CN107767440A - Historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint - Google Patents
Historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint Download PDFInfo
- Publication number
- CN107767440A CN107767440A CN201710793380.2A CN201710793380A CN107767440A CN 107767440 A CN107767440 A CN 107767440A CN 201710793380 A CN201710793380 A CN 201710793380A CN 107767440 A CN107767440 A CN 107767440A
- Authority
- CN
- China
- Prior art keywords
- image
- point
- same place
- historical relic
- dimensional
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Computer Graphics (AREA)
- Geometry (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint, including：Step 1: collection historical relic sequential images；Step 2: homotopy mapping is carried out between sequential images；Step 3: the position and attitude parameter of every image of acquisition；Step 4: in stereogram, characteristic point is extracted with Harris grids to left image and is accurately positioned using Fostner operators, the corresponding same place of search matching establishes equally distributed same place on right image, the Di Luoni triangulation networks of the same name are built using above-mentioned same place as seed point, by being matched to the continuous interpolation of triangle core point, historical relic image high density same place is obtained；Extract left image edge feature and obtain marginal point, marginal information same place is matched in stereogram；Step 5: according to the position of image and attitude parameter, historical relic image high density same place is subjected to point off density reconstruction；Step 6: using the three-dimensional point cloud of historical relic as reference, absolute orientation is realized, obtains subtle three-dimensional model.
Description
Technical field
The invention belongs to photogrammetric and laser radar technique field, it is related to a kind of text based on triangulation network interpolation and constraint
Thing sequential images subtle three-dimensional method for reconstructing.
Background technology
The three-dimensional information of real world objective objects more can intuitively, the environment and attribute of effectively expressing measurand, with
Continuous improvement of the three-dimensional reconstruction in speed and precision etc., its method is widely used in mapping, urban planning, doctor
The fields such as, military affairs, cultural heritage digital protection.Surveying and Mapping Industry application three-dimensional reconstruction generation digital terrain model, number
Word city model etc., applied to data display, data renewal, information management etc.；City three in urban planning and development control
Dimension rebuild for spatial analysis demand, show city Markov Model, City Building dynamic monitoring, reduce incorrect decision
It is significant；Military affairs are upper can carry out mapping navigation, structure Virtual Battlefield, provide accurate geography information, establish truly, can measure
The threedimensional model of survey, it is significant in target positioning, terrain visualization etc.；In medical science CT three-dimensional reconstructions promote from
Cross fault makes traditional abstract expression be represented to real visual pattern, the accuracy of diagnosis to the leap of more surface layers or even solid
Increase substantially；In terms of cultural heritage digital protection, three-dimensional reconstruction is digitized important content, its recordable historical relic shape
Achieved with the details of color, carry out disease survey, while instruct the virtual reparation of historical relic, and observer can be with polygonal
Degree, more comprehensively browse.Historical relic department requires more and more higher to digitized at present, and the historical relic three-dimensional reconstruction of efficient fine is
The focus studied at present.
The content of the invention
It is an object of the invention to solve at least the above and/or defect, and provide at least will be described later it is excellent
Point.
It is a still further object of the present invention to provide a kind of based on triangulation network interpolation and the historical relic sequential images of constraint fine three
Tie up method for reconstructing.
Therefore, technical scheme provided by the invention is：
A kind of historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint, including：
Step 1: the sequential images of historical relic are gathered using camera；
Step 2: carrying out homotopy mapping between the sequential images, sequential images orientation same place is obtained；
Step 3: resolving the orientation parameter for obtaining every image first with the same place, determined afterwards with the image
To parameter as initial value, by bundle adjustment, the position of every image of acquisition and attitude parameter；
Step 4: the dense Stereo Matching of sequential images neutral body picture pair：In stereogram, to left image with Harris lattices
Net extraction characteristic point is simultaneously accurately positioned using Fostner operators, and the corresponding same place of search matching is established equal on right image
The same place of even distribution, the Di Luoni triangulation networks of the same name are built using equally distributed same place as seed point, by triangle
The continuous interpolation matching of focus point, is continuously increased the same place of matching, obtains historical relic image high density same place；
Step 5: according to the position of the every image obtained in step 3 and attitude parameter, by the historical relic image
High density same place carries out point off density reconstruction, builds historical relic image fine-point cloud；
Step 6: using the three-dimensional point cloud of the historical relic as reference, realized by the spacial similarity transformation between coordinate system
The absolute orientation of the historical relic image fine-point cloud, make historical relic image fine-point cloud that there is actual size and yardstick, obtain historical relic
The subtle three-dimensional model of sequential images.
Preferably, in the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint,
In the step 3, specifically comprise the following steps：
3.1) same place that will be obtained in step 2, directly solved by relative orientation and obtained with reference to the method tightly solved
The relative position and attitude parameter of initial volumetric picture pair, while its three-dimensional coordinate is resolved to the same place application forward intersection；
3.2) index relative will be established between the two-dimentional same place of initial volumetric picture pair and the three-dimensional point of its forward intersection, and
According to the same place of the two or three image, determine initial volumetric picture to two-dimensional points of the same name between three-dimensional point P and the 3rd image
P, afterwards by the three-dimensional point set P of extraction and two-dimentional point set p of the same name to second as pair right image use changing based on collinearity equation
Enter Danish law iteration method with variable weights progress space resection and obtain locus and the posture of right image, then according to above-mentioned identical
The two three-dimensional indexes that method carries out next stereogram image are established, two three-dimensional same places determine and the orientation of subsequent images, are obtained
To the position of each image and posture initial value；
3.3) using the position of each image and posture initial value as the initial value of bundle adjustment, using collinearity equation as
Mathematical modeling carry out free net bundle adjustment, using Levenberg-Marquardt (LM) algorithm solution ask orientation parameter with
Three-dimensional point corresponding to SIFT+ Least squares matchings point, to obtain the position of every image and attitude parameter.
Preferably, in the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint,
In the step 4, specifically comprise the following steps：
4.1) grid partition is carried out to every image according to a fixed step size, according to each in Harris feature extraction and calculation windows
The interest value of individual pixel, the extreme point of interest value as characteristic point and application Fostner operators accurately determine using in window
Position, equally distributed characteristic point is established on the left image of stereogram；
4.2) same place between neutral body picture pair is oriented first with image, homography matrix parameter is calculated, by solid
As the characteristic point on the left image of centering is according on one-to-one relationship homograph to right image, using homograph point as
The thick value point of same place corresponding to characteristic point, afterwards using core line geometry relation, same place must be on corresponding epipolar line one-dimensional searches
Rope matching process, further reduces the hunting zone of matching, is determined in hunting zone first with correlation coefficient matching method of the same name
The initial position of point, then same place is accurately positioned by Least squares matching, what is be evenly distributed in stereogram is of the same name
Point；
4.3) based on the characteristic point on left image, Ronny Di's triangular network is built, with reference to the corresponding pass between same place
System, builds the corresponding triangulation network on right image, and the triangle that its same place is formed is referred to as triangle of the same name；
4.4) using the triangulation network on left image as matching unit, interpolation triangle core point, with corresponding triangle of the same name
Core line geometry between shape and stereogram is as constraints, and search matching is of the same name in right image accordingly triangle of the same name
Point, the triangulation network is constantly updated, according to the characteristics of Ronny Di's triangular network, side as center of gravity interpolation condition and is set using the triangle length of side
Long threshold value, no longer this triangle of interpolation when the length of side of triangle is less than threshold value are of the same name to obtain the historical relic image high density
Point.
