CN110288699A - A kind of three-dimensional rebuilding method based on structure light - Google Patents
A kind of three-dimensional rebuilding method based on structure light Download PDFInfo
- Publication number
- CN110288699A CN110288699A CN201910559303.XA CN201910559303A CN110288699A CN 110288699 A CN110288699 A CN 110288699A CN 201910559303 A CN201910559303 A CN 201910559303A CN 110288699 A CN110288699 A CN 110288699A
- Authority
- CN
- China
- Prior art keywords
- frequency
- point
- phase
- image
- testee
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 56
- 238000013507 mapping Methods 0.000 claims abstract description 38
- 238000003384 imaging method Methods 0.000 claims abstract description 13
- 230000010363 phase shift Effects 0.000 claims description 7
- 230000001351 cycling effect Effects 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000000149 penetrating effect Effects 0.000 claims 1
- 238000001028 reflection method Methods 0.000 abstract description 7
- 230000008901 benefit Effects 0.000 description 3
- 241000282693 Cercopithecidae Species 0.000 description 2
- 239000000919 ceramic Substances 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000010146 3D printing Methods 0.000 description 1
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 230000003190 augmentative effect Effects 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000007906 compression Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000004907 flux Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 239000000155 melt Substances 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000002366 time-of-flight method Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Classifications
-
- G06T3/14—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformation in the plane of the image
- G06T3/40—Scaling the whole image or part thereof
- G06T3/4038—Scaling the whole image or part thereof for image mosaicing, i.e. plane images composed of plane sub-images
-
- 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
- G06T7/85—Stereo camera calibration
Abstract
A kind of three-dimensional rebuilding method based on structure light, first setting Binocular Stereo Vision System, demarcate two cameras in system.Then on the rotating platform by testee setting, demarcate rotating platform, the point cloud model in testee different rotary face is acquired by multiple rotary, the method for wherein obtaining single point cloud model is as follows: projecting high-frequency structure light respectively, if architectures light, single-frequency structure light is to testee surface, to the high frequency imaging of two cameras acquisition, medium frequency image, single-frequency image is decoded to obtain phase information, high frequency imaging is mapped to medium frequency image using three frequency reflection methods, medium frequency image is mapped to single-frequency image again, the point cloud model that homotopy mapping obtains the current surfaces of revolution is carried out to the single-frequency image of two cameras after mapping.Finally the point cloud model of each surfaces of revolution is rotated under the same coordinate system, all point cloud models are registrated using iterative closest point algorithm, the threedimensional model for obtaining testee is merged after registration.
Description
Technical field
The invention belongs to the stereoscopic vision fields in computer vision, are related to a kind of three-dimensional rebuilding method based on structure light
It is had important application in fields such as 3D printing, augmented reality, recognition of face and reverse-engineerings.
Background technique
Computer vision, that is, computer is obtained the description to objective world and is believed by the processing to picture or sequence of pictures
Breath helps people to better understand the content for including in picture.And three-dimensional reconstruction, so that computer vision is more prone to people
The mode of class vision has shown relief visual information.
Currently, the three-dimensional rebuilding method of view-based access control model can be divided into passive means and active method, the former include close-range photogrammetry,
Stereoscopic vision etc..Wherein stereo vision method be similar to human eye perception scene depth mode, using scene under two visual angles institute
The position of imaging is different, carries out Stereo matching, and in conjunction with photography geometrical principle, restoration scenario depth is current most popular three-dimensional
Method for reconstructing.But Stereo matching often utilizes the intrinsic informations such as color similarity, is restricted reconstruction precision, and can not weigh
Build weak texture or texture-free object.Proactive includes time-of-flight method (Time of Flight, ToF), photometric stereo
Mensuration, Structure light method etc..The time difference that wherein ToF directly utilizes transmitter transmitting light wave and receiver to receive light wave calculates object
The depth information of body, principle is simple, such as the Microsoft Kinect V2 released in 2013, can be quickly obtained the depth letter of scene around
Breath, but resolution ratio is lower;Photometric stereo mensuration mainly utilizes the positional relationship of light source and camera, uses different combination of light sources
The different image of multiframe luminous flux is obtained, object surface shape is restored with this, to the more demanding of environment light;Structure light method is benefit
With spatial encoding techniques, coded structured light is projected to body surface is rebuild, is decoded after cameras capture, in conjunction with photography geometry
Basic principle, can quickly and accurately reconstruct high accuracy three-dimensional model.
