AM in FIG. 5, to the side length of the cellstart
Figure BDA0003060579880000101
Wherein x isroiAbscissa representing left boundary of roi
Calculating the characteristic length l of the first segmentstart
lstart＝hcell*Pstart*dpi (8)
Where dpi represents the pixel density
And storing the characteristic lengths and the corresponding angle categories into grids, and if the transverse and longitudinal axis conversion is performed before, cell index recovery needs to be exchanged. Corresponding lengths and angles are stored in grids as shown in equation 9
Figure BDA0003060579880000108
Otherwise, the corresponding length and angle are stored as in equation 10 without any modification
Figure BDA0003060579880000109
Calculating the end segment class, such as segment MB in FIG. 5, is the same as calculating the first segment, and calculating the subscript of the cross axis of the cell where the end point is located
Figure BDA0003060579880000102
Figure BDA0003060579880000103
Calculating the subscript of the longitudinal axis of the cell where the end point is located
Figure BDA0003060579880000104
Figure BDA0003060579880000105
Similar to equation 7, calculate the percentage P of the length of the horizontal axis of the tail segment line segment to the side length of the cellend
Figure BDA0003060579880000106
Calculating the actual length l of the tail segment line segmentend
lend＝hcell*Pend*dpi (14)
The actual length of the cell is stored in the grids according to the angle type and the actual length of the line segment
If over-conversion is performed
Figure BDA0003060579880000107
Otherwise
Figure BDA0003060579880000111
If the x coordinates of the starting point cell and the end point cell are larger than 1, intermediate line segments with the length being l integral multiple exist between the head and the tail of the line segments, and the calculation is needed section by section.
The calculation of the part can be briefly described as that the end coordinate of the first segment is taken as the starting point (e.g. point B in fig. 6), the horizontal axis length is taken as 1 to step and obtain the corresponding horizontal axis coordinate, the cell characteristics included in the line segment formed by the two points are calculated and stored, the stepping is continued and the calculation is performed until the coordinate is updated to the starting point of the last segment (e.g. point C in fig. 6), and the flow is ended.
The actual situation of calculating the middle part is taken as an example of the line AD in fig. 6, and the AD middle line BC is located in one cell. The actual coordinate of point A is (x)start,ystart) From the above equations 1 and 7, the corresponding actual vertical axis displacement offset when the length of the horizontal axis of the straight line is 1 has been derivedyAnd the percentage P of the first segment AB in the cellstartFrom this, the point B coordinate loc can be deducedb
locb＝(xstart+xstart*Pstart,ystart+offsety*Pstart) (17)
Horizontal and vertical coordinates of cell corresponding to point B
Figure BDA0003060579880000112
Figure BDA0003060579880000113
The same can derive the point C coordinate locc
locc＝(xstart+1*(Pstart+1),ystart+offsety*(Pstart+1)) (20)
Horizontal and vertical coordinates of cell corresponding to point C
Figure BDA0003060579880000114
Figure BDA0003060579880000115
It can be deduced from the equations 18, 19, 21, 22 that the abscissa and ordinate of the cell for the first n points starting from the start of the first segment and stepping with the abscissa length 1 to the start of the last segment can be determined from the equations 23, 24
Figure BDA0003060579880000121
Figure BDA0003060579880000122
Continuously analyzing the segment BC, and since the cell ordinate of B and C are the same, proving that the segment BC is located in the same cell and the length of BC is exactly l, namely the length of BC is l
In this case, the features can be directly stored in grids, and if the features are subjected to coordinate transformation, the features can be stored as in equation 25
Figure BDA0003060579880000123
Otherwise is stored as formula 26
Figure BDA0003060579880000124
If the vertical coordinates of the cells are not located in the same cell, as shown by the line segment BC in fig. 7, the horizontal coordinates of the cell where the point B, C is located are the same, and the vertical coordinates are different, it is proved that the line segment BC is not located in the same cell, and the line segment BC is divided into a first segment BM and a last segment MC for measurement.
Obviously, the actual length of the line segment BC is l, and the sum of the cell ratios of the head and the tail is 1, as shown in formula 27, the percentage P of the tail is calculated asdownCan push the percentage of the first section backwards
Figure BDA0003060579880000125
Calculating the actual length l of the tail segment line segmentdownAnd lup
ldown＝hcell*Pdown*dpi (28)
