CN103868460B - Binocular stereo vision method for automatic measurement based on parallax optimized algorithm - Google Patents
Binocular stereo vision method for automatic measurement based on parallax optimized algorithm Download PDFInfo
- Publication number
- CN103868460B CN103868460B CN201410094119.XA CN201410094119A CN103868460B CN 103868460 B CN103868460 B CN 103868460B CN 201410094119 A CN201410094119 A CN 201410094119A CN 103868460 B CN103868460 B CN 103868460B
- Authority
- CN
- China
- Prior art keywords
- parallax
- binocular
- pixel
- target object
- coordinate
- 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.)
- Expired - Fee Related
Links
Abstract
The invention discloses the binocular stereo vision method for automatic measurement based on parallax optimized algorithm, 1) the binocular view after being corrected;2) utilize Stereo Matching Algorithm to mate for base figure with left view, obtain preliminary disparity map;3) to the left view after correction, target object region is colour original, and other nontarget areas are all black;4) according to target object region, it is thus achieved that the complete disparity map in target object region;5) to complete disparity map, three-dimensional point cloud is obtained according to projection model;6) to three-dimensional point cloud, enter row-coordinate re-projection, synthesize coordinate associated pixel figure;7) morphologic method is utilized, it is achieved automatically measure the length and width of target object.This invention simplifies binocular measurement procedure;Decrease mirror-reflection, foreshortening, perspective distortion, the low texture of smooth surface and repeat texture effects;Achieve automation and intelligentification measurement, extend the range of application of binocular measurement, provide technical support for follow-up Robot Binocular Vision.
Description
Technical field
The present invention relates to the fields of measurement of binocular stereo vision, specifically automatically survey based on the binocular stereo vision of parallax optimized algorithm
Metering method.
Background technology
In recent years, the fields such as robot self-aiming, unmanned, hidden military surveillance, medical diagnosis and industrial detection should
With increasingly extensively, binocular stereo vision is increasingly becoming in Robotics research and enlivens one of branch the most, is also that subsequent technology is sent out
The important foundation of exhibition.It is a kind of efficient untouchable advanced detection, measurement based on the measurement technology of binocular stereo vision, identifies
Technology, development prospect is considerable, has wide range of applications.
Currently, the application of Binocular vision photogrammetry is largely focused on measurement distance, has certain limitation;And at measurement space thing
The size dimension application aspect of body, needs point take target pixel points or arrange special characteristic point, it is impossible to realize the automatic of object
Extract, automatically measure, so hinder industrial intellectuality, mechanization;And the artificial behaviour in traditional measuring method
Make step, not only increase labour cost and make step more complicated;The qualified detection of part realizing in terms of industrial detection,
Also it is the detection under fixed scene, it is impossible to realize the automatic measurement of relative motion dimension of object size;And the object to motion,
It is very inaccurate that manual point takes object pixel, can not realize real-time.The more important thing is, due to the minute surface of smooth surface
The impact on binocular ranging such as reflection, foreshortening, perspective distortion, low texture and repetition texture, causes manually putting the target taking
Pixel not necessarily effective three-dimensional information.
Content of the invention
It is an object of the invention to, for overcoming the deficiencies in the prior art, provide a kind of binocular stereo vision based on parallax optimized algorithm certainly
Dynamic measuring method, this method not only simplify the operating procedure of binocular measurement, more accurately can automatically obtain target object
Metrical information, and be optimized the parallax producing after its coupling, reduces the mirror-reflection of smooth surface, foreshortening, thoroughly
Depending on ectocines such as distortion, low texture and repetition textures, the three-dimensional point cloud information making target object is more complete, it is achieved automation
Intelligent measuring, extends the range of application of binocular measurement greatly, provides technical support for follow-up Robot Binocular Vision.
The technical scheme realizing the object of the invention is:
Based on the binocular stereo vision method for automatic measurement of parallax optimized algorithm, comprise the steps:
1) obtained two-way RGB color image by binocular camera simultaneously, utilize chessboard calibration method to carry out binocular camera camera
Binocular calibration, it is thus achieved that the inside and outside parameter of two binocular camera cameras, carries out binocular correction according to these parameters to binocular view,
Remove fisheye effect and site error impact, the binocular view after being corrected of camera lens;
2) to the binocular view after correction, utilize Stereo Matching Algorithm to mate for base figure with left view, obtain preliminary parallax
Figure;
3) to the left view after correction, utilizing background subtracting method to extract target object, target object region is colour original, other
Nontarget area is all black;
4) according to target object region, utilize based on foreground detection parallax optimized algorithm to preliminary disparity map local no parallax region
It is filled with, it is thus achieved that the complete disparity map in target object region;
5) to complete disparity map, three-dimensional point cloud is obtained according to projection model;
6) to three-dimensional point cloud, enter row-coordinate re-projection, three-dimensional coordinate is projected to two dimensional surface, and by the method for image procossing,
The world coordinates of two dimensional surface is associated with the pixel coordinate of picture, with gray scale picture concrete demonstrate two after re-projection
Dimension coordinate generates coordinate associated pixel figure;
7) to coordinate associated pixel figure, utilize morphologic method, use minimum enclosed rectangle to frame the profile in pixel picture,
According to the relation of world coordinates and picture pixels, the pixel value of its length of side is converted to real length information, and automatically demonstrates
Come, it is achieved automatically measure the length and width of target object.
Described binocular camera is that the double-channel collection video combined by two eyes of the camera simulation people of two same specifications sets
Standby, obtain picture pixels and be set to 640x480 pixel, keep original RGB picture.
Described chessboard calibration method is the calibration algorithm utilizing specific black and white gridiron pattern to realize, demarcates binocular camera, obtains
Obtain the inside and outside parameter of two cameras, such as: the parameters such as focal length, distortion parameter, translation vector, spin matrix;Use chessboard mark
The focal length determining two cameras during method keeps consistent, and chessboard uses the black and white lattice of 9x6 lattice, unit square length of side 30mm, gathers picture
15-20 opens;Binocular correction zoom factor is set to 0;Image after correction zooms in and out and translates so that correction chart picture only shows
Effect pixel i.e. removes irregular corner areas.