Structured light technique, for improving the matching precision of same place, on the basis of obtaining high-precision homonymy matching point, benefit
With photography Set Theory, three-dimensional reconstruction is carried out.Traditional structure light coding mainly has a Gray code and phase code, but Gray code and
Phase code has its inherent shortcoming.For Gray code, since black code and white code can have a degree of phase in intersection
The fringe resolution of Gray code, can not be arranged to a pixel wide, therefore subsequent homotopy mapping will appear by mutually interference
One-to-many situation.For phase code, single-frequency phase code, which can guarantee each column pixel all, has different phase code values, but
Since the phase value difference between adjacent pixel is too small, it is easy the interference by noise.And multifrequency phase is due to the week of phase value
Phase property, can generate period ambiguity problem, still cannot achieve one-to-one homotopy mapping.
Summary of the invention
Single-frequency phase code is held in the defect and structure light being individually present for above-mentioned traditional various three-dimensional rebuilding methods
The problem of cannot achieve one-to-one homotopy mapping vulnerable to noise jamming, multifrequency phase code, the present invention propose a kind of based on knot
The three-dimensional rebuilding method of structure light has been built a set of binocular vision system being made of two cameras and a projector, and has been designed
The a set of phase code coding and decoding scheme that can obtain high-precision homotopy mapping, is first reflected high-frequency phase using three frequency reflection methods
It penetrates as intermediate frequency phase, then by the intermediate frequency phase mapping after mapping is single-frequency phase, can be improved precision;It is acquired using rotating platform
The point cloud model of different angle, and a cloud rough registration is carried out as ICP (Iterative Closest using rotating platform
Point) the initialization of iterative closest point algorithm, the shortcomings that avoiding ICP algorithm from falling into local optimum, it is final obtain it is complete three-dimensional
Model.
The technical solution of the present invention is as follows:
A kind of three-dimensional rebuilding method based on structure light, includes the following steps:
Step 1, setting Binocular Stereo Vision System, the Binocular Stereo Vision System includes two cameras in left and right, to two
A camera carries out the inside and outside parameter that calibration obtains each camera;
Testee is arranged on the rotating platform step 2, demarcates rotating platform, acquires the quilt by multiple rotary
The point cloud model in object different rotary face is surveyed, the point cloud model of the current surfaces of revolution of the testee is wherein acquired after single rotation
Method it is as follows:
2.1, using projector by high-frequency structure pumped FIR laser pattern projection to the testee surface, and by left and right two
Camera carries out Image Acquisition to the testee respectively, and the image of collected the right and left is decoded to obtain left and right two
The phase information of each pixel in the high frequency imaging on side;
2.2, using projector by if architectures pumped FIR laser pattern projection to the testee surface, and by left and right two
Camera carries out Image Acquisition to the testee respectively, and the image of collected the right and left is decoded to obtain left and right two
The phase information of each pixel in the medium frequency image on side;
2.3, using projector by single-frequency structure light coding pattern projection to the testee surface, and by left and right two
Camera carries out Image Acquisition to the testee respectively, and the image of collected the right and left is decoded to obtain left and right two
The phase information of each pixel in the single-frequency image on side;
2.4, the phase information of pixel each in the high frequency imaging of the right and left is respectively mapped in the right and left
Frequency image, then the phase information of pixel each in the medium frequency image of the right and left after mapping is respectively mapped to the right and left
Single-frequency image;
2.5, homotopy mapping is carried out to the single-frequency image of the right and left obtained after mapping, obtains the testee and exists
The point cloud model of the current surfaces of revolution;
Step 3, using the rotating platform demarcated, the point cloud module in the different rotary face of the testee is carried out inverse
Rotation, the testee is slightly spliced in the case where the point cloud model of each surfaces of revolution rotates to the same coordinate system, will slightly be spelled
Initial value of the binding fruit as iterative closest point algorithm accurately matches all point cloud models using iterative closest point algorithm
Quasi- is merged to obtain the threedimensional model of the testee after registration.