In the formula, hcellThe side length of the cell is taken as the length of the cell,
lup＝hcell*(1-Pdown)*dpi (29)
and (3) bringing the actual lengths of the two parts into grids according to the coordinates of the cell, and if the coordinate system is converted, adjusting the actual lengths by the same formulas 9 and 10, wherein the process is consistent with the method for calculating the head and tail sections.
For a circular arc segment.
The method is suitable for the known circular arc with two end vertexes. Firstly, obtaining the radius of a circular arc fitting circle and the angle of the circle center corresponding to the vertex through circle fitting; when calculating the arc direction characteristics, the arc may be regarded as a combination of several straight line segments, and therefore, for a given arc, the horizontal and vertical intersection points of the whole circle within the minimum bounding box are calculated as shown in fig. 8. Storing the intersection points, judging whether the arc sections are closed arcs or not, if not, adding a starting point and an end point in the set, and sorting according to the angle descending order; if the point is a circular arc, the points which do not belong to the circular arc are removed, and redundant points are directly screened out according to the angle range of the circular arc.
Different from the calculation of straight-line segment characteristics, when circular arc characteristics are calculated, central angles corresponding to end points are characteristic angles, after all intersection points are obtained, sequencing is carried out according to the number of the formed central angles, adjacent end points are traversed in sequence, and arc characteristics are calculated.
When the features are stored in the grids, the calculation formulas of the corresponding cell coordinates are respectively shown as formulas 30 and 31
Figure BDA0003060579880000131
Figure BDA0003060579880000132
Different from a straight line, if the accuracy is reduced by classifying the arc angle only in eight directions, the method adopts a compromise method, namely, the angle shown in a circle figure 9 is divided into 16 parts for measuring characteristics, and the angle is classified back to the eight directions after the measurement is finished so as to keep the consistency of direction measurement.
Arc length l corresponding to each equal division anglearcIs composed of
Figure BDA0003060579880000133
Two points on the arc can be subjected to two conditions of calculating characteristics, as shown in fig. 9 and 10, respectively, when the angle formed by the two points is greater than an average angle, the arc length of the middle part is necessarily larcInteger multiples of; when the central angle formed by the two points is smaller than an even angle, if the central angle is classified under the same angle.
And calculating the arc length according to the angle difference, and if the arc length is not classified at the same angle, separately calculating the arc length of the corresponding part and storing the arc length into the corresponding grid. After the forward and reverse directions of the statistical direction are determined, the angles of two points of the known line segment are normalized, and angle classification is calculated according to the defined forward and reverse directions.
Angle classification of calculating starting point angle clockwise
Figure BDA0003060579880000134
Figure BDA0003060579880000135
Angle classification to calculate endpoint angle
Figure BDA0003060579880000141
If the two angle classifications are consistent, as shown in FIG. 11.
Calculating the arc length of the segment as shown in formula 35
Figure BDA00030605798800001411
Finally, the angle is relocated to 8 uniform classification
Figure BDA0003060579880000142
The angles and feature lengths are stored in grids.
Figure BDA0003060579880000143
If the two angles are classified differently, the arc length of the head is calculated as shown in equation 38
Figure BDA0003060579880000144
Calculating the tail arc length as in equation 39
Figure BDA0003060579880000145
The two angles are repositioned to eight-directional angle classifications, and the head angles are reclassified as shown in equation 40
Figure BDA0003060579880000146
Reclassification of the tail angle is shown in equation 41
Figure BDA0003060579880000147
And finally, storing the characteristics of the head and the tail into grids.
If the angle classification difference between the starting point and the end point is not less than 2, as shown in fig. 10, it is proved that at least one arc with the arc length being an integral multiple of the average arc length exists in the middle, and the classification is carried out according to the angle
Figure BDA0003060579880000148
As a starting point, the method comprises the following steps of,
Figure BDA0003060579880000149