Described binocular correction refers to that binocular camera calibration not only to draw the inner parameter of each camera, in addition it is also necessary to by mark
The fixed relative position measured between two cameras, i.e. right camera is relative to the vectorial t of the D translation of left camera and rotation
Matrix R;The parallax that impact point to be calculated is formed on left and right two views, first has to this point on left and right view two
Corresponding Pixel matching gets up, but it is very time-consuming for mating corresponding points on two-dimensional space, in order to reduce coupling hunting zone,
We utilize epipolar-line constraint to make the coupling of corresponding points be reduced to linear search by two-dimensional search;The effect of binocular correction seeks to handle and disappears
Except the two width images strictly row correspondence after distortion so that two width images to polar curve just in the same horizontal line, such width
On image, any point and its corresponding points on another piece image just necessarily have identical line number, only need to carry out one-dimensional at this row
Search can match corresponding points.
Described Stereo Matching Algorithm is BM (Boyer-Moore) algorithm (local matching algorithm), or SGBM algorithm (half
Global registration algorithm), BM algorithm state parameter is respectively as follows: the cutoff value (preFilterCap)=31 of pre-processing filter;SAD
Window size (SADWinSiz)=19;Minimum parallax (minDisp)=0;Parallax window (numberOfDisparities)=64;
The judgment threshold (textureThreshold)=10 of low texture region, if the x of all neighbor pixel points leads in current SAD window
Number absolute value sum is less than specifying threshold value, then the parallax value of the corresponding pixel of this window is 0;Parallax uniqueness percentage
(uniquenessRatio)=25, when in parallax window ranges, lowest costs is secondary low-cost (1+uniquenessRatio/100) times,
The corresponding parallax value of lowest costs is only the parallax of this pixel, and otherwise the parallax of this pixel is 0;Parallax change threshold
(speckleRange)=32, when in window, parallax change is more than threshold value, the parallax in this window resets.
The state parameter of SGBM algorithm from unlike BM algorithm: SAD window size=7, permissible range is [1,11];Phase
Penalty coefficient P1 during adjacent pixel parallax add drop 1;Penalty coefficient when neighbor pixel parallax changing value is more than 1 is P2,
P2 is more than P1;FullDP is set to FALSE.
Described background subtracting method, is frame difference method, and i.e. under simple background, the difference such as foreground and background brightness and color is brighter
Aobvious, deduct that frame only comprising background by a later frame comprising prospect, it is thus achieved that complete foreground area and background frames can be at any time
Choosing, the target object extracting is colored original image, and other nontarget areas are all black.
Described extraction target object is to extract automatically to scratch the form of figure, and the foreground target object extracting is original RGB figure
Picture, other background parts are completely black, i.e. the coloured image of a display foreground object.
Described parallax optimized algorithm is in foreground area, detects effective parallax and invalid parallax, and record accessing by row,
Effective parallax is the known conditions of least square method, carries out fitting a straight line to invalid parallax part, and effective parallax keeps constant,
Reaching whole target object region has complete disparity map, and isolates the disparity map in this region.
Described coordinate associated pixel figure is the method utilizing space projection, will with the left optical center of binocular camera as initial point, eyeglass
Plane is that the three-dimensional coordinate in the world coordinate system of X/Y plane projects to X/Y plane, it is achieved the three-dimensional conversion to two dimension;Then
With the corresponding 5cm of a pixel in binaryzation picture, coordinate can be believed by this respective value 5cm according to the scope modification of measurement
Breath shows with the form of binaryzation pixel so that it is visualization.
The above-mentioned binocular stereo vision method for automatic measurement based on parallax optimized algorithm, is applied in the coordinate associated pixel figure generating, from
The data that dynamic measurement obtains, as the mobile tracking of target object shows.
This advantage based on the binocular stereo vision method for automatic measurement of parallax optimized algorithm is:
The steps such as Utilization prospects detection, spatial alternation, Morphological scale-space realize automatically obtaining exactly the metrical information of target object;
Simplify binocular measurement procedure, expanded its application prospect, and the parallax producing after its coupling is optimized, subtracted
Lack mirror-reflection, foreshortening, perspective distortion, the low texture of smooth surface and repeated the ectocines such as texture, having made object
The three-dimensional point cloud information of body is more complete, it is achieved that automation and intelligentification is measured, and extends the application of binocular measurement greatly
Scope, provides technical support for follow-up Robot Binocular Vision.
Brief description
Fig. 1 is binocular stereo imaging principle schematic in embodiment.
Detailed description of the invention
With embodiment, present invention is further elaborated below in conjunction with the accompanying drawings, but is not limitation of the invention.