Specifically, being all made of N step phase shift method in the step 2.1, step 2.2, step 2.3, wherein projected in step 2.1
Instrument is by N width phase shift high-frequency structure pumped FIR laser pattern projection to the testee surface, wherein in the n-th width phase shifted images (x, y)
The light intensity of pointAre as follows:
Wherein Ap(x, y) and Bp(x, y) respectively indicates the amplitude offset and the amplitude index of modulation of phase shifted images, and f is structure
The frequency of pumped FIR laser, n ∈ [0, N-1], N are the positive integer not less than 3;
N-th width phase shifted images acquired by camera after image in (x, y) point light intensityAre as follows:
Wherein AC(x, y) and BC(x, y) respectively indicates the amplitude offset and the amplitude index of modulation of acquisition image, under
Formula is decoded the phase (x, y) of (x, y) point in acquisition image:
。
Specifically, by the phase mapping of the pixel of (x, y) coordinate in high frequency imaging to intermediate frequency figure in the step 2.4
As and by the phase mapping of the pixel of (x, y) coordinate in medium frequency image to the method for single-frequency image are as follows:
Wherein k (x, y) indicates that the phase cycling of the pixel of (x, y) coordinate, round are indicated to required numerical value four house five
Enter, fhIndicate higher-frequency image and the frequency multiple compared with low-frequency image, φu(x, y) is indicated compared with (x, y) coordinate in low-frequency image
The solution code phase of pixel, φh(x, y) indicates the solution code phase of the pixel of (x, y) coordinate in higher-frequency image, φ (x, y)
The phase of the pixel of (x, y) coordinate in image after indicating mapping.
Specifically, carrying out homotopy mapping to the single-frequency image of the right and left obtained after mapping in the step 2.5
Method are as follows: set in left figure a pixel as pl, record plAll pixels point of the point on the correspondence polar curve of right figure, from record
It is found in pixel and plThe immediate two points p of phase of point-And p+, p-The phase of point is φ-, p+Point phase value be
φ+, plThe corresponding match point p of pointrAbscissa beMatch point prOrdinate beWherein a, b, c are the p of the inside and outside parameter acquisition of the camera obtained according to step 1lThe corresponding polar curve of point
Equation In coefficient,For plThe abscissa of point,For plThe ordinate of point.
Specifically, it is as follows that the point cloud model in different rotary face rotated to the method under the same coordinate system in the step 3:
A surfaces of revolution is set as the plane of reference, remaining each surfaces of revolution is rotated into the plane of reference, if any need to rotate
The surfaces of revolution on a little be C (XC,YC,ZC), corresponding points of this on the plane of reference are
Wherein θ is the angle that the surfaces of revolution needs to rotate to the plane of reference, S (Xs,Ys,Zs) it is rotating platform center point coordinate, N
=(Nx,Ny,Nz) it is rotating platform normal direction, R (N, θ) is spin matrix,
。
Specifically, being filtered fusion using voxel filter method in the step 3.
The invention has the benefit that the present invention proposes three frequency reflection methods, the phenomenon that avoiding mapping error, and retain high
The high-precision characteristic of frequency can be realized high-precision point Yun Chongjian;Rotating platform is set, the point cloud model of each surfaces of revolution is acquired
And they are spliced under the same coordinate system, it uses spliced cloud as the initialization of ICP essence registration Algorithm, can avoid ICP
Algorithm because initial position it is undesirable caused by be registrated failure, substantially increase the success rate of ICP registration, and pass through rotating platform
After rough registration, ICP algorithm requires to reduce to the registration of cloud, can reduce the times of collection of a cloud, save a large amount of acquisition
Time and calculating time.
Detailed description of the invention
Fig. 1 is a kind of flow chart of three-dimensional rebuilding method based on structure light proposed by the present invention.
Fig. 2 is Binocular Stereo Vision System structure chart.
Fig. 3 is rotating platform schematic diagram.
Fig. 4 is to utilize a kind of Doraemon Three-dimension Reconstruction Model of the three-dimensional rebuilding method based on structure light proposed by the present invention
Figure.
Fig. 5 utilizes a kind of ceramic monkey Three-dimension Reconstruction Model of the three-dimensional rebuilding method based on structure light proposed by the present invention
Figure.
Specific embodiment
The technical schemes of the invention are described in detail in the following with reference to the drawings and specific embodiments.
The present invention proposes a kind of three-dimensional rebuilding method based on structure light, if Fig. 1 is specific flow chart, including walks as follows
It is rapid:
Binocular Stereo Vision System is arranged in step 1 first, as shown in Fig. 2, Binocular Stereo Vision System includes left and right two
Camera, projector can be placed among two cameras, carry out the inside and outside parameter that calibration obtains each camera, this reality to two cameras
It applies camera calibration in example and uses Zhang Zhengyou gridiron pattern standardization, nominal data includes the internal reference of camera, He Liangtai camera
Relatively outer ginseng.
Testee is arranged on the rotating platform step 2, demarcates rotating platform, acquires the quilt by multiple rotary
Survey the point cloud model in object different rotary face.
