CN112767497A - High-robustness calibration device based on circular calibration plate and positioning method - Google Patents

High-robustness calibration device based on circular calibration plate and positioning method Download PDF

Info

Publication number
CN112767497A
CN112767497A CN202110099590.8A CN202110099590A CN112767497A CN 112767497 A CN112767497 A CN 112767497A CN 202110099590 A CN202110099590 A CN 202110099590A CN 112767497 A CN112767497 A CN 112767497A
Authority
CN
China
Prior art keywords
calibration
image
calibration plate
point
xyz
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.)
Pending
Application number
CN202110099590.8A
Other languages
Chinese (zh)
Inventor
王朋
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.)
Smarteye Tech Ltd
Original Assignee
Smarteye Tech 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 Smarteye Tech Ltd filed Critical Smarteye Tech Ltd
Priority to CN202110099590.8A priority Critical patent/CN112767497A/en
Publication of CN112767497A publication Critical patent/CN112767497A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/80—Analysis of captured images to determine intrinsic or extrinsic camera parameters, i.e. camera calibration
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • 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
    • G06T7/10—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
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • Geometry (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Evolutionary Computation (AREA)
  • Evolutionary Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Analysis (AREA)
  • Length Measuring Devices By Optical Means (AREA)

Abstract

The invention relates to a high-robustness calibration device and a positioning method based on a circular calibration plate, wherein the circular calibration plate is placed in the public view of all cameras to be calibrated, and each camera to be calibrated xyz isiRespectively align the calibration plates XYZjImageijExtracting the contour by using the algorithm of Canny and the like to obtain an initial contour group, and then carrying out primary filtering to obtain the final contour groupValue graph ImageijComputing grid features
Figure DDA0002915531310000011
Establishing an undirected graph, carrying out cluster analysis, carrying out edge detection and contour extraction filtering again, calculating normalized world coordinates after sequencing, and not needing to pre-configure the size, the interval, the number and the like of dots or rings; the algorithm does not need to try to adaptively adjust the image threshold parameter and the like, so that the calibration speed is high; the robustness is stronger for shielding, dirty calibration plates, similar calibration plates in calibration scenes and the like.