Embodiment:
Based on the binocular stereo vision method for automatic measurement of parallax optimized algorithm, comprise the following steps:
1) obtained two-way RGB color image by binocular camera simultaneously, utilize chessboard calibration method to carry out binocular camera camera
Binocular calibration, it is thus achieved that the inside and outside parameter of two binocular camera cameras, carries out binocular correction according to these parameters to binocular view,
Remove fisheye effect and site error impact, the binocular view after being corrected of camera lens;
Described binocular camera is that the double-channel collection video combined by two eyes of the camera simulation people of two same specifications sets
Standby, obtain picture pixels and be set to 640x480 pixel, keep original RGB picture;
After adjusting binocular camera and collecting two-way RGB image, utilizing black and white lattice chessboard to demarcate it, camera is due to optics
The characteristic of lens makes imaging there is radial distortion:
xp=x (1+k1r2+k2r4+k3r6)
yp=y (1+k1r2+k2r4+k3r6) (1)
Parameter k in formula (1)1,k2,k3For radial distortion parameter, (x y) is original coordinates, (xp,yp) it is the new coordinate after correcting;
R is radius;
Error in terms of due to assembling, is not substantially parallel between sensor and optical lens, and therefore imaging exists tangential distortion:
xp=x+ [2p1y+p2(r2+2x2)]
yp=y+ [p1(r2+2y2)+2p2x)] (2)
Parameter p in formula (2)1,p2For tangential distortion parameter;The calibration of single camera mainly calculates the interior of camera
Ginseng, i.e. focal length f, imaging initial point cx、cyAnd five distortion parameters, and join outward, i.e. demarcate the world coordinates of thing.Binocular
Camera calibration not only to draw the inner parameter of each camera, in addition it is also necessary to by demarcating the phase measuring between two cameras
To position, i.e. right camera is relative to the D translation vector T of left camera and spin matrix R;Impact point to be calculated left,
The parallax being formed on right two views, first has to this point two corresponding Pixel matchings on left and right view.But,
It is very time-consuming for mating corresponding points on two-dimensional space, and in order to reduce coupling hunting zone, we can utilize epipolar-line constraint to make
The coupling obtaining corresponding points is reduced to linear search by two-dimensional search;And the effect of binocular correction seeks to two width figures after elimination is distorted
As strictly row correspondence so that two width images to polar curve just in the same horizontal line, on such piece image any point with
Its corresponding points on another piece image just necessarily have identical line number, only need to this row carry out linear search can match right
Ying Dian;
Then binocular correction is according to the monocular internal reference data (focal length, imaging initial point, distortion factor) obtaining after camera calibration
With binocular relative position relation (spin matrix and translation vector), carry out to left and right view respectively eliminating distortion and row alignment, make
The imaging origin of left and right view is consistent, two camera optical axis imaging planes parallel, left and right are coplanar, to polar curve row pair
Together;
2) to the binocular view after correction, utilize Stereo Matching Algorithm to mate for base figure with left view, obtain preliminary parallax
Figure, this disparity map is due to mirror-reflection, foreshortening, perspective distortion, low texture and repeats the impact generation sudden change parallaxes such as texture
Or local no parallax;
Stereo Matching Algorithm includes following step: first carry out Matching power flow calculating,
C (p, d)=C (x, y, d)=max{0, IL(pLi)-I'max,I'min-IL(pLi)} (3)
In equation (3), C is Matching power flow function, and d is parallax, pLiRepresent the ith pixel point p of Zuo TuLi(x,y).Left view
In figure, the gray-scale intensity of i-th is expressed as IL(pLi), I'maxAnd I'minRepresenting minimax intensity level respectively, L represents left.
Then Matching power flow superposition:
In formula (4), E is heat-supplied function, and D is whole disparity map, DPAnd DqRepresent that pixel p and q is corresponding respectively to regard
Difference, NPFor the field of pixel p, right side Section 1 represents the Matching power flow sum of all pixels.For pixel p with
Its adjacent pixel depth difference has small change and two kinds of situations of large change, and Section 2 and Section 3 use FACTOR P respectively1And P2
Being punished, function T [] is that 1 and if only if that its parameter is true here, is otherwise 0.Substantially, Section 2 and Section 3
Being smoothness constraint mentioned above, they require that the depth value of adjacent pixel is consistent as much as possible, i.e. keep smooth, aobvious
So, P1< P2;
Then direct original match cost is processed, to heat-supplied function (4), try to achieve energy by different optimized algorithms
Minimum of a value, simultaneously each point parallax value also just calculate.Some measures are used to enter parallax after initial parallax obtains
Row refinement.
3) to the left view after correction, utilizing background subtracting method to extract target object, target object region is colour original, other
Nontarget area is all black;
After left view correction, utilizing frame difference method to extract target object, its principle is:
| i (t)-i (t-1) | < T background
| i (t)-i (t-1) | >=T prospect
Wherein i (t), i (t-1) are respectively t, and the pixel value of t-1 moment corresponding pixel points, T is threshold value.Frame difference method is in background subtracting method
One, only frame difference method does not needs modeling, because its background model is exactly the figure of previous frame, so speed is very fast,
Other frame difference method is not very sensitive to the illumination of slow conversion, and its shortcoming is quite a few certainly, easily occurs that " slur " and " empty " is existing
As so next needing to be effectively improved slur and cavitation by morphology largest contours method, finally with flooding modulus method (mask)
Pluck out target object.
4) according to target object region, utilize based on foreground detection parallax optimized algorithm to preliminary disparity map local no parallax region
It is filled with, it is thus achieved that the complete disparity map in target object region;
Specifically, by row scanning binaryzation foreground picture, it is all that record continuously uninterrupted white pixel is counted more than n (threshold value)
The pixel coordinate of continuity point, obtains coordinate set V.This part is to process foreground picture, obtains the pixel coordinate of area-of-interest
Collection V, wherein V simply records the coordinate value of each pixel of area-of-interest, due to left view, disparity map and foreground picture
In area-of-interest be the same, so V is also the same.Followed by the V above obtaining, preliminary disparity map is carried out
Optimization process.Again by row scanning disparity map, obtain the pixel coordinate with a line (ordinate is identical) in pixel coordinate collection V corresponding
Parallax value, detect whether as virtual value, utilize least square line matching by row, parallax optimization is carried out to foreground area,
Obtain the complete disparity map in target object region.
5) to complete disparity map, three-dimensional point cloud is obtained according to projection model.Three-dimensional point cloud is to utilize binocular stereo imaging principle, i.e.
Projection model (relation between world coordinate system, image coordinate system and camera coordinate system) combines disparity map and generates with binocular phase
Machine left optical center point is the three dimensional space coordinate of initial point, it is thus achieved that three-dimensional point cloud comprise the tables such as three-dimensional coordinate a little, characteristic information
Show the much information between midpoint, space.
As it is shown in figure 1, left camera optical axis is parallel with right camera optical axis, baseline distance B be the projection centre line of two video cameras away from
From;Camera focus is f.
The image of existing two video cameras is in approximately the same plane, then the image coordinate Y coordinate of characteristic point P is identical, i.e. Yletf=Yright=Y
Then obtained by triangle geometrical relationship:
Then parallax is: Disparity=Xleft-Xright, i.e. the difference of the ordinate of the ordinate of left view and right view.
Thus can calculate three-dimensional coordinate under camera coordinates system for characteristic point P is:
Therefore,
On view, institute is a little simply by the presence of corresponding match point, by above-mentioned computing, just can obtain its corresponding three-dimensional coordinate.
6) to three-dimensional point cloud, enter row-coordinate re-projection, three-dimensional coordinate is projected to two dimensional surface, and by the method for image procossing,
The world coordinates of two dimensional surface is associated with the pixel coordinate of picture, with gray scale picture concrete demonstrate two after re-projection
Dimension coordinate generates coordinate associated pixel figure, and its pixel wide, length will directly reflect the actual width of object and length.
7) to coordinate associated pixel figure, utilize morphologic method, use minimum enclosed rectangle to frame the profile in pixel picture,
According to the relation of world coordinates and picture pixels, the pixel value of its length of side is converted to real length information, and automatically demonstrates
Come, it is achieved automatically measure the length and width of target object.
Described chessboard calibration method is the calibration algorithm utilizing specific black and white gridiron pattern to realize, demarcates binocular camera, it is thus achieved that two
The inside and outside parameter of individual camera, such as: the parameters such as focal length, distortion parameter, translation vector, spin matrix;Use chessboard calibration method
When two cameras focal length at same plane, chessboard uses the black and white lattice of 9x6 lattice, unit square length of side 30mm, gathers picture 15-20
?;Binocular correction zoom factor is set to 0;Image after correction zooms in and out and translates so that correction chart picture only shows effective picture
Element i.e. removes irregular corner areas.
Described binocular correction refers to that binocular camera calibration not only to draw the inner parameter of each camera, in addition it is also necessary to by demarcating
Measuring the relative position between two cameras, i.e. right camera is relative to the vectorial t of the D translation of left camera and spin matrix
R;The parallax that impact point to be calculated is formed on the view of two, left and right, first have to this point on left and right view two corresponding
Pixel matching gets up, but it is very time-consuming for mating corresponding points on two-dimensional space, in order to reduce coupling hunting zone, we
Epipolar-line constraint is utilized to make the coupling of corresponding points be reduced to linear search by two-dimensional search;The effect of binocular correction seeks to eliminating abnormal
The strictly row correspondence of two width images after change so that two width images to polar curve just in the same horizontal line, such piece image
Upper any point and its corresponding points on another piece image just necessarily have identical line number, only need to carry out linear search at this row
Corresponding points can be matched.
Described Stereo Matching Algorithm is BM algorithm or SGBM algorithm, and BM algorithm state parameter is respectively as follows: pretreatment filtering
The cutoff value (preFilterCap)=31 of device;SAD window size (SADWinSiz)=19;Minimum parallax (minDisp)
=0;Parallax window (numberOfDisparities)=64;The judgment threshold (textureThreshold)=10 of low texture region, if
In current SAD window, the x derivative absolute value sum of all neighbor pixel points is less than appointment threshold value, the then corresponding pixel of this window
Parallax value be 0;Parallax uniqueness percentage (uniquenessRatio)=25, in parallax window ranges, lowest costs is time low generation
During (1+uniquenessRatio/100) times of valency, the corresponding parallax value of lowest costs is only the parallax of this pixel, otherwise this pixel
The parallax of point is 0;Parallax change threshold (speckleRange)=32, when in window, parallax change is more than threshold value, in this window
Parallax reset;
The state parameter of SGBM algorithm from unlike BM algorithm: SAD window size=7, permissible range is [1,11];Phase
Penalty coefficient P1 during adjacent pixel parallax add drop 1;Penalty coefficient when neighbor pixel parallax changing value is more than 1 is P2,
P2 is more than P1;FullDP is set to FALSE.
Described background subtracting method, is frame difference method, and i.e. under simple background, the difference such as foreground and background brightness and color is obvious,
Deduct that frame only comprising background by a later frame comprising prospect, it is thus achieved that complete foreground area and background frames can be chosen at any time,
The target object extracting is colored original image, and other nontarget areas are all black.
Described extraction target object is to extract automatically to scratch the form of figure, and the foreground target object extracting is original RGB figure
Picture, other background parts are completely black, i.e. the coloured image of a display foreground object.
Described parallax optimized algorithm is in foreground area, detects effective parallax and invalid parallax, and record accessing by row,
Effective parallax is the known conditions of least square method, carries out fitting a straight line to invalid parallax part, and effective parallax keeps constant,
Reaching whole target object region has complete disparity map, and isolates the disparity map in this region.
Described coordinate associated pixel figure is the method utilizing space projection, will with the left optical center of binocular camera as initial point, eyeglass
Plane is that the three-dimensional coordinate in the world coordinate system of X/Y plane projects to X/Y plane, it is achieved the three-dimensional conversion to two dimension;Then
With the corresponding 5cm of a pixel in binaryzation picture, coordinate can be believed by this respective value 5cm according to the scope modification of measurement
Breath shows with the form of binaryzation pixel so that it is visualization.
This binocular stereo vision method for automatic measurement based on parallax optimized algorithm, is applied in the coordinate associated pixel figure generating, from
The data that dynamic measurement obtains, as the mobile tracking of target object shows.
The above-mentioned binocular stereo vision method for automatic measurement based on parallax optimized algorithm, only uses or binocular camera quilt first
Adjust out-of-date just needs to demarcate, as long as the binocular parameter during binocular ranging is arranged once, need not measure every time and all arrange,
And background frames can be changed according to application scenarios at any time;When target object enters view, will be identified, and pluck out mesh
Mark object is at corresponding binaryzation view interface meeting tracing detection object and completes automatically to measure, the metrical information of display target object.