Preferably, the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint,
After the step 4, also include before step 5：
Step A, it is right using the Di Luoni triangulation networks as decision condition using Canny operator extraction image edge information
Marginal point beyond the triangulation network, hunting zone is determined with thick the value point and core line geometry of homograph, in the triangulation network
Marginal point, hunting zone is determined with the triangle of the same name where marginal point and core line geometry, it is determined that hunting zone in use
Coefficient correlation and Least squares matching are accurately positioned historical relic image edge same place；
Afterwards, in step 5, according to the position of every image and attitude parameter, by the historical relic image high density
Same place carries out point off density reconstruction and the historical relic image edge same place carries out point off density reconstruction and edge reconstruction respectively, close
Marginal point cloud corresponding to fusion on the basis of collection point cloud, obtains the historical relic image fine-point cloud.
Preferably, in the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint,
The step 2, specifically comprises the following steps：
2.1) homotopy mapping is carried out between the sequential images and then passes sequentially through two-way consistency constraint,
RANSAC random sampling consistency constraints, affine transformation constraint is constrained item by item step by step, to improve the accuracy of same place,
Obtain high-quality point set of the same name；
2.2) the high-quality point set of the same name is further accurately positioned using Least squares matching method, obtained described
Sequential images orientation same place.
Preferably, the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint,
After the step 1, also include step B before step 2：
First, camera calibration is carried out to the camera, obtains the internal reference element and distortion parameter of camera, wherein, it is described interior
Join element to be used to orient image, the distortion parameter is used for image data distortion rectification；
Afterwards, distortion rectification is carried out to the sequential images data, and processing is filtered to the image after distortion；Then
Enter back into step 2.
Preferably, in the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint,
In the step B, when carrying out the camera calibration, the image for calculating camera calibration parameter will be relative to the difference of scaling board
Image series are shot at angle, different distance.
Preferably, in the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint,
In the step 6, the three-dimensional point cloud scans to obtain using Articulated arm scanners.
The present invention comprises at least following beneficial effect：
Firstth, the present invention joins to initial picture to calculating the orientation between picture pair using direct solution plus the method tightly solved first
Number, while its three-dimensional coordinate is resolved to SIFT same place application forward intersections, and establish each as to three-dimensional coordinate and tie point
Two three-dimensional indexes.Using two three-dimensional indexes of above-mentioned foundation extract next pictures to corresponding three-dimensional point and its with the 3rd shadow
The corresponding image points of picture, locus and posture as the right image of centering are resolved in the method for space resection.Pass through this side
Method resolves the orientation parameter of image successively, using obtained image orientation parameter as initial value, then by free net bundle adjustment,
The accurate orientation parameter for determining sequential images.Secondth, can be established by the inventive method in the stereogram in sequential images
Closeness is high and equally distributed same place, solves that dense Stereo Matching Midst density is low, skewness, and in dense Stereo Matching with
Triangle carries out focus point interpolation matching, with the core line Constrain Searching model in triangle of the same name as matching unit to triangle
Enclose, be continuously increased the same place of matching, to reach the purpose of dense Stereo Matching.3rd, historical relic object is carried out by the inventive method
Close shot sequential images subtle three-dimensional is rebuild, and can be generated and be become more meticulous the abundant imaging point cloud of degree height, detailed information and have reality
Size and yardstick, imaging point cloud analysis can be measured.
The present invention and for historical relic object close shot sequential images subtle three-dimensional Problems of Reconstruction, with ground laser radar point cloud and
Close-range image is data source, research camera calibration, Image Matching, image orientation, dense Stereo Matching, marginal information matching, image weight
Build, absolute orientation etc. the problems such as, by being furtherd investigate to each several part, realize the historical relic object based on triangulation network interpolation and constraint
Close shot sequential images subtle three-dimensional is rebuild.The achievement in research of the present invention can meet to become more meticulous the requirement of reconstruction, be also applied to city
Other industries such as city's planning auxiliary, antique information retention, three-dimensional display.
Further advantage, target and the feature of the present invention embodies part by following explanation, and part will also be by this
The research and practice of invention and be understood by the person skilled in the art.
Brief description of the drawings
Fig. 1 is the historical relic sequential images subtle three-dimensional method for reconstructing of the present invention based on triangulation network interpolation and constraint
Flow chart；
Fig. 2 is the camera chessboard calibration image schematic diagram obtained in the embodiment of the present invention；
Fig. 3 A and 3B are the distortion parameter calculated in the embodiment of the present invention respectively according to camera calibration, carry out the abnormal of image
Become the schematic diagram before correcting and after correction；
Fig. 4 is the laser point cloud data schematic diagram for the close-range image for being used for reconstruction in the embodiment of the present invention；
Fig. 5 A and 5B are the stereogram schematic diagram chosen in the sequential images in the embodiment of the present invention；
Fig. 6 is the result schematic diagram of the images filter pretreatment in the embodiment of the present invention；
Fig. 7 A and 7B are neutral body of embodiment of the present invention picture to SIFT+ Least squares matching result schematic diagrams；
Fig. 8 A and 8B are neutral body of embodiment of the present invention picture to SIFT+ Least squares matching result partial schematic diagrams；
Fig. 9 A and 9B are the stereo matching result schematic diagram after error hiding rejecting in accordance with the present invention；
Figure 10 A and 10B are the stereogram local matching results schematic diagram after error hiding rejecting in accordance with the present invention；
Figure 11 is the result schematic diagram that image in accordance with the present invention orients for the first time；
Figure 12 is inconsistent phenomenon schematic diagram present in image in accordance with the present invention for the first time orientation；
Figure 13 is the result schematic diagram after free net bundle adjustment in accordance with the present invention；
Figure 14 A and 14B be neutral body of embodiment of the present invention picture to Harris+Fosrtner Feature Points Matchings after, established
It is uniformly distributed the result schematic diagram of same place；
Figure 15 is the result of neutral body of embodiment of the present invention Ronny Di's triangular network of the same name as constructed by being uniformly distributed same place
Schematic diagram；
Figure 16 is the result schematic diagram according to stereogram dense Stereo Matching of the present invention.
Figure 17 is the result schematic diagram extracted according to neutral body of embodiment of the present invention picture to left image edge；
Figure 18 is the result schematic diagram according to neutral body of embodiment of the present invention picture to edge information matches；
Figure 19 is the result schematic diagram according to neutral body of embodiment of the present invention picture to edge reconstruction；
Figure 20 is the result schematic diagram to finely rebuilding according to neutral body of embodiment of the present invention picture；
Figure 21 is the result schematic diagram finely rebuild according to sequential images in the embodiment of the present invention.
Embodiment
The present invention is described in further detail below in conjunction with the accompanying drawings, to make those skilled in the art with reference to specification text
Word can be implemented according to this.
It should be appreciated that such as " having ", "comprising" and " comprising " term used herein do not allot one or more
The presence or addition of individual other elements or its combination.
As shown in figure 1, the present invention provides a kind of historical relic sequential images subtle three-dimensional weight based on triangulation network interpolation and constraint
Construction method, including：
Step 1: the sequential images of historical relic are gathered using camera；
Step 2: carrying out homotopy mapping between the sequential images, sequential images orientation same place is obtained；For example,
Homotopy mapping is carried out between sequential images, error hiding is rejected by multiple constraint step by step, obtains high-quality point set of the same name, then through most
A young waiter in a wineshop or an inn multiplies matching and obtains accurate, high-precision same place between image.