Description

High-robustness calibration device based on circular calibration plate and positioning method
Technical Field
The invention belongs to the technical field of optical image measurement and machine vision application, and particularly relates to a high-robustness calibration device and a positioning method based on a circular calibration plate.
Background
Camera calibration is the first step of many machine vision applications, and since the system application accuracy basically depends on the accuracy of the calibration result, the camera calibration plays a significant role in the machine vision applications.
The types of calibration plates common in the prior art can be roughly divided into three categories: a checkerboard calibration board, a dot/ring calibration board and calibration boards of various custom shapes. Wherein, the specially-made or custom-shaped calibration plates, such as ChAruco calibration plate and Kalibr calibration plate, have complex design and processing, which cause cost increase and poor applicability; although the checkerboard angular point calibration plate is easy to position and has high detection precision, the checkerboard angular point calibration plate is widely applied, but is easily influenced by noise and fuzziness, and the key point is that the detection precision is reduced because of the sharp black-white change at the checkerboard angular points of a system (such as a structured light measurement system) needing to calibrate the equipment similar to an inverse pinhole imaging camera of a projector; aiming at the defects of the angular point calibration plate, a dot calibration plate or a circular calibration plate is generally adopted in the industry.
For the detection of the dot or circular calibration plate, the existing methods generally include the following methods. The method is characterized in that the method is adopted by an OpenCV open source library, the method carries out circulating binarization operation on the whole image and extracts a contour, then the contour area, the number of dots, the roundness and the like are filtered according to the configuration of a user, and finally the dots which are closest to each other and meet the configuration number of the user are found out by a clustering analysis method to serve as detection results.
Other methods, such as the "projector calibration method" of high-health et al (application No. CN201710293985.5), use a homemade calibration board of dots, which is divided into three types according to the radius for determining the number and orientation of dots, and although this method can reduce the difficulty of user configuration to some extent on the calibration board, it still cannot solve the problem of detection failure under the conditions of shielding, dirt, and interference of similar objects on the calibration board, and the difficulty of distinguishing and detecting three types of dots increases when the angle of the calibration board tangential to the camera imaging plane is large, which may also cause detection failure.
Also, for example, in the patent entitled "camera calibration board, calibration method and camera" of cheng et al (application No. CN201811526921.6), the boundary size, dot pitch and size of the calibration board are constrained to improve the robustness of the calibration board detection, but this also increases the design and manufacturing cost of the calibration board, and also requires the user to configure parameters in advance to reduce the effects of dirt and similar interferents.
Disclosure of Invention
In order to solve the above problems in the prior art, the present invention aims to provide a high robustness calibration apparatus and a positioning method based on a circular calibration plate, which can calibrate multiple devices simultaneously, have the advantages of high detection speed, good detection effect, low cost and convenient operation, and can prevent the interference of shielding, dirt and the like.
The technical scheme adopted by the invention is as follows:
a high robustness calibration device based on a circular calibration plate comprises a background plate and at least one camera to be calibrated, wherein the circular calibration plate, a similar interferent, a dirty spot stain and a shielding object are arranged between the background plate and the camera to be calibrated, and the circular calibration plate, the similar interferent, the dirty spot stain and the shielding object are fixedly supported on the background plate through transparent support rods or are suspended above the background plate through steel wires; the circular calibration plate, the similar interferent, the dirty spot stain and the shelter are positioned in the public view range of all the cameras to be calibrated.
The circular calibration plate is provided with dots or rings with any size, any number and any regular arrangement.
A positioning method of the high-robustness calibration device based on the circular calibration plate comprises the following steps:
s1, marking the camera to be calibrated as xyzi(I ═ 1, 2, 3., I ≧ 1), the common field of view of all cameras to be calibrated is marked as #iFOVi；
S2, placing the circular calibration plate in the public view of all cameras to be calibrated, and marking the position and posture of the calibration plate at the moment as XYZjThe circle centers on the calibration plate are at the pose XYZjHas the coordinates of
Figure BDA0002915531290000031
Wherein
Figure BDA0002915531290000032
Representing calibration plate pose XYZjRelative to camera set { xyziConversion matrix of PkRepresenting world coordinates of each circle center on the calibration plate;
s3, each camera to be calibrated xyziRespectively align the calibration plates XYZjImage for taking one pictureijAssume that each image resolution is (w)i，hi) Then, the following pretreatment is carried out: binarization is carried out on the self-adaptive threshold value to obtain a black-white binary ImageijThen extracting the contour by using an algorithm such as Canny and the like to obtain an initial contour group { contour { constant }ijl}， l＝1，2，3...，Lij；
S4, performing preliminary filtering on the initial contour group;
using each contour contourrijlCalculating the Area of the outline AreaijlShape factor circulationijlAnd Hu rotation and scaling invariant moments HuMomentsijl: if it is not
Figure BDA0002915531290000033
And is
Figure BDA0002915531290000034
And is
Figure BDA0002915531290000035
And keeping the contour, otherwise deleting and filling the pixel point p in the contour in the binary image into white:
Imageij(p∈contourijl)＝255
in the above judgment condition
Figure BDA0002915531290000036
And
Figure BDA0002915531290000037
all represent a threshold value;
s5, utilizing the binary Image obtained in the previous stepijComputing grid features
Figure BDA0002915531290000038
S6, pair
Figure BDA0002915531290000039
Local maximum point set
Figure BDA00029155312900000310
Eij represents the number of maximum points calculated in the current grid feature, Eij≥3；
S7, obtaining the extreme point in the last step
Figure BDA00029155312900000311
As vertices, an undirected graph is built
Figure BDA00029155312900000312
Wherein
Figure BDA00029155312900000313
Representing a connecting line or edge between two vertices,
Figure BDA00029155312900000314
the construction rules of (1) are as follows: randomly selecting two vertexes as reference points, calculating the pixel distance between the two vertexes, and taking the distance as a reference distance; if extreme point setIf the distance from any extreme point to any point of the two reference points is not greater than the reference distance, connecting the two reference points to form a non-directional edge; to obtain E ij1 side, i.e. Gij＝Eij-1；
S8, mixing GijEdge
Figure BDA0002915531290000041
Clustering analysis is divided into two categories according to directions, and each category respectively calculates mean values as reference vectors and expresses the mean values as
Figure BDA0002915531290000042
And
Figure BDA0002915531290000043
s9, edge detection and contour extraction are carried out again on the binary image in S4, and the center of each contour is determined
Figure BDA0002915531290000044
Repeat S7 as the vertex, create an undirected graph
Figure BDA0002915531290000045
And using a reference
Figure BDA0002915531290000046
And
Figure BDA0002915531290000047
vector length and direction of (1) to undirected graph
Figure BDA0002915531290000048
Top point of (2)
Figure BDA0002915531290000049
And
Figure BDA00029155312900000410
filtering;
s10, establishing center point image coordinates of circle or ring
Figure BDA00029155312900000411
With corresponding physical coordinates PkOne-to-one mapping between, i.e. to, the vertices of an undirected graph
Figure BDA00029155312900000412
Sorting is carried out;
s11, calculating other points relative to OijWorld coordinates of (a):
firstly, undirected graph
Figure BDA00029155312900000413
In calculating each point
Figure BDA00029155312900000414
To OijOf points in the reference direction uijAnd vijMinimum manhattan distance of
Figure BDA00029155312900000415
Wherein.
Figure BDA00029155312900000416
And (c).
Figure BDA00029155312900000417
Respectively represent corresponding points in the figure
Figure BDA00029155312900000418
The horizontal and vertical coordinates of (1);
then calculate
Figure BDA00029155312900000419
To OijImage pixel distance of a point
Figure BDA00029155312900000420
And projects it to the reference direction uijAnd vijTo obtain
Figure BDA00029155312900000421
And
Figure BDA00029155312900000422
if it is not
Figure BDA00029155312900000423
Figure BDA00029155312900000424
Then the point is retained, otherwise the point is deleted, at which time eij≤k；
Point-aligning set
Figure BDA00029155312900000425
Push button
Figure BDA00029155312900000426
And
Figure BDA00029155312900000427
sorting from small to large and calculating the world coordinates of corresponding points
Figure BDA00029155312900000428
Wherein sign (·) represents a symbolic function, W and H represent the center distance of adjacent dots or circular ring dots on the calibration board in the horizontal direction and the vertical direction respectively;
s12, obtaining the image coordinate of the center point of the circle or the circular ring
Figure BDA0002915531290000051
And corresponding world coordinates
Figure BDA0002915531290000052
And S13, further optimizing.
In the step S4
Figure BDA0002915531290000053
Get
Figure BDA0002915531290000054
Get
Figure BDA0002915531290000055
Taking 0.6 percent,
Figure BDA0002915531290000056
Taking 0.9 percent,
Figure BDA0002915531290000057
Taking 0.13 percent,
Figure BDA0002915531290000058
Take 0.18.
The method for calculating the grid features in step S5 performs autocorrelation transformation on the image according to the following formula:
Figure BDA0002915531290000059
wherein p.x and p.y respectively represent the horizontal and vertical coordinates, C, of the image of the pixel point pijRepresenting camera xyzciPhotographed XYZjAnd (5) performing autocorrelation transformation on the calibration plate image of the pose.
In said step S6, item Eij＝9。
In the step S10, 5 great circle points are arranged on the calibration plate; first, the vertex group is matched
Figure BDA00029155312900000510
Performing cluster analysis according to area design characteristics to find directional vertexes
Figure BDA00029155312900000511
(dij is 1, 2, 3, D; 5 is taken as the number of directional vertexes, namely D is 5), and then the reference point is found according to the design characteristic of the included angle
Figure BDA00029155312900000512
And a reference direction
Figure BDA00029155312900000513
And
Figure BDA00029155312900000514
the judgment condition in the step S11
Figure BDA00029155312900000515
Taking 0 percent,
Figure BDA00029155312900000516
And taking 2.
The optimization method in step S13 is as follows:
will point set
Figure BDA00029155312900000517
As an initial value, then at each point
Figure BDA00029155312900000518
Figure BDA00029155312900000519
And
Figure BDA00029155312900000520
further optimizing the calculation process of the point center in the formed region of interest
Figure BDA00029155312900000521
Alpha and beta respectively represent scaling coefficients of the lengths of the two reference directions, and the calculation process
Figure BDA00029155312900000522
Indicating initial value using center point
Figure BDA00029155312900000523
Reference direction uij，vijAnd an original imageijThe process of calculating the center point is optimized, and ellipse fitting and projective transformation correction processes may be exemplified.
In step S13, α ═ β ═ 0.4.
The optimal point center calculation process in step S13 is an ellipse fitting and projective transformation correction process.
Setting a calibration board in the step S2 and calculating a binary Image in the steps S3-S4ijThe process of (2) is as follows: at pose XYZjBefore the calibration plate is placed, firstly, the camera xyz to be calibrated passes throughiGenerating a background image of the shooting scene
Figure BDA0002915531290000061
h＝1，2，3...，imageihRepresenting a background image of a set or a series of non-calibration plates shot by each camera to be calibrated,
Figure BDA0002915531290000062
representing the process of image computing background; then in the pose XYZjPlacing the calibration plate and using each camera to be calibrated xyziImageijAnd obtaining a calibration plate area by using a background removal technology to obtain a binary image
Figure BDA0002915531290000063
The above-mentioned
Figure BDA0002915531290000064
Calculating a background image by adopting an optical flow method; background removal process
Figure BDA0002915531290000065
Optical flow methods are also used.
The invention has the beneficial effects that:
a high robustness calibration device and a positioning method based on a circular calibration plate are provided, wherein the circular point or circular ring calibration plate is placed in the common visual field of all cameras to be calibrated, and each camera to be calibrated xyz is to be calibratediRespectively align the calibration plates XYZjImageijExtracting the contour by using an algorithm such as Canny and the like to obtain an initial contour group, and then carrying out initial filtering to obtain a value ImageijComputing grid features
Figure BDA0002915531290000066
Establishing an undirected graph, carrying out cluster analysis, carrying out edge detection and contour extraction filtering again, calculating normalized world coordinates after sequencing, and not needing to pre-configure the size, the interval, the number and the like of dots or rings; the algorithm does not need to try to adjust image threshold parameters and the like in a self-adaptive manner, and the calibration speed is high; robustness is stronger for shielding, dirty calibration plates, similar calibration plates in calibration scenes and the like;
1. the sizes, the intervals and the number of circular points or circular rings of the calibration plate are not required to be configured in advance, on one hand, the calibration plate with a specific model is not required to be specified, the applicability of the calibration algorithm is improved, on the other hand, the professional requirements of related workers can be reduced, and the full-automatic positioning can be realized;
2. the interference of other similar objects in a scene shot by a camera can be automatically eliminated, the method has good resistance to the situations of dirt, shielding and the like of a calibration plate, the detection process is quick and quick, the calibration work difficulty can be reduced, and the calibration work efficiency is further improved;
3. the calibration plate is not limited to a dot or ring calibration plate with a specific model, the dots or rings on the calibration plate can be distributed in a rectangular array or any other arrangement form of a non-rectangular array, and the calibration plate can be positioned even under the extremely severe condition that the image of the calibration plate is fuzzy;
drawings
FIGS. 1-6 are schematic diagrams illustrating defects in a calibration plate inspection method according to the prior art;
FIG. 7 is a schematic diagram of a detection scene of a circular calibration plate of the high robustness calibration device based on the circular calibration plate according to the present invention;
FIG. 8 is a schematic diagram of the orientation scheme of the calibration plate of the high robustness calibration device based on the circular calibration plate according to the present invention;
fig. 9 is a schematic diagram of a calibration plate dot positioning scheme of the high-robustness calibration device based on the circular calibration plate.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying 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.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
As shown in fig. 1 to 6, the calibration plate detection method in the prior art has the following corresponding defects:
(1) dirty spots and stains on the surface;
(2) shielding;
(3) super camera view range;
(4) similar interferences;
(5) the dip angle is large;
(6) and blurred.
The invention aims to provide a high-robustness calibration device and a positioning method based on a circular calibration plate, aiming at the defects of the prior art.
The first embodiment of the present invention provides a high robustness calibration device based on a circular calibration board, as shown in fig. 7, a background board 8 and at least one camera to be calibrated are first set, and fig. 7 shows three cameras to be calibrated, namely a first camera 1 to be calibrated, a second camera 2 to be calibrated, and a third camera to be calibrated; a circular calibration plate 4, a similar interferent 5, a dirty spot stain 6 and a shelter 7 are arranged between the background plate and all cameras to be calibrated, and the circular calibration plate 4, the similar interferent 5, the dirty spot stain 6 and the shelter 7 are respectively and fixedly supported on the background plate through transparent support rods or are suspended above the background plate through steel wires; and the circular calibration plate, similar disturbance 5, dirty spot stain 6 and obstruction 7 are all located simultaneously within the common field of view of all cameras to be calibrated.
The circular calibration plate is provided with dots or rings with any size, any number and any regular arrangement.
The hardware system structure related to the invention is shown in figure 7: one or more cameras xyz to be calibratedi(I ═ 1, 2, 3., I ≧ 1); a circular ring or dot calibration plate; and other obstructions, similar interferents, and dirty spots that are passively present in the detection scene. The following embodiments and examples are illustrated by way of example only, but the invention is not limited thereto.
The embodiment provides a positioning method of a high-robustness calibration device based on a circular calibration plate, which comprises the following steps:
(1) placing a circular ring or a circular dot calibration board on each camera xyz to be calibratediPublic view #iFOViIn the field, assume that the position and posture of the calibration plate at the moment are XYZjThe circle centers on the calibration plate are at the pose XYZjHas the coordinates of
Figure BDA0002915531290000081
Wherein
Figure BDA0002915531290000082
Representing calibration plate pose XYZjRelative to camera set { xyziConversion matrix of PkAnd the world coordinates of each circle center on the calibration plate are expressed. However, as shown in FIGS. 1-6, due to the problems of the occlusion and the scaling plate exceeding the visual field, the dots in a part of the camera visual field may not be visible in another part of the camera visual field, i.e., k ≦ mxn.
(2) Each camera to be calibrated XYZ is respectively aligned to the calibration platejImage for taking one pictureijAssume that each image resolution is (w)i，hi) Then, the following pretreatment is carried out: binarization is carried out by self-adaptive threshold value to obtain black-white binary ImageijThen extracting the contour by using the Canny algorithm and the like to obtain an initial wheelContour group { contourijl}，l＝1，2，3...，LijGenerally, the number of the extracted outlines is far larger than the number of the actual dots, namely L, due to the interference of noise, illumination, similar or non-similar impurities in the background and the likeij> m × n, these contours need to be rationalized and preliminary filtered for subsequent steps to successfully detect the calibration plate. One method of detection is to use each contour contourrijlCalculating the Area of the outline AreaijlShape factor circulationijlAnd Hu rotation and zoom invariant moments HuMomentsijl(ii) a If it is not
Figure BDA0002915531290000091
And is
Figure BDA0002915531290000092
Figure BDA0002915531290000093
And is
Figure BDA0002915531290000094
And keeping the contour, otherwise deleting and filling the pixel point p in the contour in the binary image into white:
Imageij(p∈contourijl)＝255
in the above judgment condition
Figure BDA0002915531290000095
And
Figure BDA0002915531290000096
all represent threshold values, as an example
Figure BDA0002915531290000097
Preference selection
Figure BDA0002915531290000098
Preference selection
Figure BDA0002915531290000099
Preferably 0.6,
Figure BDA00029155312900000910
Preferably 0.9,
Figure BDA00029155312900000911
Preferably 0.13,
Figure BDA00029155312900000912
Preferably 0.18.
(3) Utilizing the binary Image obtained in the last stepijComputing grid features
Figure BDA00029155312900000913
As an example, one method of computing the mesh characteristics is to perform an autocorrelation transform on the image as follows
Figure BDA00029155312900000914
Wherein p.x and p.y respectively represent the horizontal and vertical coordinates, C, of the image of the pixel point pijRepresenting camera xyzciPhotographed XYZjAnd (5) performing autocorrelation transformation on the calibration plate image of the pose.
(4) To pair
Figure BDA00029155312900000915
Local maximum point is calculated
Figure BDA00029155312900000916
(eij＝1，2，3，...，Eij)，EijRepresenting the number of maximum points calculated in the current grid feature, theoretically EijInfinite, but these extreme points are distributed in the same array as the dots or the ring points on the calibration plate, so a small number of extreme points can be selected, and E should be selected to ensure the subsequent stepsij≧ 3, E is preferred as an exampleij＝9。
(5) The extreme point obtained in the last step
Figure RE-GDA0003008637570000101
As vertices, an undirected graph is built
Figure RE-GDA0003008637570000102
Wherein
Figure RE-GDA0003008637570000103
Representing a connecting line or edge between two vertices,
Figure RE-GDA0003008637570000104
the following rules may be used: and randomly selecting two vertexes as a reference, calculating the pixel distance between the two vertexes as a reference distance, and connecting the two reference points to form a non-directional edge if the distance from the extreme point set to any one of the two reference points to another extreme point is not greater than the reference distance. Thus for EijA feature point, can obtain E ij1 side, i.e. Gij＝Eij-1。
(6) G obtained in the last stepijEdge
Figure BDA0002915531290000105
Clustering analysis is divided into two categories according to direction, each category calculates mean value as reference vector and expresses the mean value as reference vector
Figure BDA0002915531290000106
And
Figure BDA0002915531290000107
these two edges contain the pitch and orientation of the calibration plate grid.
(7) Performing edge detection and contour extraction again on the binary image obtained in the step (2), and centering each contour
Figure BDA0002915531290000108
Creating an undirected graph as a vertex using the same method as in step (5)
Figure BDA0002915531290000109
And using a reference
Figure BDA00029155312900001010
And
Figure BDA00029155312900001011
vector length and direction of (1) to undirected graph
Figure BDA00029155312900001012
Top point of (2)
Figure BDA00029155312900001013
And
Figure BDA00029155312900001014
and filtering is carried out, so that not only can interference points on the surface of the calibration plate due to dirt and the like be eliminated, but also objects similar to the calibration plate which possibly appear in a scene shot by a camera can be eliminated.
(8) When the calibration plate is used for calibrating the camera, the image coordinates of the central point of the circle or the ring are required to be established
Figure BDA00029155312900001015
With corresponding physical coordinates PkOne-to-one mapping between, i.e. requiring vertex to undirected graph
Figure BDA00029155312900001016
Sequencing; in addition to this, for a multi-camera system, i.e. I > 1, such a one-to-one mapping also needs to be established in order to calibrate the relationship between the cameras, for which purpose the circle or circle point on the calibration plate needs to be oriented and positioned. By way of example, the present embodiment is preferably illustrated with a calibration plate oriented with five great circles, as shown in fig. 8, but not limited thereto. First, the vertex group is mapped
Figure BDA00029155312900001017
Clustering analysis is performed on design features (which may be areas, as an example) to find directional vertices therein
Figure BDA0002915531290000111
( d ij1, 2, 3, ·, D; the number of orientation vertices can be preferably selected to be 5, i.e., D is 5, as an example, and then the reference point can be found according to the design feature (such design feature can be the included angle, as an example)
Figure BDA00029155312900001127
And a reference direction
Figure BDA0002915531290000112
And
Figure BDA0002915531290000113
(9) using reference point OijReference direction uijAnd vijAfter determining the orientation of the calibration plate, other points relative to O need to be calculated in the following mannerijWorld coordinates of (a): firstly, undirected graph
Figure BDA0002915531290000114
In calculating each point
Figure BDA0002915531290000115
(eijNot equal to d) to OijOf points in the reference direction uijAnd vijMinimum manhattan distance of
Figure BDA0002915531290000116
Wherein.
Figure BDA0002915531290000117
And (c).
Figure BDA0002915531290000118
Respectively represent corresponding points in the figure
Figure BDA0002915531290000119
The horizontal and vertical coordinates of (1); however, the device is not suitable for use in a kitchenPost-calculation
Figure BDA00029155312900001110
To OijImage pixel distance of a point
Figure BDA00029155312900001111
And projects it to the reference direction uijAnd vijTo obtain
Figure BDA00029155312900001112
And
Figure BDA00029155312900001113
if it is not
Figure BDA00029155312900001114
This point is retained (as an example, the condition is determined here
Figure BDA00029155312900001115
Can be taken 0,
Figure BDA00029155312900001116
2) otherwise, the point is deleted, so that the error detection of partial round points or circular ring points caused by overlarge inclination angle of the calibration plate relative to the camera imaging plane can be eliminated, and at the moment eijK is not more than k; point set
Figure BDA00029155312900001117
Push button
Figure BDA00029155312900001118
And
Figure BDA00029155312900001119
sequencing from small to large, and calculating world coordinates of corresponding points
Figure BDA00029155312900001120
Wherein sign (·) represents a sign function, and W and H represent center distances of adjacent dots or circular ring dots in horizontal and vertical directions on the calibration board, respectively.
(10) The image coordinates of the center point of the circle or the ring are calculated by the steps
Figure BDA00029155312900001121
And corresponding world coordinates
Figure BDA00029155312900001122
Due to the fact that
Figure BDA00029155312900001123
The coordinate values of (2) are calculated by contour extraction in the step (7), and in some vision systems, the precision may not be high enough, and the point set may be collected
Figure BDA00029155312900001124
As an initial value, then at each point
Figure BDA00029155312900001125
And
Figure BDA00029155312900001126
further optimizing the calculation process of the point center in the formed region of interest
Figure BDA0002915531290000121
To further improve the calibration accuracy, where α and β respectively represent scaling coefficients for two reference direction lengths, and preferably, as an example, α ═ β ═ 0.4 is selected; calculation process
Figure BDA0002915531290000122
Indicating initial value using center point
Figure BDA0002915531290000123
Reference direction uij，vijAnd an original imageijOptimizing the Process of calculating the center Point, as an example
Figure BDA0002915531290000124
May be an ellipse fitting and projective transformation correction process.
The circular calibration plate positioning method based on image autocorrelation has the following advantages:
1. the method does not need to configure the size, the interval and the number of the dots or the rings of the calibration plate in advance, does not need to specify the calibration plate with a specific model on one hand, improves the applicability of the calibration algorithm, can reduce the professional requirements of related workers on the other hand, and can realize full-automatic positioning.
2. The method can automatically eliminate the interference of other similar objects in the scene shot by the camera, has good resistance to the situations of dirt and shielding of the calibration plate, and has quick detection process, thereby reducing the difficulty of calibration work and further improving the efficiency of the calibration work.
3. The method is not limited to a specific type of dot or ring calibration plate, the distribution of dots or rings on the calibration plate can also be a non-rectangular array, and the positioning of the calibration plate can be realized even under the extremely severe condition that the image of the calibration plate is blurred.
Example two: in the first embodiment, the calibration board is placed in the step (1) and the binary Image is calculated in the step (2)ijThe following alternatives can be used for the process of (1): at pose XYZjBefore placing the calibration plate, firstly, using each camera to be calibrated xyziGenerating a background image of the shooting scene
Figure BDA0002915531290000125
Figure BDA0002915531290000125
Figure BDA0002915531290000125
Figure BDA0002915531290000125
1, 2, 3, where imageihA background image representing one or a series of non-calibration plates shot by each camera to be calibrated,
Figure BDA0002915531290000126
Representing the process of image computing context, here by way of example
Figure BDA0002915531290000127
The background image can be calculated by an optical flow method; then in the pose XYZjPlacing the calibration plate and using each camera to be calibrated xyziImageijAnd obtaining a binary image by obtaining the area of the calibration plate by using the background removal technology
Figure BDA0002915531290000128
As an example here the background removal process
Figure BDA0002915531290000129
Optical flow methods may be used.
The invention is not limited to the above alternative embodiments, and any other various products can be obtained by anyone in the light of the present invention, but any changes in shape or structure thereof, all of which fall within the scope of the claims of the present invention, fall within the protection scope of the present invention.

Claims (10)

1. The utility model provides a high robustness calibration device based on circular calibration board which characterized in that: the camera calibration device comprises a background plate and at least one camera to be calibrated, wherein a circular calibration plate, a similar interference object, a dirty spot stain and a shielding object are arranged between the background plate and the camera to be calibrated, and the circular calibration plate, the similar interference object, the dirty spot stain and the shielding object are fixedly supported on the background plate through transparent support rods or are suspended above the background plate through steel wires; the circular calibration plate, the similar interferent, the dirty spot stain and the shelter are positioned in the public view range of all the cameras to be calibrated.
2. The high robustness calibration device based on the circular calibration plate as claimed in claim 1, wherein: the circular calibration plate is provided with dots or rings with any size, any number and any regular arrangement.
3. A positioning method of the high-robustness calibration device based on the circular calibration plate as claimed in any one of claims 1-2, wherein: the method comprises the following steps:
s1, marking the camera to be calibrated as xyzi(i＝1，2，3...，I，I≥1)；
S2, marking the public view of all cameras to be calibrated as #iFOViSuppose that at this time, the pose XYZ of the calibration plate isjThe circle centers on the calibration plate are at the pose XYZjHas the coordinates of
Figure FDA0002915531280000014
Wherein
Figure FDA0002915531280000015
Representing calibration plate pose XYZjRelative to camera set { xyziConversion matrix of PkRepresenting world coordinates of each circle center on the calibration plate;
s3, each camera to be calibrated xyziRespectively align the calibration plates XYZjImage for taking one pictureijAssume that each image resolution is (w)i，hi) Then, the following pretreatment is carried out: binarization is carried out by self-adaptive threshold value to obtain black-white binary ImageijThen extracting the contour by using an algorithm such as Canny and the like to obtain an initial contour group { contour { constant }ijl}，l＝1，2，3...，Lij；
S4, performing preliminary filtering on the initial contour group;
using each contour contourrijlCalculating the Area of the outline AreaijlShape factor circulationijlAnd Hu rotation and scaling invariant moments HuMomentsijl: if it is not
Figure FDA0002915531280000011
And is
Figure FDA0002915531280000012
And is
Figure FDA0002915531280000013
And keeping the contour, otherwise deleting and filling the pixel point p in the contour in the binary image into white:
Imageij(p∈contourijl)＝255
in the above judgment condition
Figure FDA0002915531280000021
And
Figure FDA0002915531280000022
all represent a threshold value;
s5, utilizing the binary Image obtained in the previous stepijComputing grid features
Figure FDA0002915531280000023
S6, pair
Figure FDA0002915531280000024
Local maximum point set
Figure FDA0002915531280000025
EijRepresenting the number of calculated maxima points in the current grid feature, Eij≥3；
S7, obtaining the extreme point in the last step
Figure FDA0002915531280000026
As vertices, an undirected graph is built
Figure FDA0002915531280000027
Wherein
Figure FDA0002915531280000028
Representing a connecting line or edge between two vertices,
Figure FDA0002915531280000029
the construction rules of (1) are as follows: randomly selecting two vertexes as reference points, calculating the pixel distance between the two vertexes, and taking the distance as a reference distance; if the distance from any extreme point in the extreme point set to any one of the two reference points is not more than the reference distance, connecting the two reference point patternsForming a non-directional edge; to obtain Eij1 side, i.e. Gij＝Eij-1；
S8, mixing GijEdge
Figure FDA00029155312800000210
Clustering analysis is divided into two categories according to direction, each category calculates mean value as reference vector and expresses the mean value as reference vector
Figure FDA00029155312800000211
And
Figure FDA00029155312800000212
s9, edge detection and contour extraction are carried out again on the binary image in S4, and the center of each contour is determined
Figure FDA00029155312800000213
Repeating S7 as the vertex to create an undirected graph
Figure FDA00029155312800000214
And using a reference
Figure FDA00029155312800000215
And
Figure FDA00029155312800000216
vector length and direction of (1) to undirected graph
Figure FDA00029155312800000217
Top point of (2)
Figure FDA00029155312800000218
And
Figure FDA00029155312800000219
filtering;
s10, establishing center point image coordinates of circle or ring
Figure FDA00029155312800000220
With corresponding physical coordinates PkOne-to-one mapping between, i.e. to, the vertices of an undirected graph
Figure FDA00029155312800000221
Sorting is carried out;
s11, calculating other points relative to OijWorld coordinates of (a):
firstly, undirected graph
Figure FDA00029155312800000222
In calculating each point
Figure FDA00029155312800000223
To OijOf points in the reference direction uijAnd vijMinimum manhattan distance of
Figure FDA0002915531280000031
Wherein
Figure FDA0002915531280000032
And
Figure FDA0002915531280000033
respectively represent corresponding points in the figure
Figure FDA0002915531280000034
The horizontal and vertical coordinates of (1);
then calculate
Figure FDA0002915531280000035
To OijImage pixel distance of a point
Figure FDA0002915531280000036
And project itTo the reference direction uijAnd vijTo obtain
Figure FDA0002915531280000037
And
Figure FDA0002915531280000038
if it is not
Figure FDA0002915531280000039
Figure FDA00029155312800000310
Then the point is retained, otherwise the point is deleted, at which time eij≤k；
Point-aligning set
Figure FDA00029155312800000311
Push button
Figure FDA00029155312800000312
And
Figure FDA00029155312800000313
sorting from small to large, and calculating world coordinates of corresponding points
Figure FDA00029155312800000314
Wherein sign (·) represents a symbolic function, W and H represent the center distance of adjacent dots or circular ring dots on the calibration board in the horizontal direction and the vertical direction respectively;
s12, obtaining the image coordinate of the center point of the circle or the circular ring
Figure FDA00029155312800000315
And corresponding world coordinates
Figure FDA00029155312800000316
And S13, further optimizing.
4. The method for positioning the calibration device with high robustness based on the circular calibration plate as claimed in claim 3, wherein: in the step S4
Figure FDA00029155312800000317
Take 0.1 Xwi×hi、
Figure FDA00029155312800000318
Take 0.5 Xwi×hi、
Figure FDA00029155312800000319
Taking 0.6 percent,
Figure FDA00029155312800000320
Taking 0.9 percent,
Figure FDA00029155312800000321
Taking 0.13 percent,
Figure FDA00029155312800000322
Take 0.18.
5. The method for positioning the calibration device with high robustness based on the circular calibration plate as claimed in claim 3, wherein: the method for calculating the grid features in step S5 performs autocorrelation transformation on the image according to the following formula:
Figure FDA00029155312800000323
wherein p.x and p.y respectively represent the horizontal and vertical coordinates, C, of the image of the pixel point pijRepresenting camera xyzciPhotographed XYZjAnd (5) performing autocorrelation transformation on the calibration plate image of the pose.
6. Root of herbaceous plantThe method for positioning the highly robust calibration device based on the circular calibration plate as claimed in claim 3, wherein: in said step S6, item Eij＝9。
7. The method for positioning the calibration device with high robustness based on the circular calibration plate as claimed in claim 3, wherein: in the step S10, 5 dots are provided on the calibration board; first, the vertex group is matched
Figure FDA0002915531280000041
Performing clustering analysis according to area design characteristics to find directional vertexes
Figure FDA0002915531280000042
Taking 5 as the number of the orientation vertexes, namely D is 5), and finding out the datum point according to the design characteristic of the included angle
Figure FDA0002915531280000043
And a reference direction
Figure FDA0002915531280000044
And
Figure FDA0002915531280000045
8. the method for positioning a calibration device with high robustness based on a calibration plate with dots or rings as claimed in claim 3, wherein: the judgment condition in the step S11
Figure FDA0002915531280000046
Taking 0 percent,
Figure FDA0002915531280000047
And taking 2.
9. The method for positioning the calibration device with high robustness based on the circular calibration plate as claimed in claim 3, wherein: the optimization method in step S13 is as follows:
will point set
Figure FDA0002915531280000048
As an initial value, then at each point
Figure FDA0002915531280000049
Figure FDA00029155312800000410
And
Figure FDA00029155312800000411
further optimizing the calculation process of the point center in the formed region of interest
Figure FDA00029155312800000412
Alpha and beta respectively represent scaling coefficients of the lengths of the two reference directions, and the calculation process
Figure FDA00029155312800000413
Indicating initial value using center point
Figure FDA00029155312800000414
Reference direction uij，vijAnd an original imageijThe process of calculating the center point is optimized,
Figure FDA00029155312800000415
is the ellipse fitting and projective transformation correction process (claim 11 is removed), α ═ β ═ 0.4.
10. The method for positioning the calibration device with high robustness based on the circular calibration plate as claimed in claim 3, wherein: setting a calibration board in the step S2 and calculating a binary Image in the steps S3-S4ijThe process of (2) is as follows: at pose XYZjBefore the calibration plate is placed, firstly, the camera xyz to be calibrated passes throughiGenerating a pair of shooting scenesBackground picture
Figure FDA00029155312800000416
h＝1，2，3...，imageihRepresenting a background image of a or a series of non-calibration plates taken by each camera to be calibrated,
Figure FDA00029155312800000417
representing the process of image computing background; then in the pose XYZjPlacing the calibration plate and using each camera to be calibrated xyziImageijAnd obtaining a calibration plate area by using a background removal technology to obtain a binary image
Figure FDA0002915531280000051
Figure FDA0002915531280000052
Calculating a background image by adopting an optical flow method; background removal process
Figure FDA0002915531280000053
Optical flow methods are also used.
CN202110099590.8A 2021-01-25 2021-01-25 High-robustness calibration device based on circular calibration plate and positioning method Pending CN112767497A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110099590.8A CN112767497A (en) 2021-01-25 2021-01-25 High-robustness calibration device based on circular calibration plate and positioning method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110099590.8A CN112767497A (en) 2021-01-25 2021-01-25 High-robustness calibration device based on circular calibration plate and positioning method