The present invention proposes that three frequency reflection methods handle the phase of acquisition, needs to acquire high-frequency structure light, if architectures
Image after light, single-frequency project structured light to testee.The image of acquisition obtains the phase of each pixel after decoding
The phase information of pixel each in the high frequency imaging of acquisition is mapped in medium frequency image by information, then by the intermediate frequency after mapping
The phase information of each pixel is mapped in single-frequency image in image.Three frequency phase mapping methods are the bases in double frequency phase mapping
It is proposed on plinth, double frequency phase mapping method is when frequency is excessively high, and through overcompression, adjacent periods are compared with single-frequency phase, gap
It is too small, along with the modulation of object surface shape and the superposition of noise, the period raw certain influence of fixed output quota really can be caused
Mapping error, thus reconstruction failure.Three frequency reflection methods are proposed in the present invention, and high frequency is mapped to the intermediate frequency of a transition first
On, the phenomenon that re-mapping in single-frequency, can avoid mapping error, and the high-precision characteristic of reserved high-frequency, realize high-precision point
Cloud rebuild, solved high frequency map directly on unit frequency will appear mapping error caused by rebuild hole the problems such as.Three
Frequency after frequency mapping, linear characteristic is more prominent, reinforces the anti-interference ability of noise.Three frequency reflection methods not only remain
The high-precision characteristic of high frequency, and remain the integrality that intermediate frequency maps directly to single-frequency.
Image Acquisition is carried out with N step phase shift method in the present embodiment, is illustrated for projecting high-frequency structure light, due to phase
Machine is horizontal positioned, a bit in left image, corresponding polar curve slope is smaller in right image, each on polar curve close to level
Point is all by with different abscissas, so phase code is designed to that vertical striped, same row pixel have in the present embodiment
The pixel of identical phase value, different lines has different phase values.Likewise, phase code is set if the vertical f of camera is placed
It is calculated as horizontal stripe.Projector is by N width phase shift high-frequency structure pumped FIR laser pattern projection to testee surface, the coding of phase code
Formula is as shown in formula 1:
WhereinIndicate light intensity of the n-th width phase shifted images of projection in (x, y) point, Ap(x, y) and Bp(x, y) difference
It indicates amplitude offset and the amplitude index of modulation, is constant constant, does not change with the change of coordinate points (x, y).Usual feelings
It is 0 to 255 this condition to meet the tonal range of image, the present embodiment takes A under conditionp(x, y)=127.5, Bp(x, y)=
127.5, phase code is carried out to image abscissa x so being just degenerated to the phase code of entire image.N is not less than 3
Positive integer, indicate phase shift number, that is, the number for the coding pattern for needing to project, n=0,1 ..., N-1.F indicates phase code
Frequency can be understood as the lane issue for including in whole picture coded image herein.By taking experimental system of the invention as an example, throw
The size of shadow picture is 1024 × 768, and when projection single-frequency phase code, i.e., when lane issue is 1, f=1/1024 works as projection
High-frequency phase coding, when lane issue is k, f=k/1024.
After camera acquires, the picture of acquisition can be expressed as the form of formula 2:
WhereinIt indicates under the projection of n-th phase code pattern, the picture of camera shooting is in the position (x, y)
Light intensity, AC(x, y) and BC(x, y) is not constant at this time due to the influence of interchannel noise.φ (x, y) is to want decoded phase
Place value.Wherein there is AC(x, y), BCThree unknown numbers of (x, y) and φ (x, y), at least need three projection images that can just solve.
Corresponding coding/decoding method is designed according to encoded picture, the decoding formula of phase code is as shown in formula 3:
According to the above method obtain high-frequency structure light, if architectures light, single-frequency structure light acquisition image in each pixel
After the phase of point, high frequency is mapped to intermediate frequency by three frequency reflection methods, intermediate frequency is mapped to single-frequency, wherein the specific mistake of frequency mapping
Journey is divided into the compression of higher-frequency and determines the period and be mapped to compared with two steps of low frequency, respectively as shown in formula 4 and formula 5:
Wherein k (x, y) indicates the phase cycling of the pixel of (x, y) coordinate, and the range of phase code is returned from [0,2 π]
One turns to [0,1], and round expression rounds up to required numerical value.fhIndicate higher-frequency image and the frequency times compared with low-frequency image
Number, if it is the ratio between high frequency and intermediate frequency that high frequency, which is mapped to intermediate frequency, if it is intermediate frequency itself that intermediate frequency, which is mapped to single-frequency,.φu(x,
Y) the solution code phase compared with the pixel of (x, y) coordinate in low-frequency image, φ are indicatedh(x, y) indicates that (x, y) is sat in higher-frequency image
The solution code phase of target pixel, the phase code value that can not strictly meet each column at this time is identical, if high frequency maps
To intermediate frequency, φu(x, y) is just the solution code phase of medium frequency image, φh(x, y) is just the solution code phase of high frequency imaging;If intermediate frequency
It is mapped to single-frequency, φu(x, y) is just the solution code phase of single-frequency image, φh(x, y) is just the solution code phase of medium frequency image.φ(x,
Y) phase of the pixel of (x, y) coordinate in the image after indicating mapping.