Claims (2)
1. the binocular stereo vision method for automatic measurement based on parallax optimized algorithm, is characterized in that, comprise the steps:
1) obtained two-way RGB color image by binocular camera simultaneously, chessboard calibration method is utilized to carry out binocular calibration to binocular camera camera, obtain the inside and outside parameter of two binocular camera cameras, according to these parameters, binocular correction is carried out to binocular view, remove fisheye effect and site error impact, the binocular view after being corrected of camera lens;
2) to the binocular view after correction, utilize Stereo Matching Algorithm to mate for base figure with left view, obtain preliminary disparity map;
3) to the left view after correction, utilizing background subtracting method to extract target object, target object region is colour original, and other nontarget areas are all black;
4) according to target object region, the parallax optimized algorithm based on foreground detection is utilized to be filled with no parallax region, preliminary disparity map local, it is thus achieved that the complete disparity map in target object region;
5) to complete disparity map, three-dimensional point cloud is obtained according to projection model;
6) to three-dimensional point cloud, enter row-coordinate re-projection, three-dimensional coordinate is projected to two dimensional surface, and by the method for image procossing, the world coordinates of two dimensional surface is associated with the pixel coordinate of picture, and with concrete the demonstrating of gray scale picture, the two-dimensional coordinate after re-projection generates coordinate associated pixel figure;
7) to coordinate associated pixel figure, utilize morphologic method, minimum enclosed rectangle is used to frame the profile in pixel picture, relation according to world coordinates and picture pixels, the pixel value of its length of side is converted to real length information, and automatically show, it is achieved automatically measure the length and width of target object;
Binocular camera obtains picture pixels and is set to 640x480 pixel, keeps original RGB picture;
When using chessboard calibration method, the focal length of two cameras keeps consistent, and chessboard uses the black and white lattice of 9x6 lattice, unit square length of side 30mm, gathers picture 15-20 and opens;Binocular correction zoom factor is set to 0;
Described Stereo Matching Algorithm is BM algorithm or SGBM algorithm, and BM algorithm state parameter is respectively as follows: cutoff value=31 of pre-processing filter;SAD window size=19;Minimum parallax=0;Parallax window=64;Judgment threshold=10 of low texture region, if the x derivative absolute value sum of all neighbor pixel points is less than specifying threshold value in current SAD window, then the parallax value of the corresponding pixel of this window is 0;Parallax uniqueness percentage=25, when in parallax window ranges, lowest costs is secondary low-cost (1+uniquenessRatio/100) times, the corresponding parallax value of lowest costs is only the parallax of this pixel, and otherwise the parallax of this pixel is 0;Parallax change threshold=32, when in window, parallax change is more than threshold value, the parallax in this window resets;
The state parameter of SGBM algorithm from unlike BM algorithm: SAD window size=7, permissible range is [1,11];Penalty coefficient P1 during neighbor pixel parallax add drop 1;Penalty coefficient when neighbor pixel parallax changing value is more than 1 is P2, and P2 is more than P1;FullDP is set to FALSE;
Described parallax optimized algorithm is in foreground area, detect effective parallax and invalid parallax by row, and record accessing, effective parallax is the known conditions of least square method, fitting a straight line is carried out to invalid parallax part, and effectively parallax keeps constant, reaching whole target object region has complete disparity map, and isolates the disparity map in this region.
2. the binocular stereo vision method for automatic measurement based on parallax optimized algorithm according to claim 1, it is characterized in that, described coordinate associated pixel figure is the method utilizing space projection, three-dimensional coordinate with the left optical center of binocular camera as initial point, in the world coordinate system as X/Y plane for the lens plane is projected to X/Y plane, it is achieved the three-dimensional conversion to two dimension;Then with the corresponding 5cm of a pixel in binaryzation picture, coordinate information can be showed with the form of binaryzation pixel so that it is visualization by this respective value 5cm according to the scope modification of measurement.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410094119.XA CN103868460B (en) | 2014-03-13 | 2014-03-13 | Binocular stereo vision method for automatic measurement based on parallax optimized algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410094119.XA CN103868460B (en) | 2014-03-13 | 2014-03-13 | Binocular stereo vision method for automatic measurement based on parallax optimized algorithm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN103868460A CN103868460A (en) | 2014-06-18 |
CN103868460B true CN103868460B (en) | 2016-10-05 |
Family
ID=50907248
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410094119.XA Expired - Fee Related CN103868460B (en) | 2014-03-13 | 2014-03-13 | Binocular stereo vision method for automatic measurement based on parallax optimized algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN103868460B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109993781A (en) * | 2019-03-28 | 2019-07-09 | 北京清微智能科技有限公司 | Based on the matched anaglyph generation method of binocular stereo vision and system |
CN111121722A (en) * | 2019-12-13 | 2020-05-08 | 南京理工大学 | Binocular three-dimensional imaging method combining laser dot matrix and polarization vision |
Families Citing this family (95)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105335955B (en) * | 2014-07-17 | 2018-04-10 | 株式会社理光 | Method for checking object and object test equipment |
CN104111039B (en) * | 2014-08-08 | 2016-08-24 | 电子科技大学 | For arbitrarily putting the scaling method of fringe projection three-dimension measuring system |
CN104236468A (en) * | 2014-09-26 | 2014-12-24 | 上海未来伙伴机器人有限公司 | Method and system for calculating coordinates of target space and mobile robot |
CN104363375B (en) * | 2014-11-28 | 2018-03-27 | 广东欧珀移动通信有限公司 | A kind of method, apparatus and terminal of photo noise reduction |
CN104406538B (en) * | 2014-12-01 | 2017-01-11 | 南京航空航天大学 | Three-dimensional mark point matching method for point cloud registration and three-dimensional scanning method |
CN104636724B (en) * | 2015-02-02 | 2018-09-04 | 华中科技大学 | A kind of quick Pedestrians and vehicles detection method of in-vehicle camera based on goal congruence |
CN104794713B (en) * | 2015-04-15 | 2017-07-11 | 同济大学 | Chamber crop digitalized image method based on ARM and binocular vision |
CN105551020B (en) * | 2015-12-03 | 2019-03-26 | 浙江大华技术股份有限公司 | A kind of method and device detecting object size |
WO2017032335A1 (en) | 2015-08-26 | 2017-03-02 | Zhejiang Dahua Technology Co., Ltd. | Methods and systems for traffic monitoring |
CN105046749A (en) * | 2015-09-10 | 2015-11-11 | 深圳市神州龙资讯服务有限公司 | Method for automatically generating 3D model based on three-view aerial photos |
CN105651258B (en) * | 2015-12-30 | 2018-07-13 | 杨正林 | Active visual angle binocular range-measurement system and method |
CN105550670B (en) * | 2016-01-27 | 2019-07-12 | 兰州理工大学 | A kind of target object dynamically track and measurement and positioning method |
CN105783731A (en) * | 2016-03-08 | 2016-07-20 | 上海易景信息科技有限公司 | Method for measuring length of measured object by means of double cameras |
CN106504276B (en) * | 2016-10-25 | 2019-02-19 | 桂林电子科技大学 | Non local solid matching method |
CN106651911A (en) * | 2016-11-17 | 2017-05-10 | 天津大学 | Fish behavior acquiring method based on image tracking |
CN107636679B (en) * | 2016-12-30 | 2021-05-25 | 达闼机器人有限公司 | Obstacle detection method and device |
CN106625713A (en) * | 2017-01-11 | 2017-05-10 | 长春工业大学 | Method of improving gumming accuracy of gumming industrial robot |
CN106952332B (en) * | 2017-03-06 | 2020-05-05 | 华侨大学 | Automatic 3D picture construction drawing generation method based on perspective transformation |
CN106803913A (en) * | 2017-03-10 | 2017-06-06 | 武汉东信同邦信息技术有限公司 | A kind of detection method and its device of the action that taken the floor for Auto-Sensing student |
CN108961383B (en) * | 2017-05-19 | 2021-12-14 | 杭州海康威视数字技术股份有限公司 | Three-dimensional reconstruction method and device |
CN107256562A (en) * | 2017-05-25 | 2017-10-17 | 山东师范大学 | Image defogging method and device based on binocular vision system |
CN107274400B (en) * | 2017-06-21 | 2021-02-12 | 歌尔光学科技有限公司 | Space positioning device, positioning processing method and device, and virtual reality system |
CN109215044B (en) * | 2017-06-30 | 2020-12-15 | 京东方科技集团股份有限公司 | Image processing method and system, storage medium, and mobile system |
CN107392898B (en) * | 2017-07-20 | 2020-03-20 | 海信集团有限公司 | Method and device for calculating pixel point parallax value applied to binocular stereo vision |
CN107578397A (en) * | 2017-07-25 | 2018-01-12 | 西南交通大学 | A kind of novel non-contact abrasion of contact wire detection method |
CN108230235B (en) * | 2017-07-28 | 2021-07-02 | 北京市商汤科技开发有限公司 | Disparity map generation system, method and storage medium |
CN107481267A (en) * | 2017-08-14 | 2017-12-15 | 华南理工大学 | A kind of shooting projection interactive system and method based on binocular vision |
CN107588721A (en) * | 2017-08-28 | 2018-01-16 | 武汉科技大学 | The measuring method and system of a kind of more sizes of part based on binocular vision |
CN108401460B (en) * | 2017-09-29 | 2021-10-12 | 深圳市大疆创新科技有限公司 | Method, system, storage medium, and computer program product for generating a disparity map |
CN108010085B (en) * | 2017-11-30 | 2019-12-31 | 西南科技大学 | Target identification method based on binocular visible light camera and thermal infrared camera |
CN108230372B (en) * | 2017-11-30 | 2022-01-14 | 努比亚技术有限公司 | Image processing method, terminal and computer readable storage medium |
CN107945220B (en) * | 2017-11-30 | 2020-07-10 | 华中科技大学 | Binocular vision-based reconstruction method |
CN107862721B (en) * | 2017-12-01 | 2020-06-16 | 杭州晶一智能科技有限公司 | Grassland secondary recognition system based on binocular stereoscopic vision |
CN107992871B (en) * | 2017-12-21 | 2020-06-02 | 陕西伟景机器人科技有限公司 | Restaurant automatic checkout method based on image recognition |
CN108171739A (en) * | 2017-12-25 | 2018-06-15 | 清华大学苏州汽车研究院(吴江) | Binocular vision solid matching method based on movable information |
CN108458668A (en) * | 2018-01-05 | 2018-08-28 | 燕山大学 | Slab edge and Head and Tail Shape automatic checkout system based on binocular vision and method |
CN108259881A (en) * | 2018-01-30 | 2018-07-06 | 深圳市得色科技有限公司 | 3D synthetic methods and its system based on parallax estimation |
CN108335350A (en) * | 2018-02-06 | 2018-07-27 | 聊城大学 | The three-dimensional rebuilding method of binocular stereo vision |
CN108921842A (en) * | 2018-07-02 | 2018-11-30 | 上海交通大学 | A kind of cereal flow detection method and device |
CN109389648B (en) * | 2018-09-19 | 2022-04-22 | 晓智未来(成都)科技有限公司 | Method for reducing measurement data error through data iteration |
CN109186947B (en) * | 2018-09-26 | 2019-11-12 | 福州大学 | A kind of electric appliance Three-Dimensional Dynamic characteristic Auto-Test System and method |
CN111047638B (en) * | 2018-10-12 | 2023-08-18 | 深圳富联富桂精密工业有限公司 | Monocular distance measuring device |
CN109048926A (en) * | 2018-10-24 | 2018-12-21 | 河北工业大学 | A kind of intelligent robot obstacle avoidance system and method based on stereoscopic vision |
CN109584312B (en) * | 2018-11-30 | 2020-09-11 | Oppo广东移动通信有限公司 | Camera calibration method, device, electronic equipment and computer-readable storage medium |
CN111380458A (en) * | 2018-12-29 | 2020-07-07 | 顺丰科技有限公司 | Method and device for measuring carriage cargo volume |
CN109612437B (en) * | 2019-01-14 | 2020-07-31 | 华南农业大学 | Binocular positioning detection method based on combination of Bumblebe and total station |
CN111627067B (en) * | 2019-02-28 | 2023-08-22 | 海信集团有限公司 | Calibration method of binocular camera and vehicle-mounted equipment |
CN109903327B (en) * | 2019-03-04 | 2021-08-31 | 西安电子科技大学 | Target size measurement method of sparse point cloud |
JP7078173B2 (en) * | 2019-03-14 | 2022-05-31 | オムロン株式会社 | Image processing device and 3D measurement system |
CN111721194A (en) * | 2019-03-19 | 2020-09-29 | 北京伟景智能科技有限公司 | Multi-laser-line rapid detection method |
CN109974659A (en) * | 2019-03-31 | 2019-07-05 | 徐州工程学院 | A kind of embedded range-measurement system based on binocular machine vision |
CN110120074B (en) * | 2019-05-10 | 2020-08-25 | 清研同创机器人(天津)有限公司 | Cable positioning method for live working robot in complex environment |
CN110260801A (en) * | 2019-05-13 | 2019-09-20 | 平安科技(深圳)有限公司 | Method and apparatus for measuring volume of material |
CN110675454A (en) * | 2019-10-16 | 2020-01-10 | 深圳前海达闼云端智能科技有限公司 | Object positioning method, device and storage medium |
CN110838148B (en) * | 2019-11-11 | 2023-05-05 | 杭州凌像科技有限公司 | Binocular camera self-calibration method |
CN110889873A (en) * | 2019-11-26 | 2020-03-17 | 中国科学院光电研究院 | Target positioning method and device, electronic equipment and storage medium |
CN111047709B (en) * | 2019-11-29 | 2023-05-05 | 暨南大学 | Binocular vision naked eye 3D image generation method |
CN111198563B (en) * | 2019-12-30 | 2022-07-29 | 广东省智能制造研究所 | Terrain identification method and system for dynamic motion of foot type robot |
CN111243018A (en) * | 2019-12-31 | 2020-06-05 | 北京航空航天大学 | Method and system for automatically performing nevus-removing operation |
CN111351446B (en) * | 2020-01-10 | 2021-09-21 | 奕目(上海)科技有限公司 | Light field camera calibration method for three-dimensional topography measurement |
CN111402309B (en) * | 2020-03-05 | 2023-11-03 | 中国科学院上海微系统与信息技术研究所 | Calibration and depth image processing method for multi-view stereo camera |
CN111369447B (en) * | 2020-03-09 | 2023-06-16 | 湖南警察学院 | Method for correcting image point in monocular stereoscopic vision image |
CN113393413B (en) * | 2020-03-11 | 2022-09-13 | 河海大学 | Water area measuring method and system based on monocular and binocular vision cooperation |
CN111429523B (en) * | 2020-03-16 | 2021-06-15 | 天目爱视(北京)科技有限公司 | Remote calibration method in 3D modeling |
CN111462096A (en) * | 2020-04-03 | 2020-07-28 | 浙江商汤科技开发有限公司 | Three-dimensional target detection method and device |
CN111508068B (en) * | 2020-04-20 | 2023-05-30 | 华中科技大学 | Three-dimensional reconstruction method and system applied to binocular endoscopic image |
CN111553252B (en) * | 2020-04-24 | 2022-06-07 | 福建农林大学 | Road pedestrian automatic identification and positioning method based on deep learning and U-V parallax algorithm |
CN111739359A (en) * | 2020-06-30 | 2020-10-02 | 上海乂学教育科技有限公司 | Augmented reality courseware generation system |
CN112085802A (en) * | 2020-07-24 | 2020-12-15 | 浙江工业大学 | Method for acquiring three-dimensional finger vein image based on binocular camera |
CN111899282A (en) * | 2020-07-30 | 2020-11-06 | 平安科技(深圳)有限公司 | Pedestrian trajectory tracking method and device based on binocular camera calibration |
CN112097732A (en) * | 2020-08-04 | 2020-12-18 | 北京中科慧眼科技有限公司 | Binocular camera-based three-dimensional distance measurement method, system, equipment and readable storage medium |
CN112053781B (en) * | 2020-09-16 | 2023-04-28 | 四川大学华西医院 | Dynamic and static stereoscopic vision testing method and terminal |
CN112171635A (en) * | 2020-09-24 | 2021-01-05 | 上海思寒环保科技有限公司 | Remote control robot |
CN112013792B (en) * | 2020-10-19 | 2021-02-02 | 南京知谱光电科技有限公司 | Surface scanning three-dimensional reconstruction method for complex large-component robot |
CN112258635B (en) * | 2020-10-26 | 2023-07-21 | 北京石油化工学院 | Three-dimensional reconstruction method and device based on improved binocular matching SAD algorithm |
CN112379605B (en) * | 2020-11-24 | 2023-03-28 | 中国人民解放军火箭军工程大学 | Bridge crane semi-physical simulation control experiment system and method based on visual servo |
CN112465899A (en) * | 2020-11-25 | 2021-03-09 | 国网福建省电力有限公司电力科学研究院 | Binocular vision-based automatic wire peeling and positioning method |
CN112734652B (en) * | 2020-12-22 | 2023-03-31 | 同济大学 | Near-infrared blood vessel image projection correction method based on binocular vision |
CN112581542A (en) * | 2020-12-24 | 2021-03-30 | 北京百度网讯科技有限公司 | Method, device and equipment for evaluating automatic driving monocular calibration algorithm |
CN112767442B (en) * | 2021-01-18 | 2023-07-21 | 中山大学 | Pedestrian three-dimensional detection tracking method and system based on top view angle |
CN113177565B (en) * | 2021-03-15 | 2023-03-24 | 北京理工大学 | Binocular vision position measuring system and method based on deep learning |
CN112991369B (en) * | 2021-03-25 | 2023-11-17 | 湖北工业大学 | Method for detecting outline size of running vehicle based on binocular vision |
CN114004880B (en) * | 2021-04-08 | 2023-04-25 | 四川大学华西医院 | Point cloud and strong reflection target real-time positioning method of binocular camera |
CN113099204B (en) * | 2021-04-13 | 2022-12-13 | 北京航空航天大学青岛研究院 | Remote live-action augmented reality method based on VR head-mounted display equipment |
CN113077524B (en) * | 2021-04-22 | 2023-12-08 | 中德(珠海)人工智能研究院有限公司 | Automatic calibration method, device and equipment for binocular fisheye camera and storage medium |