traversing for the end point, and determining the angle and arc length along the way
Figure BDA00030605798800001410
And storing in grids.
When calculating along the counterclockwise direction, calculating the index corresponding to the angle according to the defined searching direction, such as calculating the starting point angle classification according to formula 42
Figure BDA0003060579880000151
End point angle classification as calculated by equation 43
Figure BDA0003060579880000152
If the two angle classifications are consistent, as shown in FIG. 8, the arc length is calculated by equation 32, the angle is repositioned to the eight-even index by equation 36, and the arc characteristics are calculated by combining the angle classifications and storing in grids.
If the two angle classifications are different, the starting point classification is calculated by equation 44
Figure BDA0003060579880000153
Calculation of endpoint Classification by equation 45
Figure BDA0003060579880000154
Finally, the two angles are relocated to eight even index, and the characteristics of the head and tail are stored in grids. If the angle classification difference between the starting point and the end point is not less than 2, the steps are repeated to
Figure BDA0003060579880000155
As a starting point, the method comprises the following steps of,
Figure BDA0003060579880000156
and traversing for the end point.
And converting the information recorded in the grids into a window, as shown in fig. 12, the dark part is the current position of the window, and traversing all the regions in the grids by step 1, so that the row and column layout of the created window is consistent with the grids.
The invention provides that each cell has a window with the coordinate of the cell as the starting point of the upper left corner, the windows are traversed in sequence, and the cell characteristics in the window range are merged into the eight-direction characteristic vector V of the windowdirAs shown in formula 46.
Figure BDA0003060579880000157
Where w is the length of a single window.
Calculating scores for all the characteristic windows, and the starting points of the evaluation algorithm are as follows:
1. in the orthogonal direction, the closer the quantity is, the more beneficial the characteristics are to be identified, and the product of the difference rate and the sine value is taken as a weight factor;
2. the direction count must be considered, otherwise a small orthogonal pair also has a higher score;
3. four pairs of orthogonal direction pairs exist, the more the pairs are suitable for alignment, at least one pair of orthogonal direction pairs exists, the marginal effect of linear scoring is decreased progressively, and therefore the linear scoring is distributed according to the following fixed values: 1 st pair 8/15, 2 nd pair 4/15, 3 rd pair 2/15, 4 th pair 1/15;
4. the denser the contour is, the richer the boundary is, the more suitable the alignment is, and this factor is used as the second weight factor, and like the third point, the weight is set to a fixed value of 20% without linear scoring.
The orthogonality degree between the two directions of each line segment and the circular arc segment is fully considered, the total score and the complete orthogonality quantity are related, only the optimal pair locus is considered, and the uniqueness is not considered at all.
The orthogonal efficiency of each direction and all other directions in the vector is calculated sequentially as formula 47
Figure BDA0003060579880000161
If η infinity is close to 1, it is verified that the two vectors are orthogonal and the orthogonal number ortho is increased by 1.
Calculating a window total score
Figure BDA0003060579880000162
In the formula IiIs the total length of the features in the ith direction of the eight directions, ljIs the total length of the features in the j direction, αiI direction corresponds to an angle, αjThe j direction corresponds to an angle.
Finally, the total score is adjusted according to the orthogonal condition
Figure BDA0003060579880000163
When there is no orthogonal pair in the window, the adjustment value Z is 0.5 times of the original value, 1 time of 1 orthogonal pair, 1.4 times of 2 orthogonal pairs, 1.7 times of 3 orthogonal pairs, and 2 times of 4 orthogonal pairs.
Since a large amount of similarity may exist in a local area, the feature points do not have uniqueness in a certain area, and a plurality of results exist in the later stage of feature point matching. The problem that the feature points may not have uniqueness is not solved well only by calculating the direction features, so the feature points with highest uniqueness and highest difference in each area need to be selected to solve the problem of feature point repetition.
The invention defines the uniqueness region and the degree of difference as follows:
the uniqueness region: taking the original window as the center, expanding the adjacent area with the length lambda outwards, as shown in fig. 13, respectively showing the uniqueness area under the window viewing angle and under the cell viewing angle, the light blue area being the original window range, and the gray area being the comparison range of the uniqueness area.
The difference degree: the difference can measure the repetition degree of a part of characteristics in the whole area, and the higher the difference is, the lower the repetition degree is.
Obtaining the uniqueness consideration range u according to the uniqueness region side length lambda and the cell side length as the formula 50
Figure BDA0003060579880000171
The sum of the values of l is calculated as
Figure BDA0003060579880000172
Wherein the content of the first and second substances,
Figure BDA0003060579880000173
is the length of the feature in the ith direction in a certain feature vector v.
Traversing all windows, calculating the difference degree of each window and all windows in the domain once, and calculating the characteristic direction of any two windowsQuantity vm,vnThe degree of difference diff (v) is calculated by equation 52m,vn)
Figure BDA0003060579880000174
As shown in fig. 13, the difference between the center point and the other windows is calculated in the unique region, and after the difference is obtained, the difference is compared with the current difference between the two windows and is replaced with the minimum value, otherwise, the difference is not changed.
The scanning order affects the effect of the algorithm, and the scanning order of the sliding window is set as from bottom to top (inner loop) and from left to right (outer loop), and the condition setting is noted in the loop, and the overlapped windows are skipped.
Taking the example of the Sudoku center, if compared to the surrounding 8 neighborhoods, this results in twice as many computations, splitting the text into two cycles to save time and to do so
Figure BDA0003060579880000175
And
Figure BDA0003060579880000176
respectively represent the sum of 1 st and 2 nd cycle diff, and the formula is as follows
Figure BDA0003060579880000177
Figure BDA0003060579880000178
Where row, col is the number of rows and columns of all cells in the entire image domain.
The calculation of the total score here takes into account a summary of the uniqueness score and the direction score. And traversing all windows to calculate the total score, and if the difference degree of the windows is less than 0.2, indicating that the windows do not have uniqueness, wherein the window score is 0.
The final total score is
scorei,j＝scorei,j*diff(i,j) (55)
And after scoring all windows, sorting the windows in a descending order to obtain the first n optimal feature points.
The calculation of the benchmark score is performed on the feature by creating a window-sized circular contour.
And after the benchmark score is obtained, screening the window sets which are sorted according to the descending order of the total score.
If the window score is larger than the reference time, the window is necessarily the area with the highest score, and the neighborhood window in the unique area of the window is directly cleaned.
And if the window score is smaller than the reference score, searching all surrounding windows by eight neighborhoods and replacing the windows with the maximum difference values.
After the screening is finished, a plurality of optimal feature areas (points) are obtained, and the whole feature point extraction algorithm based on the geometric figure is finished.
A vector diagram of a complete PCB (202.5345 mm in length and 104.0879mm in width) is read by CAM + GTL, the vector diagram has high resolution and a large number of independent outlines, a filling diagram is displayed for facilitating the observation of the extraction effect, as shown in FIG. 14, the extracted characteristic region diagram is represented in a box form, the characteristic point extraction is visually carried out on the diagram,
the parameter settings are as follows:
the cell side length is set to 0.05 pixel length,
the side length of the window is set to four times the side length of the cell,
the length of the uniqueness neighborhood is set to 0.5, i.e. 10 times the side length of the cell,
dpi is set to 1000.
The characteristic points of the PCB are extracted, and the software displays the characteristic area obtained by the algorithm on the outline map in a box form, as shown in FIG. 15.
From these, a certain feature region is selected, as shown in fig. 16, the contour in the region has the lowest degree of repetition in the whole region, and includes excellent features such as circular arcs and rectangles, which are similar to the manually selected region.
The examples should not be construed as limiting the present invention, but any modifications made based on the spirit of the present invention should be within the scope of protection of the present invention.