Homotopy mapping then is carried out to each pixel after three frequencies map, testee is obtained and is rotated currently
The point cloud model in face.Photography geometrical principle is based on binocular stereo vision, and as shown in Fig. 2, the point P in three-dimensional space is logical
Physics of photography is crossed in the imaging plane Π of left and right cameral,ΠrUpper imaging is respectively pl,pr, pl,prReferred to as a pair of of the same name
Point.Line segment OlOrIt is the baseline (baseline) in binocular stereo vision, two of them endpoint Ol,OrIt is two video cameras respectively
Optical center.Ol,OrWith Πl,ΠrIntersection location be respectively el,er, it is referred to as pole.Connect plAnd elStraight line LlAnd connection
prAnd erStraight line LrReferred to as polar curve.Any point P and O in three-dimensional scenicl,OrThe face constituted is referred to as polar plane.
The matching process of same place is converted to the matching of phase value.It is believed that phase value is immediate in two images
Two pixels are a pair of of same place.In the present invention, the linear characteristic of phase code is utilized, sub- picture is realized by straight line fitting
The homotopy mapping of plain grade.Specific method is described as follows: some pixel p in left figurelIt is decoded with the phase code that size is φ
As a result.First according to epipolar-line constraint, records this and correspond to the pixel that polar curve is passed through in right image;Then phase code is searched
Closest to two point p of φ-And p+, p-Phase value be φ-, p+Phase value be φ+, can be with according to the distribution mode of phase
Obtain p-And p+It must be two adjacent pixels, utilize φ-And φ+Size, fit a virtual subpixel coordinates
Point is used as plMatch point pr, match point prCoordinate it is as shown in formula 6.
Homotopy mapping is carried out according to the above method, using photography geometrical principle, realizes the reconstruction of three-dimensional point cloud, is obtained every
The point cloud model of one surfaces of revolution.
Step 3, by testee in the case where the point cloud model of each surfaces of revolution rotates to the same coordinate system, changed using closest approach
All point cloud models are registrated for algorithm ICP, the threedimensional model for obtaining testee is merged after registration.
The rotating platform that the present invention uses is a circular platform, and rotary shaft is located at the center point, as shown in Fig. 3, wherein
World coordinate system Ow-XwYwZwFor current reference coordinates system, S (Xs,Ys,Zs) it is rotating platform center point coordinate, N=(Nx,Ny,
Nz) it is rotating platform normal direction, and N is the rotary shaft of rotating platform by central point S, i.e. N.Assuming that initial position object
Surface has a bitAfter rotating platform rotates θ counterclockwise, which is C (XC,YC,ZC).By rotating θ
The process of the C0 point position before the rotation of C point position acquisition afterwards is exactly point cloud.In method proposed by the present invention, one is devised
Set is directly from the method for the point cloud information of rotating platform calibration rotating platform.Rotation is reconstructed using above-mentioned method of structured light first
Turn the surface of platform, plane fitting then is carried out to the surface, the normal direction of plane is determined, finally to the point cloud of rotating platform
Position is averaged, and the coordinate of Platform center point is obtained.It, can be by different angle after obtaining center point coordinate and normal direction
Point cloud rotates under the same coordinate system, as shown in formula 7:
Wherein R (N, θ) is a spin matrix, only with rotating platform normal vector N and to rotate angle, θ related, such as formula (8) and
(9) shown in.
Rotation angle, θ can be manually set and record, according to object surface shape, it is ensured that have between consecutive points cloud certain
Registration.Rotating platform normal vector direction N=(Nx,Ny,Nz) and center position S (XS,YS,ZS) can be obtained by calibration
?.
Three-dimensional rebuilding method proposed by the present invention is adopted in data acquisition phase in order to obtain the point cloud data of different angle
With rotating platform.It is reconstructed object to place on the rotating platform, every specific angle of rotation carries out the acquisition of a data.
In the point cloud registering stage, rotating platform is demarcated first, the point cloud of different angle is subjected to a reverse rotation operation, splicing
To under the same coordinate system, use spliced cloud as the initialization of ICP essence registration Algorithm, ICP algorithm is sensitive to initial position,
After slightly being spliced using rotating platform, ICP registration success rate and registration accuracy can be improved, ICP algorithm is avoided to pay no attention to because of initial position
It is registrated failure caused by thinking, substantially increases the success rate of ICP registration.And after passing through rotating platform rough registration, ICP algorithm
The registration of cloud is required to reduce, the times of collection of a cloud can be reduced, save a large amount of acquisition time and calculates the time.
It after multi-disc point cloud registering, is merely superimposed and will cause a cloud Density inhomogeneity, point cloud is later and information is superfluous
It is remaining, need to obtain the threedimensional model of testee by fusion, the present embodiment uses voxel filter method, is filtered and melts to a cloud
It closes, the consistent complete three-dimensional model of even density can be obtained.It is to utilize three-dimensional proposed by the present invention as shown in figures 4 and 5
The threedimensional model that method for reconstructing handles Doraemon and ceramic monkey.
Method provided by the present invention is described in detail above, specific case pair is applied in this illustration method
The principle of the present invention and embodiment are expounded, and build mode such as camera and projector, the calibration of rotating platform is special
Property etc., the core concept described above for being served only for helping to understand the method for the present invention, those of ordinary skill in the art, according to this hair
Bright thought can be changed in specific implementation.Therefore, the content of the present specification should not be construed limitation of the present invention.
Claims (6)
1. a kind of three-dimensional rebuilding method based on structure light, which comprises the steps of:
Step 1, setting Binocular Stereo Vision System, the Binocular Stereo Vision System includes two cameras in left and right, to two phases
Machine carries out the inside and outside parameter that calibration obtains each camera;
Testee is arranged on the rotating platform step 2, demarcates rotating platform, acquires the measured object by multiple rotary
The point cloud model in body different rotary face wherein acquires the side of the point cloud model of the current surfaces of revolution of the testee after single rotation
Method is as follows:
2.1, using projector by high-frequency structure pumped FIR laser pattern projection to the testee surface, and by left and right two cameras
Image Acquisition is carried out to the testee respectively, is decoded the image of collected the right and left to obtain the right and left
The phase information of each pixel in high frequency imaging;
2.2, using projector by if architectures pumped FIR laser pattern projection to the testee surface, and by left and right two cameras
Image Acquisition is carried out to the testee respectively, is decoded the image of collected the right and left to obtain the right and left
The phase information of each pixel in medium frequency image;
2.3, using projector by single-frequency structure light coding pattern projection to the testee surface, and by left and right two cameras
Image Acquisition is carried out to the testee respectively, is decoded the image of collected the right and left to obtain the right and left
The phase information of each pixel in single-frequency image;
2.4, the phase information of pixel each in the high frequency imaging of the right and left is respectively mapped to the intermediate frequency figure of the right and left
Picture, then the phase information of pixel each in the medium frequency image of the right and left after mapping is respectively mapped to the list of the right and left
Frequency image;
2.5, homotopy mapping is carried out to the single-frequency image of the right and left obtained after mapping, obtains the testee current
The point cloud model of the surfaces of revolution;
Step 3, the inverse rotation of point cloud module progress using the rotating platform demarcated, to the different rotary face of the testee
Turn, the testee is slightly spliced in the case where the point cloud model of each surfaces of revolution rotates to the same coordinate system, will slightly be spliced
As a result as the initial value of iterative closest point algorithm, accuracy registration is carried out to all point cloud models using iterative closest point algorithm,
It is merged to obtain the threedimensional model of the testee after registration.
2. the three-dimensional rebuilding method according to claim 1 based on structure light, which is characterized in that the step 2.1, step
2.2, N step phase shift method is all made of in step 2.3, wherein projector throws N width phase shift high-frequency structure pumped FIR laser pattern in step 2.1
It is mapped to the testee surface, wherein the light intensity that (x, y) is put in the n-th width phase shifted imagesAre as follows:
Wherein Ap(x, y) and Bp(x, y) respectively indicates the amplitude offset and the amplitude index of modulation of phase shifted images, and f is structure light volume
The frequency of code, n ∈ [0, N-1], N are the positive integer not less than 3;
N-th width phase shifted images acquired by camera after image in (x, y) point light intensityAre as follows:
Wherein AC(x, y) and BC(x, y) respectively indicates the amplitude offset and the amplitude index of modulation of acquisition image, right according to the following formula
The phase (x, y) of (x, y) point is decoded in acquisition image:
3. the three-dimensional rebuilding method according to claim 1 based on structure light, which is characterized in that will in the step 2.4
The phase mapping of the pixel of (x, y) coordinate is to medium frequency image and by the picture of (x, y) coordinate in medium frequency image in high frequency imaging
Method of the phase mapping of vegetarian refreshments to single-frequency image are as follows:
Wherein k (x, y) indicates the phase cycling of the pixel of (x, y) coordinate, and round expression rounds up to required numerical value, fhTable
Show higher-frequency image and the frequency multiple compared with low-frequency image, φu(x, y) indicates the pixel compared with (x, y) coordinate in low-frequency image
Solution code phase, φh(x, y) indicates the solution code phase of the pixel of (x, y) coordinate in higher-frequency image, and φ (x, y) expression is reflected
The phase of the pixel of (x, y) coordinate in image after penetrating.