CN113446958A (en) * | 2021-06-02 | 2021-09-28 | 南京航空航天大学 | Binocular vision system based on symmetrical arrangement and curved surface reconstruction method of DIC technology |
CN113340241B (en) * | 2021-06-09 | 2022-12-02 | 河南德朗智能科技有限公司 | Binocular vision concrete joint surface roughness measurement method and system |
CN113379831B (en) * | 2021-06-22 | 2022-09-09 | 北京航空航天大学青岛研究院 | Augmented reality method based on binocular camera and humanoid robot |
CN113487679B (en) * | 2021-06-29 | 2023-01-03 | 哈尔滨工程大学 | Visual ranging signal processing method for automatic focusing system of laser marking machine |
CN113610741A (en) * | 2021-08-18 | 2021-11-05 | 梅卡曼德(北京)机器人科技有限公司 | Point cloud processing method and device based on laser line scanning |
CN113963052B (en) * | 2021-09-22 | 2023-08-18 | 西安交通大学 | Large aerostat volume real-time monitoring method based on binocular vision |
CN113959398B (en) * | 2021-10-09 | 2024-04-12 | 广东汇天航空航天科技有限公司 | Distance measurement method and device based on vision, drivable equipment and storage medium |
CN114742876B (en) * | 2022-06-13 | 2022-09-06 | 菏泽市土地储备中心 | Land vision stereo measurement method |
CN115049980A (en) * | 2022-06-16 | 2022-09-13 | 威海经济技术开发区天智创新技术研究院 | Target object determination method and device based on image and electronic equipment |
CN116258759B (en) * | 2023-05-15 | 2023-09-22 | 北京爱芯科技有限公司 | Stereo matching method, device and equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009124308A (en) * | 2007-11-13 | 2009-06-04 | Tokyo Denki Univ | Multi-viewpoint image creating system and multi-viewpoint image creating method |
CN101887589A (en) * | 2010-06-13 | 2010-11-17 | 东南大学 | Stereoscopic vision-based real low-texture image reconstruction method |
CN101976455A (en) * | 2010-10-08 | 2011-02-16 | 东南大学 | Color image three-dimensional reconstruction method based on three-dimensional matching |
-
2014
- 2014-03-13 CN CN201410094119.XA patent/CN103868460B/en not_active Expired - Fee Related
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2009124308A (en) * | 2007-11-13 | 2009-06-04 | Tokyo Denki Univ | Multi-viewpoint image creating system and multi-viewpoint image creating method |
CN101887589A (en) * | 2010-06-13 | 2010-11-17 | 东南大学 | Stereoscopic vision-based real low-texture image reconstruction method |
CN101976455A (en) * | 2010-10-08 | 2011-02-16 | 东南大学 | Color image three-dimensional reconstruction method based on three-dimensional matching |
Non-Patent Citations (3)
Title |
---|
基于双目立体视觉的弹目空间交会测量系统;许凤良;《中国优秀硕士学位论文全文数据库工程科技II辑》;20091115(第11期);22-25、34-40、44、51-52、58、63-66 * |
基于视差和帧差的图割优化运动目标分割算法;朱秋煜等;《电视技术》;20120702;第36卷(第13期);135-138 * |
大尺寸高温锻件双目视觉测量技术;吴瑞敏;《中国优秀硕士学位论文全文数据库信息科技辑》;20090515(第5期);61-66 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109993781A (en) * | 2019-03-28 | 2019-07-09 | 北京清微智能科技有限公司 | Based on the matched anaglyph generation method of binocular stereo vision and system |
CN111121722A (en) * | 2019-12-13 | 2020-05-08 | 南京理工大学 | Binocular three-dimensional imaging method combining laser dot matrix and polarization vision |
Also Published As
Publication number | Publication date |
---|---|
CN103868460A (en) | 2014-06-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103868460B (en) | Binocular stereo vision method for automatic measurement based on parallax optimized algorithm | |
CN107945268B (en) | A kind of high-precision three-dimensional method for reconstructing and system based on binary area-structure light | |
CN104330074B (en) | Intelligent surveying and mapping platform and realizing method thereof | |
CN103411553B (en) | The quick calibrating method of multi-linear structured light vision sensors | |
CN102999892B (en) | Based on the depth image of region mask and the intelligent method for fusing of RGB image | |
CN107635129B (en) | Three-dimensional trinocular camera device and depth fusion method | |
CN103607584B (en) | Real-time registration method for depth maps shot by kinect and video shot by color camera | |
CN110264567A (en) | A kind of real-time three-dimensional modeling method based on mark point | |
CN101329771B (en) | Method for rapidly modeling of urban street base on image sequence | |
CN102509348B (en) | Method for showing actual object in shared enhanced actual scene in multi-azimuth way | |
CN105550670A (en) | Target object dynamic tracking and measurement positioning method | |
CN104376552A (en) | Virtual-real registering algorithm of 3D model and two-dimensional image | |
CN104299220B (en) | A kind of method that cavity in Kinect depth image carries out real-time filling | |
CN105258673B (en) | A kind of target ranging method based on binocular synthetic aperture focusing image, device | |
CN107560592A (en) | A kind of precision ranging method for optronic tracker linkage target | |
CN103971404A (en) | 3D real-scene copying device having high cost performance | |
CN109712232B (en) | Object surface contour three-dimensional imaging method based on light field | |
CN105277144A (en) | Land area rapid detection method based on binocular vision and detection device thereof | |
CN108629828B (en) | Scene rendering transition method in the moving process of three-dimensional large scene | |
CN109035307B (en) | Set area target tracking method and system based on natural light binocular vision | |
CN106295657A (en) | A kind of method extracting human height's feature during video data structure | |
CN109801360A (en) | Stomach and intestine three-dimensionalreconstruction and method for visualizing based on image | |
CN108230242A (en) | A kind of conversion method from panorama laser point cloud to video flowing | |
CN110909571B (en) | High-precision face recognition space positioning method | |
CN106709432A (en) | Binocular stereoscopic vision based head detecting and counting method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20161005 |