Step 3: resolving the orientation parameter for obtaining every image first with the same place, determined afterwards with the image
To parameter as initial value, by bundle adjustment, the position of every image of acquisition and attitude parameter；For example, first calculate initial picture
Orientation parameter between pair, while its three-dimensional coordinate is resolved to same place application forward intersection, and establish as to three-dimensional coordinate and company
The three-dimensional index of the two of contact.Above-mentioned three-dimensional point is extracted with next picture to corresponding three-dimensional using two three-dimensional indexes of above-mentioned foundation
Point and its corresponding image points with the 3rd image, with the method for space resection resolve as the right image of centering locus and
Posture.Resolve the orientation parameter of every image successively by this method, using obtained image orientation parameter as initial value, then lead to
Free net bundle adjustment is crossed, the accurate orientation parameter for determining sequential images.
Step 4: the dense Stereo Matching of sequential images neutral body picture pair：In stereogram, to left image with Harris lattices
Net extraction characteristic point is simultaneously accurately positioned using Fostner operators, and the corresponding same place of search matching is established equal on right image
The same place of even distribution, the Di Luoni triangulation networks of the same name are built using equally distributed same place as seed point, by triangle
The continuous interpolation matching of focus point, is continuously increased the same place of matching, obtains historical relic image high density same place；
Step 5: according to the position of the every image obtained in step 3 and attitude parameter, by the historical relic image
High density same place carries out point off density reconstruction, builds historical relic image fine-point cloud；
Step 6: using the three-dimensional point cloud of the historical relic as reference, realized by the spacial similarity transformation between coordinate system
The absolute orientation of the historical relic image fine-point cloud, make historical relic image fine-point cloud that there is actual size and yardstick, obtain historical relic
The subtle three-dimensional model of sequential images.For example, using joint arm scanning element cloud as reference, control point is chosen, by imaging point cloud
Image space rectangular coordinate system is transformed under the partial sweep coordinate system of joint arm, passes through the spacial similarity transformation between Two coordinate system
The absolute orientation of imaging point cloud is realized, makes imaging point cloud that there is actual size and yardstick on the basis of detailed information is fine.
In one of embodiment of the present invention, preferably, in the step 3, specifically comprise the following steps：
3.1) same place that will be obtained in step 2, directly solved by relative orientation and obtained with reference to the method tightly solved
The relative position and attitude parameter of initial volumetric picture pair, while its three-dimensional coordinate is resolved to the same place application forward intersection；
3.2) index relative will be established between the two-dimentional same place of initial volumetric picture pair and the three-dimensional point of its forward intersection, and
According to the same place of the two or three image, determine initial volumetric picture to two-dimensional points of the same name between three-dimensional point P and the 3rd image
P, afterwards by the three-dimensional point set P of extraction and two-dimentional point set p of the same name to second as pair right image use changing based on collinearity equation
Enter Danish law iteration method with variable weights progress space resection and obtain locus and the posture of right image, then according to above-mentioned identical
The two three-dimensional indexes that method carries out next stereogram image are established, two three-dimensional same places determine and the orientation of subsequent images, are obtained
To the position of each image and posture initial value；
3.3) using the position of each image and posture initial value as the initial value of bundle adjustment, using collinearity equation as
Mathematical modeling carry out free net bundle adjustment, using Levenberg-Marquardt (LM) algorithm solution ask orientation parameter with
Three-dimensional point corresponding to SIFT+ Least squares matchings point, to obtain the position of every image and attitude parameter.
In some embodiments of the invention, preferably, in the step 4, specifically comprise the following steps：
4.1) grid partition is carried out to every image according to a fixed step size, according to each in Harris feature extraction and calculation windows
The interest value of individual pixel, the extreme point of interest value as characteristic point and application Fostner operators accurately determine using in window
Position, equally distributed characteristic point is established on the left image of stereogram；
4.2) same place between neutral body picture pair is oriented first with image, homography matrix parameter is calculated, by solid
As the characteristic point on the left image of centering is according on one-to-one relationship homograph to right image, using homograph point as
The thick value point of same place corresponding to characteristic point, afterwards using core line geometry relation, same place must be on corresponding epipolar line one-dimensional searches
Rope matching process, further reduces the hunting zone of matching, is determined in hunting zone first with correlation coefficient matching method of the same name
The initial position of point, then same place is accurately positioned by Least squares matching, what is be evenly distributed in stereogram is of the same name
Point；
4.3) based on the characteristic point on left image, Ronny Di's triangular network is built, with reference to the corresponding pass between same place
System, builds the corresponding triangulation network on right image, and the triangle that its same place is formed is referred to as triangle of the same name；
4.4) using the triangulation network on left image as matching unit, interpolation triangle core point, with corresponding triangle of the same name
Core line geometry between shape and stereogram is as constraints, and search matching is of the same name in right image accordingly triangle of the same name
Point, the triangulation network is constantly updated, according to the characteristics of Ronny Di's triangular network, side as center of gravity interpolation condition and is set using the triangle length of side
Long threshold value, no longer this triangle of interpolation when the length of side of triangle is less than threshold value are of the same name to obtain the historical relic image high density
Point.
In one of embodiment of the present invention, preferably, after the step 4, also wrapped before step 5
Include：
Step A, it is right using the Di Luoni triangulation networks as decision condition using Canny operator extraction image edge information
Marginal point beyond the triangulation network, hunting zone is determined with thick the value point and core line geometry of homograph, in the triangulation network
Marginal point, hunting zone is determined with the triangle of the same name where marginal point and core line geometry, it is determined that hunting zone in use
Coefficient correlation and Least squares matching are accurately positioned historical relic image edge same place；
Afterwards, in step 5, according to the position of every image and attitude parameter, by the historical relic image high density
Same place carries out point off density reconstruction and the historical relic image edge same place carries out point off density reconstruction and edge reconstruction respectively, close
Marginal point cloud corresponding to fusion on the basis of collection point cloud, obtains the historical relic image fine-point cloud.
In one of embodiment of the present invention, preferably, the step 2, specifically comprises the following steps：
2.1) homotopy mapping is carried out between the sequential images and then passes sequentially through two-way consistency constraint,
RANSAC random sampling consistency constraints, affine transformation constraint is constrained item by item step by step, to improve the accuracy of same place,
Obtain high-quality point set of the same name；
2.2) the high-quality point set of the same name is further accurately positioned using Least squares matching method, obtained described
Sequential images orientation same place.
In one of embodiment of the present invention, preferably, after the step 1, also include before step 2
Step B：
First, camera calibration is carried out to the camera, obtains the internal reference element and distortion parameter of camera, wherein, it is described interior
Join element to be used to orient image, the distortion parameter is used for image data distortion rectification；
Afterwards, distortion rectification is carried out to the sequential images data, and processing is filtered to the image after distortion；Then
Enter back into step 2.
In one of embodiment of the present invention, preferably, in the step B, when carrying out the camera calibration, use
Will be relative to shooting image series at the different angle of scaling board, different distance in the image for calculating camera calibration parameter.
In one of embodiment of the present invention, preferably, in the step 6, the three-dimensional point cloud uses joint
Arm scanner scanning obtains.