4. the three-dimensional rebuilding method according to claim 1 based on structure light, which is characterized in that right in the step 2.5
The method that the single-frequency image of the right and left obtained after mapping carries out homotopy mapping are as follows: set in left figure a pixel as pl,
Record plAll pixels point of the point on the correspondence polar curve of right figure, finds and p from the pixel of recordlThe phase of point most connects
Two close point p-And p+, p-The phase of point is φ-, p+The phase value of point is φ+, plThe corresponding match point p of pointrAbscissa beMatch point prOrdinate beWherein a, b, c are obtained according to step 1
The p that the inside and outside parameter of camera obtainslThe corresponding polar curve equation of pointIn coefficient,For plThe cross of point
Coordinate,For plThe ordinate of point.
5. the three-dimensional rebuilding method according to claim 1 based on structure light, which is characterized in that will not in the step 3
The method rotated under the same coordinate system with the point cloud model of the surfaces of revolution is as follows:
A surfaces of revolution is set as the plane of reference, remaining each surfaces of revolution is rotated into the plane of reference, if any rotation for needing to rotate
Turn on face to be some C (XC,YC,ZC), corresponding points of this on the plane of reference are
Wherein θ is the angle that the surfaces of revolution needs to rotate to the plane of reference, S (Xs,Ys,Zs) it is rotating platform center point coordinate, N=(Nx,
Ny,Nz) it is rotating platform normal direction, R (N, θ) is spin matrix,
。
6. the three-dimensional rebuilding method according to claim 1 based on structure light, which is characterized in that utilized in the step 3
Voxel filter method is filtered fusion.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910559303.XA CN110288699A (en) | 2019-06-26 | 2019-06-26 | A kind of three-dimensional rebuilding method based on structure light |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910559303.XA CN110288699A (en) | 2019-06-26 | 2019-06-26 | A kind of three-dimensional rebuilding method based on structure light |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110288699A true CN110288699A (en) | 2019-09-27 |
Family
ID=68005854
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910559303.XA Pending CN110288699A (en) | 2019-06-26 | 2019-06-26 | A kind of three-dimensional rebuilding method based on structure light |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110288699A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110852979A (en) * | 2019-11-12 | 2020-02-28 | 广东省智能机器人研究院 | Point cloud registration and fusion method based on phase information matching |
CN111028295A (en) * | 2019-10-23 | 2020-04-17 | 武汉纺织大学 | 3D imaging method based on coded structured light and dual purposes |
CN111275776A (en) * | 2020-02-11 | 2020-06-12 | 北京淳中科技股份有限公司 | Projection augmented reality method and device and electronic equipment |
CN111754462A (en) * | 2020-05-28 | 2020-10-09 | 新拓三维技术(深圳)有限公司 | Visual detection method and system for three-dimensional bent pipe |
CN113192179A (en) * | 2021-04-28 | 2021-07-30 | 沈阳工业大学 | Three-dimensional reconstruction method based on binocular stereo vision |
WO2021185218A1 (en) * | 2020-03-16 | 2021-09-23 | 左忠斌 | Method for acquiring 3d coordinates and dimensions of object during movement |
CN113781305A (en) * | 2021-09-08 | 2021-12-10 | 浙江汉振智能技术有限公司 | Point cloud fusion method of double-monocular three-dimensional imaging system |
CN113838112A (en) * | 2021-09-24 | 2021-12-24 | 东莞市诺丽电子科技有限公司 | Trigger signal determining method and trigger signal determining system of image acquisition system |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107170043A (en) * | 2017-06-19 | 2017-09-15 | 电子科技大学 | A kind of three-dimensional rebuilding method |
-
2019
- 2019-06-26 CN CN201910559303.XA patent/CN110288699A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107170043A (en) * | 2017-06-19 | 2017-09-15 | 电子科技大学 | A kind of three-dimensional rebuilding method |
Non-Patent Citations (6)
Title |
---|
CHAO ZUO: "Temporal phase unwrapping algorithms for fringe projection profilometry: A comparative review", < OPTICS AND LASERS IN ENGINEERING> * |
吴奇: "基于结构光与旋转平台的三维重建技术", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
左超: "一种新的点云拼接算法", 《中国激光》 * |
钟莹: "基于改进ICP算法的点云自动配准技术", 《控制工程》 * |
陶海跻: "一种基于法向量的点云自动配准方法", 《中国激光》 * |
高爽: "基于结构光的小型物体三维重建技术", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111028295A (en) * | 2019-10-23 | 2020-04-17 | 武汉纺织大学 | 3D imaging method based on coded structured light and dual purposes |
CN110852979A (en) * | 2019-11-12 | 2020-02-28 | 广东省智能机器人研究院 | Point cloud registration and fusion method based on phase information matching |
CN111275776A (en) * | 2020-02-11 | 2020-06-12 | 北京淳中科技股份有限公司 | Projection augmented reality method and device and electronic equipment |
WO2021185218A1 (en) * | 2020-03-16 | 2021-09-23 | 左忠斌 | Method for acquiring 3d coordinates and dimensions of object during movement |
CN111754462A (en) * | 2020-05-28 | 2020-10-09 | 新拓三维技术(深圳)有限公司 | Visual detection method and system for three-dimensional bent pipe |
CN113192179A (en) * | 2021-04-28 | 2021-07-30 | 沈阳工业大学 | Three-dimensional reconstruction method based on binocular stereo vision |
CN113192179B (en) * | 2021-04-28 | 2024-03-26 | 沈阳工业大学 | Three-dimensional reconstruction method based on binocular stereo vision |
CN113781305A (en) * | 2021-09-08 | 2021-12-10 | 浙江汉振智能技术有限公司 | Point cloud fusion method of double-monocular three-dimensional imaging system |
CN113838112A (en) * | 2021-09-24 | 2021-12-24 | 东莞市诺丽电子科技有限公司 | Trigger signal determining method and trigger signal determining system of image acquisition system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110288699A (en) | A kind of three-dimensional rebuilding method based on structure light | |
CN107170043B (en) | A kind of three-dimensional rebuilding method | |
CN103955920B (en) | Binocular vision obstacle detection method based on three-dimensional point cloud segmentation | |
US7573489B2 (en) | Infilling for 2D to 3D image conversion | |
US7573475B2 (en) | 2D to 3D image conversion | |
US9357206B2 (en) | Systems and methods for alignment, calibration and rendering for an angular slice true-3D display | |
CN103971404B (en) | 3D real-scene copying device having high cost performance | |
CN108475327A (en) | three-dimensional acquisition and rendering | |
JP2017532847A (en) | 3D recording and playback | |
CN101487703B (en) | Fast full-view stereo photography measuring apparatus | |
CN107407866A (en) | Laser radar stereoscopic fusion true man's outdoor scene threedimensional model video reconstruction for 360 ° of body virtual reality videos of six degree of freedom | |
JP2002519791A (en) | Method and apparatus for capturing a stereoscopic image using an image sensor | |
CN105791800B (en) | Three-dimensional display system and stereo display method | |
CN108876926A (en) | Navigation methods and systems, AR/VR client device in a kind of panoramic scene | |
JP2000503177A (en) | Method and apparatus for converting a 2D image into a 3D image | |
CN109379578A (en) | Omnidirectional three-dimensional video-splicing method, apparatus, equipment and storage medium | |
CN107464280A (en) | The matching process and device of user's 3D modeling | |
Moreau et al. | Equisolid fisheye stereovision calibration and point cloud computation | |
AU2022271399A1 (en) | Techniques for multi-view neural object modeling | |
CN105812772B (en) | Medical image three-dimensional display system and method | |
Tahara et al. | Interference-free epipole-centered structured light pattern for mirror-based multi-view active stereo | |
Evers‐Senne et al. | Image based interactive rendering with view dependent geometry | |
Budinger | An analysis of 3-D display strategies | |
CN107079145B (en) | For collecting the method for the image data for generating immersion video and for the method based on image data viewing space | |
CN111899293B (en) | Virtual and real shielding processing method in AR application |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190927 |
|
WD01 | Invention patent application deemed withdrawn after publication |