Claims (10)

1. A vector graphic feature point extraction method based on direction features and local uniqueness is characterized by comprising the following steps: which comprises the following steps:
the method comprises the following steps of firstly, preprocessing a vector diagram to obtain a graph outline, cutting a line segment and an arc of the graph outline, calculating the total area, and equally dividing a frame-wrapping area of each outline into a plurality of cells for the obtained graph outline, the line segment and the arc, wherein the cells form a table grid for recording the positions of the cells;
step two, obtaining angle classification of the line segments, obtaining corresponding angles and characteristic lengths of the line segments, and storing the corresponding angles and characteristic lengths into table grids;
step three, obtaining the angle classification of the circular arc, obtaining the central angle and the corresponding arc length of the circular arc, and storing the central angle and the corresponding arc length into a table grids;
step four, converting the table grids into window windows, traversing all the areas in the table grids by a set step length, and merging the features in the window range into the eight-direction feature vector VdirIn (1),
Figure FDA0003060579870000011
w is the length of a single window, and the score calculation is carried out on all the windows to obtain the total score
Figure FDA0003060579870000012
liIs the total characteristic length of i direction in eight directions, ljIs the total length of the features in the j direction, αiI direction corresponds to an angle, αjThe direction of the j is corresponding to the angle,
finally, the total score is adjusted according to the orthogonal logarithm to obtain
Figure FDA0003060579870000013
ortho is an orthogonal logarithm, and Z is an adjustment value of no orthogonal time synchronization in a window;
step five, taking the original window as the center, expanding the adjacent area with the length lambda outwards to form a unique area, and obtaining the difference diff (m, n) of the characteristic vector in each window,
Figure FDA0003060579870000014
vmfeature vector, v, representing the mth windownRepresenting the direction feature vector of the nth window;
and step six, obtaining the final total score (score) diff (m, N) according to the total score obtained in the step four and the difference diff (m, N) obtained in the step five, scoring all windows, and then sorting the windows in a descending order to obtain the first N optimal feature points.
2. The vector graphics feature point extraction method based on directional features and local uniqueness according to claim 1, characterized in that: in the second step, the corresponding vertical axis offset when the unit length of the x axis is 1 is obtained through the slopey，
Figure FDA0003060579870000021
Offset according to vertical axisyDetermining the number of the line segments crossing the cell; h iscellIs the side length of the cell, (x)start,ystart) Represents the coordinates of the start point of the line segment, (x)end,yend) Representing the coordinates of the line segment end point; simultaneously, the line segment angle is obtained according to the slope
Figure FDA0003060579870000022
And classifies the acquired line segment angle into one of eight directions,
Figure FDA0003060579870000023
3. the vector graphics feature point extraction method based on directional features and local uniqueness according to claim 2, wherein: if the line span is larger than one cell, and the middle line is located in the same cellAnd the sum of the lengths of the line segments in the adjacent cells is a fixed value l,
Figure FDA0003060579870000024
obtaining the cross axis index of the cell where the starting point of the line segment is
Figure FDA0003060579870000025
Obtaining cell longitudinal axis index of line segment starting point
Figure FDA0003060579870000026
Obtaining the percentage P of the length of the cross shaft of the first segment line segment to the side length of the cellstart，
Figure FDA0003060579870000027
xroiThe abscissa representing the left boundary of the roi,
then, the characteristic length l of the first segment line segment is obtainedstart，lstart＝hcell*PstartDpi, dpi representing pixel density,
and obtaining the characteristic length l of the first segment line segmentstartAnd storing the corresponding angle into table grids;
obtaining cell cross axis index of line segment terminal point
Figure FDA0003060579870000028
Obtaining cell longitudinal axis index of line segment terminal point
Figure FDA0003060579870000029
Acquiring the percentage P of the length of the transverse shaft of the tail segment line segment to the side length of the cellend，
Figure FDA00030605798700000210
Then obtaining the characteristic length l of the tail segment line segmentend，lend＝hcell*Pend*dpi，
And obtaining the characteristic length l of the tail segment line segmentendAnd the corresponding angle is stored in the table grids.
4. The vector graphics feature point extraction method based on directional features and local uniqueness according to claim 2, wherein: if the middle line segment is not located in the same cell, cutting the middle line segment to form a first segment and a tail segment,
obtaining the percentage of the cut tail segment in the cell
Figure FDA0003060579870000031
Respectively obtaining the characteristic length l of the first section and the tail sectiondownAnd lup，
ldown＝hcell*Pdown*dpi
lup＝hcell*(1-Pdown)*dpi。
5. The vector graphics feature point extraction method based on directional features and local uniqueness according to claim 1, characterized in that: in the third step, obtaining the radius r of the arc fitting circle and the angle of the circle center corresponding to the vertex through circle fitting;
acquiring the horizontal and vertical intersection points of the whole circle in the given arc minimum frame, storing the acquired intersection points, judging whether the arc section is a closed arc, if not, adding a starting point and an end point in the set, and sorting according to the angle descending order; if the arc is the arc, removing points which do not belong to the arc, and directly screening redundant points according to the angle range of the arc;
and after all the intersection points are obtained, sequencing according to the number of the formed central angles, sequentially traversing adjacent end points, and calculating the arc characteristics.
6. The vector graphics feature point extraction method based on directional features and local uniqueness according to claim 5, wherein: when the characteristics of two points on the arc are calculated, after the forward and reverse directions of the statistical direction are determined, the angles of the two points of the known line segment are normalized, and the angles are calculated according to the defined forward and reverse directions for classification.
7. The vector graphics feature point extraction method based on directional features and local uniqueness according to claim 6, wherein: if the direction is in the clockwise direction, the direction is,
obtaining angle classifications of starting point angles
Figure FDA0003060579870000032
γstartIs the angle of the circle center corresponding to the starting point of the arc line,
angle classification to obtain endpoint angle
Figure FDA0003060579870000041
γendIs the angle of the circle center corresponding to the end point of the arc line,
if the two angles are classified to be consistent, the arc length of the section is obtained
Figure FDA0003060579870000042
Relocating angle classifications to 8 evenly classified classifications
Figure FDA0003060579870000043
Storing the angle classification and the total arc length into grids,
Figure FDA0003060579870000044
Figure FDA0003060579870000045
is the average cell abscissa of the arc starting and ending points,
Figure FDA0003060579870000046
the mean cell ordinate of the starting point and the end point of the arc line;
if the two angle classifications are not consistent, then
Obtaining the first arc length
Figure FDA0003060579870000047
Obtaining the arc length of the tail section
Figure FDA0003060579870000048
Relocating the two-angle classification to an eight-directional angle classification,
first segment angle classification
Figure FDA0003060579870000049
Tail section angle classification
Figure FDA00030605798700000410
Classifying the arc lengths and angles of the first section and the tail section and storing the arc lengths and angles into grids;
if the two angle classifications differ by no less than 2,
sorting by angle
Figure FDA00030605798700000411
As a starting point, the method comprises the following steps of,
Figure FDA00030605798700000412
for the traversal of the end point, the angle and the arc length along the way
Figure FDA00030605798700000413
And storing in grids.
8. The vector graphics feature point extraction method based on directional features and local uniqueness according to claim 6, wherein: if in the counterclockwise direction, the first and second,
obtaining starting point angle classifications
Figure FDA00030605798700000414
Terminal acquisitionPoint angle classification
Figure FDA00030605798700000415
If the two angle classifications are consistent, the total arc length is obtained
Figure FDA00030605798700000416
Relocating angle classifications to 8 evenly classified classifications
Figure FDA00030605798700000417
Storing the angle classification and the total arc length into grids,
Figure FDA0003060579870000051
if the two angle classifications are not consistent, then
Obtaining the first arc length
Figure FDA0003060579870000052
Obtaining the arc length of the tail section
Figure FDA0003060579870000053
Relocating the two-angle classification to an eight-directional angle classification,
classifying the arc lengths and angles of the first section and the tail section and storing the arc lengths and angles into grids;
if the two angle classifications differ by not less than 2, to
Figure FDA0003060579870000054
As a starting point, the method comprises the following steps of,
Figure FDA0003060579870000055
traversal is performed for the endpoint.
9. The vector graphics feature point extraction method based on directional features and local uniqueness according to claim 1, characterized in that: in step five, go throughCalculating the difference degree of all windows, each window and all windows in the domain once, and calculating the feature vector v of any two windowsm、vnAnd through
Figure FDA0003060579870000056
And obtaining the difference, calculating the difference between the central point and other windows in the unique region, obtaining the difference, comparing the difference with the current difference of the two windows, and replacing the difference with the minimum value.
10. The vector graphics feature point extraction method based on directional features and local uniqueness according to claim 1, characterized in that: a window-sized clean circular profile is created, and the feature is subjected to a benchmark score calculation,
after the benchmark score is obtained, the window sets which are sorted according to the descending order of the total score are screened,
if the window score is larger than the reference score, the window is the area with the highest score in the area, the neighborhood window in the unique area is directly cleaned,
and if the window score is smaller than the reference score, searching all surrounding windows by eight neighborhoods and replacing the windows with the maximum difference values.
CN202110513993.2A 2021-05-11 2021-05-11 Vector graphic feature point extraction method based on direction feature and local uniqueness Active CN113256571B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110513993.2A CN113256571B (en) 2021-05-11 2021-05-11 Vector graphic feature point extraction method based on direction feature and local uniqueness