To make those skilled in the art more fully understand the present invention, examples below is now provided and illustrated：
As shown in figure 1, the present invention provides a kind of historical relic sequential images subtle three-dimensional weight based on triangulation network interpolation and constraint
Construction method, including：
Step 1: camera calibration：Is obtained by uncalibrated image, and is pressed in diverse location, different distance for the camera for gathering image
The internal reference element and distortion parameter of camera are calculated according to the camera calibration method of computer vision, wherein Fig. 2 is the demarcation collected
Image, it is specially：
With reference to pin-hole model generally principal point position in the picture and camera focus composition in camera calibration
Matrix is referred to as internal reference matrix, and the relation between image pixel coordinates and scaling board coordinate is as follows：
Q=MQ (1)
Wherein
Q is the pixel coordinate using the image upper left corner as the origin of coordinates in formula；F in matrix Mx fyFor camera x-axis and y-axis direction
Focal length；S represents imaging plane x-axis and the non-orthogonal property in y-axis direction, and general s is 0；
It can be represented by the formula for the imaging process of point to camera at gridiron pattern diverse location：
M=λ A [R t] M (3)
Wherein M=[X Y Z 1]TFor the space point coordinates in world coordinate system under homogeneous coordinates；M=[u v 1]TTo be neat
Two-dimensional points coordinate under secondary coordinate in pixel coordinate system；λ is scaling；
Scaling board is assumed in camera calibration in world coordinate system Z=0 plane, then formula 2-20 can further simplify
For：
Take H=λ A [r1 r2T]=[h1 h2 h3], matrix H is also referred to as homography matrix, then
Rotating vector is mutually orthogonal in construction in above formula, scaling zoom factor is mentioned into outside, then r1With r2Mutually
It is orthogonal.Then have
According to the implication that two vectors are orthogonal, then there are following two basic constraintss：
In order that being expressed below relatively easily, B=A is taken-TA-1, embody as follows：
Actually matrix B has the closing solution of common version：
Matrix B is incorporated into two basic constraintss, then has general constraint typeIn view of square
Battle array B symmetry, and matrix is deployed to the form of each element, rearrange to obtain a new vectorial b, then have：
With reference toDefinition, two basic definition can also be write as：
If K checkerboard image, accumulation is listed these equations, then had：
Vb=0 (14)
Wherein V is 2*K*6 matrix, is obtained in the closing solution of matrix B common version：
Wherein
Obtained for rotating with being translated across following formula：
When actually solving, R=[r are taken1 r2 r3], to matrix R carry out singular value decomposition obtain a diagonal matrix D and
Two orthogonal matrix U and V, while it is unit battle array to set diagonal matrix, makes R=UDVT, so as to obtain internal reference matrix M and distortion matrix P.
WhereinP=[k1 k2 p1 p2 k3]。
[k1 k2 k3] it is coefficient of radial distortion, [p1 p2] it is tangential distortion coefficient.
With reference to pin-hole model, using distortion parameter to (xd,yd) distortion point corrected to obtain correct point (xp,yp), it is as follows
It is shown：
Calculate distortion parameter p and the internal reference element M of camera in calibrating procedure, and to space three-dimensional point inverse subpoint,
Labeling projection error V is equal to 0.1708 pixel.
P=[0.987627-40.8725 0.00451 0.01172 816.1225] (19)
Step 2: the distortion rectification and filtering process of image；
The close-range image data of measurand acquired in this paper and corresponding laser point cloud data are as shown in figure 4, take it
In a solid object as shown in figure 5, being corrected demarcation image distortion as shown in figure 3, equally to all near using distortion parameter
Scape image carries out distortion rectification, and images filter is pre-processed with Wallis wave filters as shown in fig. 6, wherein, Wallis is filtered
The form of expression be：
Wallis wave filters are also referred to as
gc(x, y)=g (x, y) r1+r0 (21)
r1=(csf)/(csg+sf/c) (22)
r0=bmf+(1-b)mg (23)
Wherein, r1To multiply property coefficient, r0For additivity coefficient, c represents image contrast extension constant；B represents image brilliance system
Number, mgFor the image greyscale average of certain neighborhood of a certain pixel in image.mfFor the ash of certain neighborhood of a certain pixel in image
Spend variance, sfFor the desired value of image variance, sgFor the gray variance of certain neighborhood of a certain pixel in image.
Step 3: the homotopy mapping between sequential images：
(1), the result of SIFT initial matchings is as shown in Figure 7.It can be seen that in Fig. 8 close-up schematic view, in match point
A large amount of error hidings can be contained, therefore, by two-way consistency constraint, RANSAC random sampling consistency constraints, affine transformation is about
The sequencing of beam is constrained item by item step by step, to improve the accuracy of same place；
(2), after multiple constraint obtains high-quality matching set, Least squares matching is recycled further to be accurately positioned of the same name
Point, the same place that the same place so obtained is oriented as image.As a result as shown in figure 9, wherein Figure 10 is the part of matching
Schematic diagram, it can be seen that the same place point position of matching is accurate, and precision is high.
Step 4: sequential images orient, position and the attitude parameter of each image are determined：
S1, using SIFT+ Least squares matchings the same place between stereogram is obtained, with reference to the same place of matching, led to
Relative orientation directly solution plus the tightly accurate relative position and attitude parameter for obtaining initially picture pair of solution are crossed, is specially：
Photographic base S1S2And corresponding image rays S1M、S2M is coplanar, that is, meets coplanar condition, its mathematical modeling
B·(R1×R2)=0 (24)
Wherein vectorial B represents photographic base S1S2；R1And R2Corresponding image rays S is represented respectively1M、S2M。
Coplanar condition equation is represented with coordinate form, its matrix form is as follows：
Wherein [X1 Y1 Z1]T[X2 Y2 Z2]TFor point coordinates under the auxiliary coordinates of image space, may particularly denote for：
Wherein a1a2a3b1b2b3c1c2c3For the matrix of left image rotation parameter composition；a’1a’2a’3b’1b’2b’3c’1c’2c’3For the matrix of right image rotation parameter composition.
The image space auxiliary coordinates of left image are generally taken to be overlapped with the direct coordinate system in image space, then (26) can represent as follows
Form：
Same place is substituted into the coplanar condition equation of (25) using formula (26) with (27), its specific expanded form is as follows：
L1y1x2+L2y1y2-L3y1f+L4fx2+L5fy2-L6ff+L7x1x2+L8x1y2-L9x1F=0 (28)
By formula (28) equation the right and left while divided by L5, it is as follows to arrange equation result：
Wherein
After the unknown number in solving (29) formula, if given BxNumerical value, then it can solve required orientation parameter, each ginseng
Number expression formula is as follows：
In (30) formula, L5ValueDue to flat
The effect of side, L5Value have positive and negative two values, can be seen that in formula (30), L5Value can produce shadows to 9 rotation parameters
Ring, therefore two groups of different positions and attitude angle can be calculated, it is contemplated that corresponding image rays is that difference takes the photograph station same culture point is obtained
Take what image information was formed.Then the angle element of right image should meet following condition：
In order to obtain the orientation parameter that the degree of accuracy is good, precision is high, on the basis of the direct solution of relative orientation, with what is directly solved
Parameter is as initial value, the iterative in the formula that relative orientation tightly solves.Tight solution formula represents as follows：
Wherein Q is vertical parallax,
S2, index relative will be established between the two-dimentional same place of initial volumetric picture pair and the three-dimensional point of its forward intersection, and
According to the same place of two or three images, it is determined that initial as to two-dimensional points p of the same name between three-dimensional point P and the 3rd image.It will carry
The three-dimensional point set P taken and two-dimentional point set p of the same name is selected using the improvement Danish law based on collinearity equation the right image of the second picture pair
Power iterative method carries out space resection and obtains locus and the posture of right image, carries out next picture by the same way to shadow
Two three-dimensional indexes of picture are established, two three-dimensional control points of the same name determine and the orientation of subsequent images, obtain the position of each image with
Posture initial value.The result of its first sparse reconstruction is as shown in figure 11, and to reconstructed results rotation, scaling, there is offset error
As shown in figure 12.