Publications (1)

Publication Number Publication Date
CN112767497A true CN112767497A (en) 2021-05-07

Family

ID=75707358

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110099590.8A Pending CN112767497A (en) 2021-01-25 2021-01-25 High-robustness calibration device based on circular calibration plate and positioning method

Country Status (1)

Country Link
CN (1) CN112767497A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114708164A (en) * 2022-04-08 2022-07-05 四川焱飞科技有限公司 Method for correcting image large and small head distortion caused by object inclination in machine vision measurement
CN115222825A (en) * 2022-09-15 2022-10-21 湖南视比特机器人有限公司 Calibration method, computer storage medium and calibration system
CN116563388A (en) * 2023-04-28 2023-08-08 北京优酷科技有限公司 Calibration data acquisition method and device, electronic equipment and storage medium

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114708164A (en) * 2022-04-08 2022-07-05 四川焱飞科技有限公司 Method for correcting image large and small head distortion caused by object inclination in machine vision measurement
CN115222825A (en) * 2022-09-15 2022-10-21 湖南视比特机器人有限公司 Calibration method, computer storage medium and calibration system
CN116563388A (en) * 2023-04-28 2023-08-08 北京优酷科技有限公司 Calibration data acquisition method and device, electronic equipment and storage medium
CN116563388B (en) * 2023-04-28 2024-05-07 神力视界(深圳)文化科技有限公司 Calibration data acquisition method and device, electronic equipment and storage medium

Similar Documents

Publication Publication Date Title
CN111243032B (en) Full-automatic detection method for checkerboard corner points
CN112767497A (en) High-robustness calibration device based on circular calibration plate and positioning method
CN109859226B (en) Detection method of checkerboard corner sub-pixels for graph segmentation
CN109035276B (en) Image edge extraction method and device and automatic driving system
CN111260731B (en) Self-adaptive detection method for checkerboard sub-pixel level corner points
CN106683137B (en) Artificial mark based monocular and multiobjective identification and positioning method
CN107941808B (en) 3D printing forming quality detection system and method based on machine vision
CN108562250B (en) Keyboard keycap flatness rapid measurement method and device based on structured light imaging
CN112132907B (en) Camera calibration method and device, electronic equipment and storage medium
CN107767456A (en) A kind of object dimensional method for reconstructing based on RGB D cameras
CN114494045B (en) Large spur gear geometric parameter measurement system and method based on machine vision
CN113012234B (en) High-precision camera calibration method based on plane transformation
CN114998571B (en) Image processing and color detection method based on fixed-size markers
CN106952262B (en) Ship plate machining precision analysis method based on stereoscopic vision
CN113313116A (en) Vision-based accurate detection and positioning method for underwater artificial target
CN115082538A (en) System and method for three-dimensional reconstruction of surface of multi-view vision balance ring part based on line structure light projection
CN109671084B (en) Method for measuring shape of workpiece
CN114332758A (en) Method for panoramic reconstruction based on multiple visual angles under high-shielding multi-target environment
JP5254545B2 (en) Image processing apparatus and image processing method
CN111667429B (en) Target positioning correction method for inspection robot
CN115239801B (en) Object positioning method and device
CN115184362B (en) Rapid defect detection method based on structured light projection
CN108596981B (en) Aerial view angle re-projection method and device of image and portable terminal
CN116596987A (en) Workpiece three-dimensional size high-precision measurement method based on binocular vision
CN214670725U (en) High robustness calibration device based on circular calibration plate

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