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110513993.2A CN113256571B (en) 2021-05-11 2021-05-11 Vector graphic feature point extraction method based on direction feature and local uniqueness

Publications (2)

Publication Number Publication Date
CN113256571A true CN113256571A (en) 2021-08-13
CN113256571B CN113256571B (en) 2022-05-27

Family

ID=77222852

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110513993.2A Active CN113256571B (en) 2021-05-11 2021-05-11 Vector graphic feature point extraction method based on direction feature and local uniqueness

Country Status (1)

Country Link
CN (1) CN113256571B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114248100A (en) * 2021-12-03 2022-03-29 武汉纺织大学 Screw hole positioning algorithm and screw locking device based on image processing

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006209353A (en) * 2005-01-26 2006-08-10 Sharp Corp Image determination apparatus, image formimg apparatus, image determination method, image determination program, image formimg program, and computer readable recording medium
CN101064009A (en) * 2006-04-28 2007-10-31 夏普株式会社 Image processing apparatus, image forming apparatus, image reading apparatus and image processing method
CN102496161A (en) * 2011-12-13 2012-06-13 浙江欧威科技有限公司 Method for extracting contour of image of printed circuit board (PCB)
CN105320783A (en) * 2014-07-01 2016-02-10 北京华大九天软件有限公司 Three-dimensional vector object representation method based on two-dimensional SVG vector graphics in integrated circuit design tool
US20170099438A1 (en) * 2015-10-05 2017-04-06 Canon Kabushiki Kaisha Image processing apparatus and method
CN107341837A (en) * 2017-06-26 2017-11-10 华中师范大学 Grid and vector data conversion and continuous yardstick expression based on image pyramid
CN109447160A (en) * 2018-10-31 2019-03-08 武汉大学 A kind of method of image and vector road junction Auto-matching
CN109948653A (en) * 2019-02-15 2019-06-28 广东智媒云图科技股份有限公司 A kind of image similarity judgment method, electronic equipment and storage medium
CN111805051A (en) * 2020-06-30 2020-10-23 梅卡曼德（北京）机器人科技有限公司 Groove cutting method, device, electronic equipment and system