S3, the result to orient for the first time are carried out freely as the initial value of bundle adjustment by mathematical modeling of collinearity equation
Net bundle adjustment, quickly solution orientation parameter and SIFT+ least squares can be asked using Levenberg-Marquardt (LM) algorithm
Three-dimensional point corresponding to match point.Specially：
Using collinearity equation as mathematical modeling in bundle adjustment, its mathematical modeling represents as follows：
On the basis of the image space auxiliary coordinate of first photo influenceed herein by the sequence of shooting, wherein (x, y) table
Show the picpointed coordinate under image plane rectangular coordinate system；(X0,Y0,Z0) represent photo centre space coordinates, f is camera focus；
(X, Y, Z) is space three-dimensional point coordinates；(ak,bk,ck) (k=1,2,3) expression is by the angle element φ of photo, the side that ω, κ are formed
To cosine.
For several images, according to the order of sequential images, by the position for calculating every width image and the initial value of posture
It is represented sequentially as：
It is reconstruction point coordinate X to take p=[△ X, △ Y, △ Z]j,Yj,ZjIncrement,
For elements of exterior orientation eiIncrement, collinearity equation formula (33) Taylor linearization is deployed, and establishes error equation, expression formula is：
Wherein, vx vyFor correction；lx lyFor x, y approximation；aij(i=1,2；J=1,2 ... 6) it is weighted number.
Then above formula can further be write as
It can be reduced to:
Wherein
On the basis of above-mentioned conventional beam method adjustment, image position is resolved using Levenberg-Marquardt (LM) algorithm
Put and posture and three-dimensional point coordinate.
Bundle adjustment mainly solves the problems, such as it is to calculate accurate orientation parameter and SIFT+ Least squares matching point pair
The three-dimensional point answered, the levels of precision of its adjustment directly influence the observation of imaging point and the size of estimate error.It is now assumed that
By on n spot projection in space to m image, x is usedijRepresent the point that i-th of spatial point is projected on jth image.
Then error size V can be represented by the formula：
Wherein, cijRepresent indicator parameter, general cij=1, d (x, y) represent observation station x to inverse estimation point y between away from
From bundle adjustment is exactly the process for minimizing error V in fact, and orientation parameter member is constantly corrected during error minimizes
Element and three-dimensional point coordinate, therefore the Rule of judgment by the use of error size V as adjustment precision.
By a in above formulajWith biIt is unified into set P, and by observation xijIt is incorporated into observation vector X, is expressed as：
Wherein,Coordinates of the representation space point i in the image space rectangular coordinate system of image.
Take P0For initial parameter, ∑ x is as covariance, covariance matrix as unit under no any priori conditions
Matrix, use P0Estimation observation vector X can be expressed as below：
Wherein
Based on the principle of least square, bundle adjustment, which is converted to ask, minimizes mahalanobis distance such as following formula (41) expression,
According to LM algorithms, normal equation is represented by following formula (42)
JT∑x-1J δ=JT∑x-1ε (42)
Wherein J is Jacobian matrix, and δ represents step-size in search.Due to consideration that the irrelevance of projection matrix, above formula (42)
It is middle a sparsity structure to be present, and sparsity structure computational efficiency is high, quickly solution orientation parameter can be asked corresponding with SIFT match points
Three-dimensional point.It is as shown in figure 13 that free net bundle adjustment correction result is carried out to first orientation result.
Step 5: the dense Stereo Matching of sequential images neutral body picture pair：
Ith, grid partition is carried out to image according to a fixed step size, according to each pixel in Harris feature extraction and calculation windows
The interest value of point, the extreme point of interest value is established on the left image of stereogram and is uniformly distributed as characteristic point using in window
Characteristic point, it is specially：
After Harris establishes equally distributed characteristic point, the characteristic point of extraction is accurately positioned using Fostner operators.
Wherein Fostner operators positioning normal equation formula be：
IIth, the SIFT+ Least squares matching high accuracy same places in being oriented using image between picture pair, homography matrix ginseng is calculated
Number.By as the Harris+Fostner characteristic points on the left image of centering are according on one-to-one relationship homograph to right image,
Thick value point using homograph point as same place corresponding to characteristic point.Using core line geometry relation, same place must be in core of the same name
Linear search matching process on line, further to reduce the hunting zone of matching, first with phase relation in hunting zone
Number matching determines the initial position of same place, then is accurately positioned same place by Least squares matching, is obtained in stereogram
Equally distributed same place.As a result as shown in figure 14, wherein being specially：
Homography matrix be for describing as to a kind of mapping relations between same place, in being oriented using image as pair between
SIFT+ Least squares matching high accuracy same places, calculate homography matrix parameter, and mathematical modeling is expressed as：
Wherein (x, y) is the pixel coordinate put on left image；H is homography matrix；Zoom factors of the s between coordinate.
After the completion of image orientation, according to coplanar condition equation, as the core line geometry relation between it has been determined that following institute
Show
Wherein, each term coefficient in formula has resolved in orientation, for the point (x on left image1,y1), if known
X on right image2, then the y on core line can be calculated according to formula 462Value.
The hunting zone corresponding to left image feature point is found on right image, utilizes coefficient correlation and Least squares matching
Determine same place.The maximum pixel of coefficient correlation is found in hunting zone first above and below core line, passes through coefficient correlation threshold
Whether value is same place to screen.Coefficient correlation is represented by：
Wherein m, n is search box size, general m=n；gi,jFor grey scale pixel value size in m*n target windows；
g′i+r,j+cThe grey scale pixel value being expressed as in search window；It is expressed as pixel grey scale average in target window；For search window
Pixel grey scale average in mouthful；
Correlation coefficient matching method is based on grayscale image, and camera collection is all color digital image, in order to improve phase
The degree of accuracy of coefficients match is closed, the coefficient correlations of R, G, B primary display channels is contemplated herein, it is final similar to determine with average
Spend Pend。
IIIth, the same place that Harris+Fostner characteristic matchings are evenly distributed on the image of left and right, with left image
Same place based on, build Ronny Di's triangular network, with reference to the corresponding relation between same place, built on right image corresponding
The triangulation network, the triangle that its same place is formed are referred to as triangle of the same name, and the triangle of the same name constructed by stereogram is shaped like Figure 15 institutes
Show；
IVth, using the triangle on left image as matching unit, interpolation triangle core point, with corresponding triangle of the same name
Core line geometry between picture pair is as constraints, and search matching same place is continuous more in right image accordingly triangle of the same name
The new triangulation network, according to the characteristics of Ronny Di's triangular network, length of side threshold value as center of gravity interpolation condition and is set using the triangle length of side, should
Length of side threshold value determines according to the density of finally desired three-dimensional point, when the length of side of triangle is less than threshold value just no longer interpolation this three
Angular, dense Stereo Matching result is as shown in figure 16, wherein being specially：
The area of left side Harris+Fostner matching structure Ronny Di's triangular network intermediate cam shapes is calculated first, is judged and is set
Determine the size of area threshold, its area S calculation formula are as follows
Wherein, (x1,y1), (x2,y2), (x3,y3) be same place pixel coordinate.
If the area of triangle map unit is more than area threshold, interpolation triangle core point, center of gravity point coordinates calculates
Formula is as follows
According to the core line geometry relation of orientation result neutral body picture pair, interpolation focus point is calculated in the core corresponding to right image
Line, the linear equation on Atria bar side of the same name corresponding to calculating, and the intersection point with core line is obtained, determine core line according to two intersection points
Scope in triangle of the same name, searched for correlation coefficient matching method in the range of the core line in triangle of the same name and obtain matching initial
Value, then it is accurately positioned same place position with Least squares matching.
Focus point in left image triangle is matched into the same place obtained with search in right image triangle of the same name to split
The triangulation network is updated simultaneously into three triangles, according to corresponding point position relation storage, fractionation rear triangle is still kept of the same name
Triangle principle, and repeat the above steps.
Step 6: the marginal information matching result of stereogram：
It is that the edge feature of the object extracted from image (may be three using Canny operator extraction image edge information
In angular, it can also be overlapped with triangulation network summit, it is also possible to and the side of triangle is intersected).As shown in figure 17, with dense Stereo Matching
The Di Luoni triangulation networks of seed point structure are as decision condition, to the marginal point beyond the triangulation network, with the thick value of homograph
Point and core line geometry determine hunting zone.To the marginal point in the triangulation network, with the triangle of the same name where marginal point and core line
Geometry determines hunting zone, it is determined that hunting zone in using coefficient correlation and Least squares matching to be accurately positioned high accuracy same
Famous cake, the marginal information matching result of stereogram are as shown in figure 18.
Step 7: sequential images are finely rebuild：
Highdensity same place is obtained according to dense Stereo Matching method, marginal information matches to obtain high-precision edge of the same name
Point, utilize edge same place of the orientation parameter of image to sequential images neutral body picture pair, high density same place result progress side
Edge is rebuild and point off density is rebuild, the marginal point cloud corresponding to fusion on the basis of point off density cloud, as Figure 19,20 show wherein one
Individual stereogram edge reconstruction result and edge reconstruction and point off density rebuild the fine reconstructed results blended, finally give details
The high fine reconstruction point cloud of sequential images of abundant information, point cloud density, as a result as shown in figure 21.
Step 8: imaging point cloud absolute orientation
Using joint arm scanning element cloud as reference, control point is chosen, by the image space rectangular coordinate system conversion of imaging point cloud
To under the partial sweep coordinate system of joint arm, the definitely fixed of imaging point cloud is realized by the spacial similarity transformation between Two coordinate system
To making imaging point cloud that there is actual size and yardstick on the basis of detailed information is fine, can be specially：
The arbitrfary point image space rectangular coordinate system for taking imaging point cloud is (XP YP ZP), corresponding joint arm partial sweep
Coordinate system is (Xtp Ytp Ztp), spacial similarity transformation model is
Wherein, △ X, △ Y, △ Z are translation parameters；ai,bi,ciFor rotation parameter；λ is image space rectangular coordinate system to pass
The zoom scale of joint arm partial sweep coordinate system.
Module number and treatment scale described herein are the explanations for simplifying the present invention.To the present invention based on three
The application of angle net interpolation and the historical relic sequential images subtle three-dimensional method for reconstructing of constraint, modifications and variations are to those skilled in the art
It is obvious for member.
Although embodiment of the present invention is disclosed as above, it is not restricted in specification and embodiment listed
With it can be applied to various suitable the field of the invention completely, can be easily for those skilled in the art
Other modification is realized, therefore under the universal limited without departing substantially from claim and equivalency range, it is of the invention and unlimited
In specific details and shown here as the legend with description.
Claims (8)
- A kind of 1. historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint, it is characterised in that including：Step 1: the sequential images of historical relic are gathered using camera；Step 2: carrying out homotopy mapping between the sequential images, sequential images orientation same place is obtained；Step 3: resolving the orientation parameter for obtaining every image first with the same place, oriented join with the image afterwards Number is used as initial value, by bundle adjustment, the position of every image of acquisition and attitude parameter；Step 4: the dense Stereo Matching of sequential images neutral body picture pair：In stereogram, left image is carried with Harris grids Take characteristic point and be accurately positioned using Fostner operators, the corresponding same place of matching is searched on right image and is established and is uniformly divided The same place of cloth, the Di Luoni triangulation networks of the same name are built using equally distributed same place as seed point, by triangle core The continuous interpolation matching of point, is continuously increased the same place of matching, obtains historical relic image high density same place；Step 5: according to the position of the every image obtained in step 3 and attitude parameter, the historical relic image is highly dense Spend same place and carry out point off density reconstruction, build historical relic image fine-point cloud；Step 6: using the three-dimensional point cloud of the historical relic as reference, by described in the spacial similarity transformation realization between coordinate system The absolute orientation of historical relic image fine-point cloud, make historical relic image fine-point cloud that there is actual size and yardstick, obtain historical relic sequence The subtle three-dimensional model of image.
- 2. the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint as claimed in claim 1, its It is characterised by, in the step 3, specifically comprises the following steps：3.1) same place that will be obtained in step 2, directly solved by relative orientation and obtained initially with reference to the method tightly solved The relative position and attitude parameter of stereogram, while its three-dimensional coordinate is resolved to the same place application forward intersection；3.2) index relative will be established between the two-dimentional same place of initial volumetric picture pair and the three-dimensional point of its forward intersection, and according to The same place of two or three image, initial volumetric picture is determined to two-dimensional points p of the same name between three-dimensional point P and the 3rd image, it It is afterwards that the three-dimensional point set P of extraction and two-dimentional point set p of the same name is red using the improvement based on collinearity equation to the right image of the second picture pair Wheat method iteration method with variable weights carries out space resection and obtains locus and the posture of right image, then according to above-mentioned same procedure The two three-dimensional orientations for indexing foundation, two three-dimensional same places determinations and subsequent images of next stereogram image are carried out, are obtained each The position of individual image and posture initial value；3.3) using the position of each image and posture initial value as the initial value of bundle adjustment, using collinearity equation as mathematics Model carries out free net bundle adjustment, and orientation parameter and SIFT+ are asked most using Levenberg-Marquardt (LM) algorithm solution A young waiter in a wineshop or an inn multiplies three-dimensional point corresponding to match point, to obtain the position of every image and attitude parameter.