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006209353A (en) * 2005-01-26 2006-08-10 Sharp Corp Image determination apparatus, image formimg apparatus, image determination method, image determination program, image formimg program, and computer readable recording medium
CN101064009A (en) * 2006-04-28 2007-10-31 夏普株式会社 Image processing apparatus, image forming apparatus, image reading apparatus and image processing method
CN102496161A (en) * 2011-12-13 2012-06-13 浙江欧威科技有限公司 Method for extracting contour of image of printed circuit board (PCB)
CN105320783A (en) * 2014-07-01 2016-02-10 北京华大九天软件有限公司 Three-dimensional vector object representation method based on two-dimensional SVG vector graphics in integrated circuit design tool
US20170099438A1 (en) * 2015-10-05 2017-04-06 Canon Kabushiki Kaisha Image processing apparatus and method
CN107341837A (en) * 2017-06-26 2017-11-10 华中师范大学 Grid and vector data conversion and continuous yardstick expression based on image pyramid
CN109447160A (en) * 2018-10-31 2019-03-08 武汉大学 A kind of method of image and vector road junction Auto-matching
CN109948653A (en) * 2019-02-15 2019-06-28 广东智媒云图科技股份有限公司 A kind of image similarity judgment method, electronic equipment and storage medium
CN111805051A (en) * 2020-06-30 2020-10-23 梅卡曼德（北京）机器人科技有限公司 Groove cutting method, device, electronic equipment and system

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114248100A (en) * 2021-12-03 2022-03-29 武汉纺织大学 Screw hole positioning algorithm and screw locking device based on image processing

Also Published As

Publication number Publication date
CN113256571B (en) 2022-05-27

Similar Documents

Publication Publication Date Title
CN115082467B (en) Building material welding surface defect detection method based on computer vision
CN109711288A (en) Remote sensing ship detecting method based on feature pyramid and distance restraint FCN
CN107248159A (en) A kind of metal works defect inspection method based on binocular vision
CN108288271A (en) Image detecting system and method based on three-dimensional residual error network
US8548253B2 (en) Fast line linking
US20070036434A1 (en) Topology-Based Method of Partition, Analysis, and Simplification of Dynamical Images and its Applications
CN110544233B (en) Depth image quality evaluation method based on face recognition application
CN104933709A (en) Automatic random-walk CT lung parenchyma image segmentation method based on prior information
CN114219794B (en) Method and system for evaluating surface quality of shaving board based on machine vision
CN103940708B (en) A kind of Quick Measurement, sophisticated category method of steel holotype state crystal grain
CN104850822B (en) Leaf identification method under simple background based on multi-feature fusion
CN101487838A (en) Extraction method for dimension shape characteristics of profiled fiber
CN109376740A (en) A kind of water gauge reading detection method based on video
CN105005764A (en) Multi-direction text detection method of natural scene
CN115100206B (en) Printing defect identification method for textile with periodic pattern
CN116664559B (en) Machine vision-based memory bank damage rapid detection method
CN114862855B (en) Textile defect detection method and system based on template matching
CN107016680B (en) A kind of pest image background minimizing technology based on conspicuousness detection
CN113256571B (en) Vector graphic feature point extraction method based on direction feature and local uniqueness
CN108182705A (en) A kind of three-dimensional coordinate localization method based on machine vision
CN114022439A (en) Flexible circuit board defect detection method based on morphological image processing
CN113609984A (en) Pointer instrument reading identification method and device and electronic equipment
CN115331245A (en) Table structure identification method based on image instance segmentation
CN113689431B (en) Industrial product appearance defect detection method and device
CN113688846B (en) Object size recognition method, readable storage medium, and object size recognition system

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