- 3. being rebuild based on triangulation network interpolation and the historical relic sequential images subtle three-dimensional of constraint as described in claim 1 or 2 is any Method, it is characterised in that in the step 4, specifically comprise the following steps：4.1) grid partition is carried out to every image according to a fixed step size, according to each picture in Harris feature extraction and calculation windows The interest value of vegetarian refreshments, the extreme point of interest value is accurately positioned as characteristic point and using Fostner operators using in window, Equally distributed characteristic point is established on the left image of stereogram；4.2) same place between neutral body picture pair is oriented first with image, homography matrix parameter is calculated, by stereogram The characteristic point on middle left image is according on one-to-one relationship homograph to right image, using homograph point as feature The thick value point of same place corresponding to point, afterwards must be in the linear search on corresponding epipolar line using core line geometry relation, same place Method of completing the square, the hunting zone of matching is further reduced, same place is determined first with correlation coefficient matching method in hunting zone Initial position, then same place is accurately positioned by Least squares matching, the same place being evenly distributed in stereogram；4.3) based on the characteristic point on left image, Ronny Di's triangular network is built, with reference to the corresponding relation between same place, The corresponding triangulation network is built on right image, the triangle that its same place is formed is referred to as triangle of the same name；4.4) using the triangulation network on left image as matching unit, interpolation triangle core point, with corresponding triangle of the same name and Core line geometry between stereogram is as constraints, the search matching same place in right image accordingly triangle of the same name, no The disconnected renewal triangulation network, according to the characteristics of Ronny Di's triangular network, as center of gravity interpolation condition and length of side threshold is set using the triangle length of side Value, no longer this triangle of interpolation when the length of side of triangle is less than threshold value, to obtain the historical relic image high density same place.
- 4. the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint as claimed in claim 3, its It is characterised by, after the step 4, also includes before step 5：Step A, using Canny operator extraction image edge information, using the Di Luoni triangulation networks as decision condition, to three Marginal point beyond the net of angle, determines hunting zone, to the edge in the triangulation network with thick the value point and core line geometry of homograph Point, hunting zone is determined with the triangle of the same name where marginal point and core line geometry, it is determined that hunting zone in using correlation Coefficient and Least squares matching are accurately positioned historical relic image edge same place；Afterwards, it is according to the position of every image and attitude parameter, the historical relic image high density is of the same name in step 5 Point carries out point off density reconstruction and the historical relic image edge same place carries out point off density reconstruction and edge reconstruction respectively, in point off density Marginal point cloud corresponding to fusion, obtains the historical relic image fine-point cloud on the basis of cloud.
- 5. the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint as claimed in claim 1, its It is characterised by, the step 2, specifically comprises the following steps：2.1) homotopy mapping is carried out between the sequential images and then passes sequentially through two-way consistency constraint, RANSAC with Machine sampling consistency constraint, affine transformation constraint is constrained item by item step by step, to improve the accuracy of same place, is obtained high-quality same Name point set；2.2) the high-quality point set of the same name is further accurately positioned using Least squares matching method, obtains the sequence Image orientation same place.
- 6. the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint as claimed in claim 1, its It is characterised by, after the step 1, also includes step B before step 2：First, camera calibration is carried out to the camera, obtains the internal reference element and distortion parameter of camera, wherein, the internal reference member Element is used to orient image, and the distortion parameter is used for image data distortion rectification；Afterwards, distortion rectification is carried out to the sequential images data, and processing is filtered to the image after distortion；Then enter again Enter step 2.
- 7. the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint as claimed in claim 6, its It is characterised by, in the step B, when carrying out the camera calibration, the image for calculating camera calibration parameter will be relative to mark Image series are shot at the different angle of fixed board, different distance.
- 8. the historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint as claimed in claim 1, its It is characterised by, in the step 6, the three-dimensional point cloud scans to obtain using Articulated arm scanners.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710793380.2A CN107767440B (en) | 2017-09-06 | 2017-09-06 | Cultural relic sequence image fine three-dimensional reconstruction method based on triangulation network interpolation and constraint |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710793380.2A CN107767440B (en) | 2017-09-06 | 2017-09-06 | Cultural relic sequence image fine three-dimensional reconstruction method based on triangulation network interpolation and constraint |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107767440A true CN107767440A (en) | 2018-03-06 |
| CN107767440B CN107767440B (en) | 2021-01-26 |
Family
ID=61264984
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710793380.2A Active CN107767440B (en) | 2017-09-06 | 2017-09-06 | Cultural relic sequence image fine three-dimensional reconstruction method based on triangulation network interpolation and constraint |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107767440B (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109118587A (en) * | 2018-07-14 | 2019-01-01 | 武汉华宇世纪科技发展有限公司 | The production method and device of digital orthoimage |
| CN109238173A (en) * | 2018-08-16 | 2019-01-18 | 中国矿业大学 | A kind of coal yard three-dimensional live reconfiguration system and coal amount Method of fast estimating |
| CN109945853A (en) * | 2019-03-26 | 2019-06-28 | 西安因诺航空科技有限公司 | A kind of geographical coordinate positioning system and method based on 3D point cloud Aerial Images |
| CN111311731A (en) * | 2020-01-23 | 2020-06-19 | 深圳市易尚展示股份有限公司 | Random gray level map generation method and device based on digital projection and computer equipment |
| CN111595299A (en) * | 2020-04-14 | 2020-08-28 | 武汉倍思凯尔信息技术有限公司 | Clothing information acquisition system based on photogrammetry |
| CN112419380A (en) * | 2020-11-25 | 2021-02-26 | 湖北工业大学 | High-precision registration method for static orbit satellite sequence images based on cloud mask |
| CN112837389A (en) * | 2021-02-01 | 2021-05-25 | 北京爱奇艺科技有限公司 | Object model construction method and device and electronic equipment |
| CN112929626A (en) * | 2021-02-02 | 2021-06-08 | 辽宁工程技术大学 | Three-dimensional information extraction method based on smartphone image |
| CN113379822A (en) * | 2020-03-16 | 2021-09-10 | 天目爱视（北京）科技有限公司 | Method for acquiring 3D information of target object based on pose information of acquisition equipment |
| CN113589340A (en) * | 2021-06-15 | 2021-11-02 | 北京道达天际科技有限公司 | High-precision positioning method and device for satellite images assisted by reference network |
| CN114266830A (en) * | 2021-12-28 | 2022-04-01 | 北京建筑大学 | Underground large-space high-precision positioning method |
| CN114353780A (en) * | 2021-12-31 | 2022-04-15 | 高德软件有限公司 | Attitude optimization method and equipment |
| CN114913246A (en) * | 2022-07-15 | 2022-08-16 | 齐鲁空天信息研究院 | Camera calibration method and device, electronic equipment and storage medium |
| CN115830246A (en) * | 2023-01-09 | 2023-03-21 | 中国地质大学（武汉） | Spherical panoramic image three-dimensional reconstruction method based on incremental SFM |
| CN117115336A (en) * | 2023-07-13 | 2023-11-24 | 中国工程物理研究院计算机应用研究所 | Point cloud reconstruction method based on remote sensing stereoscopic image |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2003088085A1 (en) * | 2002-04-04 | 2003-10-23 | Arizona Board Of Regents | Three-dimensional digital library system |
| CN102074052A (en) * | 2011-01-20 | 2011-05-25 | 山东理工大学 | Sampling point topological neighbor-based method for reconstructing surface topology of scattered point cloud |
| CN104952107A (en) * | 2015-05-18 | 2015-09-30 | 湖南桥康智能科技有限公司 | Three-dimensional bridge reconstruction method based on vehicle-mounted LiDAR point cloud data |
| FR3040798A1 (en) * | 2015-09-08 | 2017-03-10 | Safran | PLENOPTIC CAMERA |
-
2017
- 2017-09-06 CN CN201710793380.2A patent/CN107767440B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2003088085A1 (en) * | 2002-04-04 | 2003-10-23 | Arizona Board Of Regents | Three-dimensional digital library system |
| CN102074052A (en) * | 2011-01-20 | 2011-05-25 | 山东理工大学 | Sampling point topological neighbor-based method for reconstructing surface topology of scattered point cloud |
| CN104952107A (en) * | 2015-05-18 | 2015-09-30 | 湖南桥康智能科技有限公司 | Three-dimensional bridge reconstruction method based on vehicle-mounted LiDAR point cloud data |
| FR3040798A1 (en) * | 2015-09-08 | 2017-03-10 | Safran | PLENOPTIC CAMERA |
Non-Patent Citations (2)
| Title |
|---|
| 张莉等: "Forstner 算子及其改进", 《北京工业职业技术学院学报》 * |
| 董友强: "影像匹配与密集点云生成研究", 《中国优秀硕士学位论文全文数据库（电子期刊）基础科学辑》 * |
Cited By (23)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109118587A (en) * | 2018-07-14 | 2019-01-01 | 武汉华宇世纪科技发展有限公司 | The production method and device of digital orthoimage |
| CN109238173A (en) * | 2018-08-16 | 2019-01-18 | 中国矿业大学 | A kind of coal yard three-dimensional live reconfiguration system and coal amount Method of fast estimating |
| CN109238173B (en) * | 2018-08-16 | 2020-03-13 | 煤炭科学研究总院 | Three-dimensional live-action reconstruction system for coal storage yard and rapid coal quantity estimation method |
| CN109945853A (en) * | 2019-03-26 | 2019-06-28 | 西安因诺航空科技有限公司 | A kind of geographical coordinate positioning system and method based on 3D point cloud Aerial Images |
| CN109945853B (en) * | 2019-03-26 | 2023-08-15 | 西安因诺航空科技有限公司 | Geographic coordinate positioning system and method based on 3D point cloud aerial image |
| CN111311731B (en) * | 2020-01-23 | 2023-04-07 | 深圳市易尚展示股份有限公司 | Random gray level map generation method and device based on digital projection and computer equipment |
| CN111311731A (en) * | 2020-01-23 | 2020-06-19 | 深圳市易尚展示股份有限公司 | Random gray level map generation method and device based on digital projection and computer equipment |
| CN113379822A (en) * | 2020-03-16 | 2021-09-10 | 天目爱视（北京）科技有限公司 | Method for acquiring 3D information of target object based on pose information of acquisition equipment |
| CN113379822B (en) * | 2020-03-16 | 2024-03-22 | 天目爱视（北京）科技有限公司 | Method for acquiring 3D information of target object based on pose information of acquisition equipment |
| CN111595299A (en) * | 2020-04-14 | 2020-08-28 | 武汉倍思凯尔信息技术有限公司 | Clothing information acquisition system based on photogrammetry |
| CN112419380A (en) * | 2020-11-25 | 2021-02-26 | 湖北工业大学 | High-precision registration method for static orbit satellite sequence images based on cloud mask |
| CN112419380B (en) * | 2020-11-25 | 2023-08-15 | 湖北工业大学 | Cloud mask-based high-precision registration method for stationary orbit satellite sequence images |
| CN112837389A (en) * | 2021-02-01 | 2021-05-25 | 北京爱奇艺科技有限公司 | Object model construction method and device and electronic equipment |
| CN112929626A (en) * | 2021-02-02 | 2021-06-08 | 辽宁工程技术大学 | Three-dimensional information extraction method based on smartphone image |
| CN112929626B (en) * | 2021-02-02 | 2023-02-14 | 辽宁工程技术大学 | Three-dimensional information extraction method based on smartphone image |
| CN113589340A (en) * | 2021-06-15 | 2021-11-02 | 北京道达天际科技有限公司 | High-precision positioning method and device for satellite images assisted by reference network |
| CN114266830A (en) * | 2021-12-28 | 2022-04-01 | 北京建筑大学 | Underground large-space high-precision positioning method |
| CN114353780A (en) * | 2021-12-31 | 2022-04-15 | 高德软件有限公司 | Attitude optimization method and equipment |
| CN114353780B (en) * | 2021-12-31 | 2024-04-02 | 高德软件有限公司 | Gesture optimization method and device |
| CN114913246A (en) * | 2022-07-15 | 2022-08-16 | 齐鲁空天信息研究院 | Camera calibration method and device, electronic equipment and storage medium |
| CN115830246A (en) * | 2023-01-09 | 2023-03-21 | 中国地质大学（武汉） | Spherical panoramic image three-dimensional reconstruction method based on incremental SFM |
| CN115830246B (en) * | 2023-01-09 | 2023-04-28 | 中国地质大学（武汉） | Spherical panoramic image three-dimensional reconstruction method based on incremental SFM |
| CN117115336A (en) * | 2023-07-13 | 2023-11-24 | 中国工程物理研究院计算机应用研究所 | Point cloud reconstruction method based on remote sensing stereoscopic image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107767440B (en) | 2021-01-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107767440A (en) | Historical relic sequential images subtle three-dimensional method for reconstructing based on triangulation network interpolation and constraint | |
| CN110296691B (en) | IMU calibration-fused binocular stereo vision measurement method and system | |
| CN111242990B (en) | 360-degree three-dimensional reconstruction optimization method based on continuous phase dense matching | |
| CN104376552B (en) | A kind of virtual combat method of 3D models and two dimensional image | |
| Pepe et al. | Techniques, tools, platforms and algorithms in close range photogrammetry in building 3D model and 2D representation of objects and complex architectures | |
| CN108765298A (en) | Unmanned plane image split-joint method based on three-dimensional reconstruction and system | |
| CN104240289B (en) | Three-dimensional digitalization reconstruction method and system based on single camera | |
| Grussenmeyer et al. | Solutions for exterior orientation in photogrammetry: a review | |
| Barazzetti et al. | True-orthophoto generation from UAV images: Implementation of a combined photogrammetric and computer vision approach | |
| Moussa | Integration of digital photogrammetry and terrestrial laser scanning for cultural heritage data recording | |
| CN111060006A (en) | Viewpoint planning method based on three-dimensional model | |
| Moussa et al. | Automatic fusion of digital images and laser scanner data for heritage preservation | |
| Gong et al. | Point cloud and digital surface model generation from high resolution multiple view stereo satellite imagery | |
| CN108830921A (en) | Laser point cloud reflected intensity correcting method based on incident angle | |
| Previtali et al. | Rigorous procedure for mapping thermal infrared images on three-dimensional models of building façades | |
| Rodríguez‐Gonzálvez et al. | A hybrid approach to create an archaeological visualization system for a Palaeolithic cave | |
| CN104964669A (en) | Orthoimage generation method of cylinder-like antique object | |
| Wu | Photogrammetry: 3-D from imagery | |
| CN112819900B (en) | Method for calibrating internal azimuth, relative orientation and distortion coefficient of intelligent stereography | |
| Zhang et al. | Reconstruction of measurable three-dimensional point cloud model based on large-scene archaeological excavation sites | |
| CN114252058A (en) | Aviation fine large scene manufacturing and measuring method | |
| Yang et al. | Three-dimensional panoramic terrain reconstruction from aerial imagery | |
| Re et al. | Evaluation of an area-based matching algorithm with advanced shape models | |
| Oniga | Comparative study on methods for 3D modelling of urban areas | |
| Ye et al. | Lunar Terrain Reconstruction from Multi-View Lroc Nac Images Based on Semi-Global Matching in Object Space |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |