CN110399931A - A kind of fish eye images matching process and system - Google Patents
A kind of fish eye images matching process and system Download PDFInfo
- Publication number
- CN110399931A CN110399931A CN201910693447.4A CN201910693447A CN110399931A CN 110399931 A CN110399931 A CN 110399931A CN 201910693447 A CN201910693447 A CN 201910693447A CN 110399931 A CN110399931 A CN 110399931A
- Authority
- CN
- China
- Prior art keywords
- triangular mesh
- eye images
- fish eye
- matching
- flake
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Studio Devices (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of fish eye images matching process and systems.The described method includes: obtaining the fish eye images to be matched obtained using fisheye camera；It is imaged in half spherical model in flake and constructs inscribe triangular grid polyhedron；Using the deflection of triangular mesh normal and fish eye images normal as the longitude and latitude angle in ASIFT algorithm, the first fish eye images and the second fish eye images are matched using ASIFT algorithm, obtain flake matching figure；Triangular grid polyhedron is projected to respectively in the first flake matching image and the second flake matching image, triangular mesh flake matching image is obtained；The sparse matching result of fish eye images is obtained by the error matching points in triangular mesh flake matching image to deletion using the method for dynamic statistics.The present invention can fast and accurately realize that fish eye images match.
Description
Technical field
The present invention relates to technical field of image processing, more particularly to a kind of fish eye images matching process and system.
Background technique
Fisheye camera is widely used in the neck such as fixation and recognition, vision monitoring, robot vision, virtual reality in recent years
Domain, the main reason is that fish eye images have big wide-angle feature, visual angle can achieve 180 °, even greater advantage.But with
Common transmissive image is compared, and nonlinear distortion has also been introduced while obtaining ultra wide-angle in fish eye images, using SURF,
The matching algorithms such as DAISY can not match the effect obtained to fish eye images.
Currently, also proposed a lot of algorithms for the sparse matching of fish eye images, Hansen is on the basis of SIFT algorithm
The method for introducing hemisphere imaging model is improved, and two kinds of new descriptors based on spherical surface: parabolic are proposed
SIFT (P-SIFT) and spherical sift (S-SIFT).However, the two descriptors are only applicable to the lesser image that distorts.
If image fault is larger, it can only generally guarantee that central area matching is correct, need to carry out large-scale interpolation, will certainly reduce
Key point matching precision.Spherical diagram is attempted to use by the SIFT on the Sphere (SIFT-Sphere) that Cruz-Mota is proposed
The low deficiency of matching accuracy rate of the SIFT algorithm in distortion critical regions is filled up as model or specific non-linear distortion model.
But the key point of small scale cannot be detected in SIFT-Sphere, the key point for causing SIFT-Sphere to detect is less.
YUHAO SHAN proposes a kind of method for extracting feature descriptor from discrete spherical surface image using convolutional neural networks, this
Algorithm distorts large area on fish eye images and central area can accomplish the matching of high accuracy, but this algorithm
It is too slow with speed, and this algorithm cannot be removed effectively error hiding.In conclusion the sparse matching technique of present fish eye images is not
There is practical, fast accurate a solution.
It would therefore be highly desirable to which one kind should guarantee matching precision, guarantee the sparse match party of fish eye images of matching rapidity again
Method.
Summary of the invention
Based on this, it is necessary to a kind of fish eye images matching process and system are provided, fast and accurately to realize fish eye images
Matching.
To achieve the above object, the present invention provides following schemes:
A kind of fish eye images matching process, comprising:
Obtain fish eye images to be matched；The fish eye images to be matched shoot to obtain using fisheye camera；It is described to
Matching fish eye images includes the first fish eye images and the second fish eye images；
It is imaged in half spherical model in flake and constructs inscribe triangular grid polyhedron；The triangular grid polyhedron is by multiple
What triangular mesh was constituted；
Feature point extraction is carried out to the fish eye images to be matched, obtains fish eye images characteristic point；
By on the fish eye images projecting characteristic points to the triangular grid polyhedron, and by triangular mesh normal and fish
The deflection of eye image normal is as the longitude and latitude angle in ASIFT algorithm, using ASIFT algorithm to first fish eye images and institute
It states the second fish eye images to be matched, obtains flake matching figure；
The triangular grid polyhedron is projected to respectively in the first flake matching image and the second flake matching image, is obtained
To triangular mesh flake matching image；The triangular mesh flake matching image is the flake after being covered by triangular mesh
Image；The triangular mesh is that the triangular grid polyhedric projection obtains；The triangular mesh flake matching image
Including the first triangular mesh fish eye images and the second triangular mesh fish eye images；The first triangular mesh fish eye images
Corresponding first fish eye images in figure are matched for the flake；The second triangular mesh fish eye images are flake matching
Corresponding second fish eye images in figure；
Using the method for dynamic statistics, by the error matching points in the triangular mesh flake matching image to deletion,
Obtain the sparse matching result of fish eye images；The error matching points are to the match point for wrong triangular mesh centering；The mistake
Accidentally triangular mesh is to the triangular mesh pair for being less than preset threshold for match point number in triangular mesh.
Optionally, described by the fish eye images projecting characteristic points to the triangular grid polyhedron, and by triangle
The deflection of grid normal and fish eye images normal is as the longitude and latitude angle in ASIFT algorithm, using ASIFT algorithm to described first
Fish eye images and second fish eye images are matched, and are obtained flake matching figure, are specifically included:
By on the fish eye images projecting characteristic points to the triangular grid polyhedron, by the triangular mesh normal with
The inclination angle of fish eye images normal is determined as angle of latitude, and the rotation angle of the triangular mesh normal and fish eye images normal determines
For longitude angle；
Angle is rotated according to the angle of latitude, the longitude angle, fisheye camera focal length and fisheye camera, constructs affine transformation
Matrix；The affine transformation matrix
Wherein, λ indicates that fisheye camera focal length, λ > 0, γ indicate that fisheye camera rotates angle, and θ indicates that angle of latitude, φ are warp
Spend angle；
According to the affine transformation matrix, the triangular mesh fish eye images are matched using ASIFT algorithm, are obtained
It matches and schemes to flake.
Optionally, the method using dynamic statistics, by the mistake in the triangular mesh flake matching image
With point to deletion, the sparse matching result of fish eye images is obtained, is specifically included:
Inside the triangular mesh flake matching image intermediate cam shape grid is determined using the method for dynamic statistics
The number of matching internal point is compared by the number with point with preset threshold, deletes the triangular mesh flake matching figure
Error matching points pair as in, obtain the first matching result；
The corresponding triangular mesh of the first triangular mesh fish eye images is rotated into predetermined angle, is obtained postrotational
First triangular mesh fish eye images, so that falling into triangular mesh in the match point on triangular mesh edge in rotation anteposition
It is internal；
The corresponding triangular mesh of postrotational first triangular mesh fish eye images is determined using the method for dynamic statistics
The number of match point inside flake matching image intermediate cam shape grid, by the number of postrotational matching internal point and default threshold
Value is compared, and is deleted the error matching points pair in the triangular mesh flake matching image, is obtained the second matching result；
First matching result and second matching result are merged, the sparse matching result of fish eye images is obtained.
Optionally, the method using dynamic statistics determines the triangular mesh flake matching image intermediate cam shape net
The number of matching internal point is compared with preset threshold, deletes the triangular mesh by the number of the match point inside lattice
The error matching points pair of flake matching image obtain the first matching result, specifically include:
Determine preferred triangular mesh；The preferred triangular mesh is in the first triangular mesh fish eye images
It is greater than or equal to the triangular mesh of the first preset threshold with number；
Determine the first corresponding region；First corresponding region is first in the second triangular mesh fish eye images
Corresponding triangular mesh；The first corresponding triangular mesh be in the second triangular mesh fish eye images with it is described preferably
The corresponding triangular mesh most comprising match point number of triangular mesh；
If the number of the matching double points in the preferred triangular mesh and first corresponding region is greater than or equal to the
The preferred triangular mesh and first corresponding region are then determined as the first matching result by two preset thresholds.
Optionally, the method using dynamic statistics determines that postrotational first triangular mesh fish eye images are corresponding
The number of match point inside triangular mesh flake matching image intermediate cam shape grid, by the number of postrotational matching internal point
Mesh is compared with preset threshold, is deleted the error matching points pair in the triangular mesh flake matching image, is obtained second
Matching result specifically includes:
Determine preferred triangular mesh after rotating；Preferred triangular mesh is postrotational first triangle after the rotation
Match point number is greater than or equal to the triangular mesh of first preset threshold in grid fish eye images；
Determine the second corresponding region；Second corresponding region is second in the second triangular mesh fish eye images
Corresponding triangular mesh；The second corresponding triangular mesh be in the second triangular mesh fish eye images with the rotation
The corresponding triangular mesh most comprising match point number of preferred triangular mesh afterwards；
If the number of the matching double points after the rotation in preferred triangular mesh and second corresponding region be greater than or
Equal to the second preset threshold, then triangular mesh preferred after the rotation and second corresponding region are determined as the second matching
As a result.
The present invention also provides a kind of fish eye images matching systems, comprising:
Image collection module, for obtaining fish eye images to be matched；The fish eye images to be matched are using fisheye camera
What shooting obtained；The fish eye images to be matched include the first fish eye images and the second fish eye images；
Polyhedron constructs module, constructs inscribe triangular grid polyhedron for being imaged in half spherical model in flake；Described three
Angle grid polyhedron is made of multiple triangular mesh；
Feature point extraction module obtains fish eye images spy for carrying out feature point extraction to the fish eye images to be matched
Sign point；
Matching module, for by the fish eye images projecting characteristic points to the triangular grid polyhedron, and by triangle
The deflection of shape grid normal and fish eye images normal is as the longitude and latitude angle in ASIFT algorithm, using ASIFT algorithm to described the
One fish eye images and second fish eye images are matched, and flake matching figure is obtained；
Projection module, for the triangular grid polyhedron to be projected to the first flake matching image and the second flake respectively
In matching image, triangular mesh flake matching image is obtained；The triangular mesh flake matching image is by network of triangle
Fish eye images after lattice covering；The triangular mesh is that the triangular grid polyhedric projection obtains；The network of triangle
Lattice flake matching image includes the first triangular mesh fish eye images and the second triangular mesh fish eye images；First triangle
Shape grid fish eye images are that the flake matches corresponding first fish eye images in figure；The second triangular mesh fish eye images
Corresponding second fish eye images in figure are matched for the flake；
Match point rejects module, will be in the triangular mesh flake matching image for the method using dynamic statistics
Error matching points to deletion, obtain the sparse matching result of fish eye images；The error matching points are to for wrong triangular mesh
The match point of centering；The mistake triangular mesh is to the triangle for being less than preset threshold for match point number in triangular mesh
Grid pair.
Optionally, the matching module, specifically includes:
Angle of latitude determination unit is used for the fish eye images projecting characteristic points to the triangular grid polyhedron, will
The inclination angle of the triangular mesh normal and fish eye images normal is determined as angle of latitude, the triangular mesh normal and flake
The rotation angle of image normal is determined as longitude angle；
Matrix construction unit, for being rotated according to the angle of latitude, the longitude angle, fisheye camera focal length and fisheye camera
Angle constructs affine transformation matrix；The affine transformation matrix
Wherein, λ indicates that fisheye camera focal length, λ > 0, γ indicate that fisheye camera rotates angle, and θ indicates that angle of latitude, φ are warp
Spend angle；
Matching unit is used for according to the affine transformation matrix, using ASIFT algorithm to the triangular mesh fish-eye image
As being matched, flake matching figure is obtained.
Optionally, the match point rejects module, specifically includes:
First match point culling unit, for determining the triangular mesh flake matching figure using the method for dynamic statistics
As the number of the match point inside intermediate cam shape grid, the number of matching internal point is compared with preset threshold, deletes institute
The error matching points pair in triangular mesh flake matching image are stated, the first matching result is obtained；
Grid rotary unit, it is default for rotating the corresponding triangular mesh of the first triangular mesh fish eye images
Angle obtains postrotational first triangular mesh fish eye images, so that in rotation anteposition on triangular mesh edge
It is fallen into inside triangular mesh with point；
Second match point culling unit, for determining postrotational first triangular mesh fish using the method for dynamic statistics
The number of match point inside the corresponding triangular mesh flake matching image intermediate cam shape grid of eye image, will be postrotational interior
The number of portion's match point is compared with preset threshold, deletes the error matching points in the triangular mesh flake matching image
It is right, obtain the second matching result；
It is dilute to obtain fish eye images for merging first matching result and second matching result for integrated unit
Dredge matching result.
Optionally, the first match point culling unit, specifically includes:
First determines subelement, for determining preferred triangular mesh；The preferred triangular mesh is the described 1st
Match point number is greater than or equal to the triangular mesh of the first preset threshold in hexagonal lattice fish eye images；
Second determines subelement, for determining the first corresponding region；First corresponding region is second triangle
The first corresponding triangular mesh in grid fish eye images；Described first corresponding triangular mesh is second triangular mesh
The triangular mesh most comprising match point number corresponding with the preferred triangular mesh in fish eye images；
Third determines subelement, if for the matching double points in the preferred triangular mesh and first corresponding region
Number be greater than or equal to the second preset threshold, then the preferred triangular mesh and first corresponding region are determined as the
One matching result.
Optionally, the second match point culling unit, specifically includes:
4th determines subelement, for preferred triangular mesh after determining rotation；Preferred triangular mesh after the rotation
It is greater than or equal to the triangle of first preset threshold for match point number in postrotational first triangular mesh fish eye images
Shape grid；
5th determines subelement, for determining the second corresponding region；Second corresponding region is second triangle
The second corresponding triangular mesh in grid fish eye images；Described second corresponding triangular mesh is second triangular mesh
The triangular mesh most comprising match point number corresponding with triangular mesh preferred after the rotation in fish eye images；
6th determines subelement, if in triangular mesh preferred after the rotation and second corresponding region
Number with point pair is greater than or equal to the second preset threshold, then triangular mesh preferred after the rotation is corresponding with described second
Region is determined as the second matching result.
Compared with prior art, the beneficial effects of the present invention are:
The invention proposes a kind of fish eye images matching process and systems, are based on triangular grid structure, and triangular grid is more
The normal of face body intermediate cam shape and the deflection of fish eye images normal are as longitude and latitude angle, using ASIFT algorithm to triangular mesh
Fish eye images are matched, and flake matching figure is obtained, and are optimized the feature description of ASIFT, are improved matching speed and matching essence
Degree；Using the method for dynamic statistics, flake is matched into the error matching points in figure to deletion, finally obtains fish eye images sparse
With as a result, wherein error matching points are to the match point for wrong triangular mesh centering, mistake triangular mesh is to for triangle
Match point number is less than the triangular mesh pair of preset threshold in grid, in this way by the triangle gridding where two match points
The threshold value of centering setting coupling number is true or false come the matching determined in triangle gridding, passes through of statistical method removal mistake
Match, further improves matching speed and matching precision.
Detailed description of the invention
It in order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, below will be to institute in embodiment
Attached drawing to be used is needed to be briefly described, it should be apparent that, the accompanying drawings in the following description is only some implementations of the invention
Example, for those of ordinary skill in the art, without any creative labor, can also be according to these attached drawings
Obtain other attached drawings.
Fig. 1 is a kind of flow chart of fish eye images matching process of the embodiment of the present invention；
Fig. 2 is fish eye images of embodiment of the present invention imaging schematic diagram；
Fig. 3 is the structural schematic diagram of triangular grid of embodiment of the present invention polyhedron and polyhedral triangle normal；
Fig. 4 is the structural schematic diagram of triangular mesh of the embodiment of the present invention；
Fig. 5 is fish eye images of embodiment of the present invention left image；
Fig. 6 is fish eye images of embodiment of the present invention right image；
Fig. 7 is the sparse matching result figure of fish eye images of the embodiment of the present invention；
Fig. 8 is a kind of structural schematic diagram of fish eye images matching system of the embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
In order to make the foregoing objectives, features and advantages of the present invention clearer and more comprehensible, with reference to the accompanying drawing and specific real
Applying mode, the present invention is described in further detail.
Fig. 1 is a kind of flow chart of fish eye images matching process of the embodiment of the present invention.
Referring to Fig. 1, the fish eye images matching process of embodiment, comprising:
Step S1: fish eye images to be matched are obtained；The fish eye images to be matched shoot to obtain using fisheye camera.
The fish eye images to be matched include the first fish eye images and the second fish eye images.In the present embodiment, the first flake
Image is left image, and the second fish eye images are right image.Since fish eye images to be matched are there are nonlinear distortion, it connects down
To use the inscribe triangular grid polyhedron of half spherical model of imaging to be fitted half spherical model.
Step S2: it is imaged in half spherical model in flake and constructs inscribe triangular grid polyhedron.
The flake imaging model is as shown in Fig. 2, flake imaging model multinomial is
R (θ)=k1θ+k2θ3+k3θ5+k4θ7, wherein r (θ) be on fish eye images point to world coordinate system origin away from
From；k1、k2、k3、k4For flake distortion parameter；θ is indicated in 1 spherical model upslide shadow point of flake angle corresponding with z-axis.
In the present embodiment, triangular grid polyhedron is positive the half of 320 face bodies, that is, the inscribe of half spherical model of flake is more
The face face Ti Wei160 body, and each face is isosceles triangle, apex angle is 72 °, and triangular grid polyhedron is as shown in Figure 3.
Step S3: feature point extraction is carried out to the fish eye images to be matched, obtains fish eye images characteristic point.
Step S4: by the fish eye images projecting characteristic points to the triangular grid polyhedron, and by triangular mesh
The deflection of normal and fish eye images normal is as the longitude and latitude angle in ASIFT algorithm, using ASIFT algorithm to first flake
Image and second fish eye images are matched, and flake matching figure is obtained.
The step S4 is specifically included:
41) by the fish eye images projecting characteristic points to the triangular grid polyhedron, by the triangular mesh method
The inclination angle of line and fish eye images normal is determined as angle of latitude, the rotation angle of the triangular mesh normal and fish eye images normal
It is determined as longitude angle.A in Fig. 3 indicates triangular mesh normal, ZcIndicate fish eye images normal.
42) due to needing to simulate two parameters of affine matrix when ASIFT algorithm is matched to fault image, foundation
The angle of latitude, the longitude angle, fisheye camera focal length and fisheye camera rotate angle, construct affine transformation matrix；It is described imitative
Penetrate transformation matrix
Wherein, λ indicates that fisheye camera focal length, λ > 0, γ indicate that fisheye camera rotates angle, and θ indicates that angle of latitude, φ are warp
Spend angle.Angle of latitude, longitude angle.
43) according to the affine transformation matrix, using ASIFT algorithm to triangular mesh fish eye images progress
Match, obtains flake matching figure.Specifically, being matched using ASIFT to polyhedral each delta-shaped region, each triangle
The normal of grid and the inclination angle of image normal are exactly the angle of latitude in this triangle ASIFT, the method for each triangular mesh
The rotation angle of line and image normal is exactly the longitude angle in this triangle ASIFT, and ASIFT thus can be used and complete flake
Image Feature Matching.
Step S5: the triangular grid polyhedron is projected into the first flake matching image and the second flake matching figure respectively
As in, triangular mesh flake matching image is obtained；The triangular mesh flake matching image includes the first triangular mesh
Fish eye images and the second triangular mesh fish eye images.
The triangular mesh flake matching image is the fish eye images after being covered by triangular mesh；The network of triangle
Lattice are that the triangular grid polyhedric projection obtains；The first triangular mesh fish eye images are that the flake matches in figure
Corresponding first fish eye images；The second triangular mesh fish eye images are that the flake matches corresponding second flake in figure
Image.
The step constructs dynamic statistics triangle gridding, the network of triangle in flake matching image in flake matching image
Lattice are obtained on the triangular grid polyhedric projection to fish eye images being imaged in half spherical model by flake, will in the present embodiment
160 face bodies in step S2 project to an available triangular on fish eye images at grid, inside triangular mesh
It include 160 isosceles triangles, remoter apart from fish eye images center its apex angle of triangle is bigger in triangular mesh, place
The fish eye images distortion in region is bigger.Triangular mesh is as shown in Figure 4.
Step S6: using the method for dynamic statistics, by the error matching points in the triangular mesh flake matching image
To deletion, the sparse matching result of fish eye images is obtained；The error matching points are to the match point for wrong triangular mesh centering；
The mistake triangular mesh is to the triangular mesh pair for being less than preset threshold for match point number in triangular mesh.
The step S6, specifically includes:
1) number of the match point inside the flake matching figure intermediate cam shape grid is determined using the method for dynamic statistics,
The number of matching internal point is compared with preset threshold, the error matching points pair in the flake matching figure is deleted, obtains
First matching result.It is specific:
Determine preferred triangular mesh；The preferred triangular mesh is that the first triangular mesh fish eye images are (left
Image) in match point number be greater than or equal to the first preset threshold triangular mesh.Match point number is less than the first default threshold
Value, illustrates that data volume is too small with a low credibility, does not handle these regions.
Determine the first corresponding region；First corresponding region is the second triangular mesh fish eye images (right image)
In the first corresponding triangular mesh；The first corresponding triangular mesh be in the second triangular mesh fish eye images with
The corresponding triangular mesh most comprising match point number of the preferred triangular mesh.
If the number of the matching double points in the preferred triangular mesh and first corresponding region is greater than or equal to the
The preferred triangular mesh and first corresponding region are then determined as the first matching result by two preset thresholds.
Above-mentioned steps 1) it is the removal that error matching points are carried out to the match point inside triangular mesh, and for triangle
The problem of grid edge match point divides, proposes and revolves the corresponding triangular mesh of any piece image in two images
The thinking turned, so that the match point on triangular mesh edge is fallen in inside triangular mesh after rotation, to solve Grid Edge
Edge match point partition problem.In the present embodiment, by the first triangular mesh fish eye images (left figure in the flake matching image
Picture) it is 36 ° of triangular mesh rotation corresponding, postrotational first triangular mesh fish eye images are obtained, so that in rotation anteposition
The central area of triangular mesh is fallen into the match point on triangular mesh edge.To the match point on triangular mesh edge
Shown in the concrete mode such as step 2) for carrying out the removal of error matching points.
2) the corresponding flake matching of postrotational first triangular mesh fish eye images is determined using the method for dynamic statistics
The number for scheming the match point inside (left image) intermediate cam shape grid, by the number and preset threshold of postrotational matching internal point
It is compared, deletes the error matching points pair in the flake matching figure, obtain the second matching result.It is specific:
Determine preferred triangular mesh after rotating；Preferred triangular mesh is postrotational first triangle after the rotation
Match point number is greater than or equal to the triangular mesh of first preset threshold in grid fish eye images.
Determine the second corresponding region；Second corresponding region is the second triangular mesh fish eye images (right image)
In the second corresponding triangular mesh；The second corresponding triangular mesh be in the second triangular mesh fish eye images with
The corresponding triangular mesh most comprising match point number of preferred triangular mesh after the rotation.
If the number of the matching double points after the rotation in preferred triangular mesh and second corresponding region be greater than or
Equal to the second preset threshold, then triangular mesh preferred after the rotation and second corresponding region are determined as the second matching
As a result.
3) first matching result and second matching result are merged, obtains the sparse matching result of fish eye images.
It adopts and the fish eye images matching process of above-described embodiment is verified below experimentally.
1) fish eye images to be matched are obtained using fisheye camera, wherein Fig. 5 is fish eye images left image, and Fig. 6 is flake
Image right image.
2) characteristic point is extracted on fish eye images by FAST algorithm, and half spherical model internal structure inscribe is imaged in flake
Triangular grid polyhedron, by the triangle on the polyhedral Grid Projection to fish eye images of triangular grid, and in use grid
Normal direction angle carries out Feature Points Matching as the longitude and latitude angle in ASIFT, then by ASIFT algorithm.
3) identical triangular mesh is constructed to left image right image, next counts each triangle in left image
It include match point number in region, if delta-shaped region matching internal point illustrates that data volume is too small with a low credibility less than threshold value,
These regions are not handled.
4) match point for meeting the delta-shaped region of threshold value to left image counts, and counts these match points in right image
Match point number in the most delta-shaped region of middle distribution, match point number is more than in the grid pair where two match points
The threshold value of coupling number is set, then matching is correct, retains matching result.
5) grid edge match point partition problem is treated, we use two kinds of grids to right image, the first is and left figure
As identical grid division, the grid division of left image is rotated into 36 ° of second of grid divisions as right image, then to a left side
The match point of image and the delta-shaped region of the right image with second of grid is counted, and counts these match points in right figure
The match point number being distributed in most delta-shaped regions as in, match point number is super in the grid pair where two match points
The threshold value of setting coupling number is crossed, then matching is correct, retains matching result.
6) removal for finally matching result that the matching number counted in two kinds of grids meets threshold value being merged to the end
The sparse matching result of the fish eye images of error hiding.The sparse matching result of fish eye images as shown in fig. 7,, can by intuitively observing
To find out that the fish eye images matching process of the present embodiment can accomplish matching efficiency height to the sparse matching of fish eye images, removal is missed
The features such as matching speed is fast.
The fish eye images characteristic point that the fish eye images matching process of the present embodiment has high robust, quickly removes error hiding
The advantages of.The Feature Points Matching speed that this method solve fish eye images in distortion is slow, poor robustness problem；By drawing
Enter triangular grid polyhedron and solve ASIFT to carry out sparse matching problem on fish eye images；By introducing triangular grid dynamic
It is slow-footed that statistical method solves the problems, such as that fish eye images reject error hiding.
Fig. 8 is a kind of structural schematic diagram of fish eye images matching system of the embodiment of the present invention.
Referring to Fig. 8, the fish eye images matching system of embodiment, comprising:
Image collection module 801, for obtaining fish eye images to be matched；The fish eye images to be matched are using flake phase
What machine was shot；The fish eye images to be matched include the first fish eye images and the second fish eye images.
Polyhedron constructs module 802, constructs inscribe triangular grid polyhedron for being imaged in half spherical model in flake；It is described
Triangular grid polyhedron is made of multiple triangular mesh.
Feature point extraction module 803 obtains fish eye images for carrying out feature point extraction to the fish eye images to be matched
Characteristic point.
Matching module 804, for by the fish eye images projecting characteristic points to the triangular grid polyhedron, and by three
The deflection of angular grid normal and fish eye images normal is as the longitude and latitude angle in ASIFT algorithm, using ASIFT algorithm to described
First fish eye images and second fish eye images are matched, and flake matching figure is obtained.
Projection module 805, for the triangular grid polyhedron to be projected to the first flake matching image and second respectively
In flake matching image, triangular mesh flake matching image is obtained；The triangular mesh flake matching image is by triangle
Fish eye images after the covering of shape grid；The triangular mesh is that the triangular grid polyhedric projection obtains；The triangle
Shape grid flake matching image includes the first triangular mesh fish eye images and the second triangular mesh fish eye images；Described first
Triangular mesh fish eye images are that the flake matches corresponding first fish eye images in figure；The second triangular mesh flake
Image is that the flake matches corresponding second fish eye images in figure.
Match point rejects module 806, for the method using dynamic statistics, by the triangular mesh flake matching image
In error matching points to deletion, obtain the sparse matching result of fish eye images；The error matching points are to for wrong network of triangle
The match point of lattice centering；The mistake triangular mesh is to the triangle for being less than preset threshold for match point number in triangular mesh
Shape grid pair.
As an alternative embodiment, the matching module 804, specifically includes:
Angle of latitude determination unit is used for the fish eye images projecting characteristic points to the triangular grid polyhedron, will
The inclination angle of the triangular mesh normal and fish eye images normal is determined as angle of latitude, the triangular mesh normal and flake
The rotation angle of image normal is determined as longitude angle.
Matrix construction unit, for being rotated according to the angle of latitude, the longitude angle, fisheye camera focal length and fisheye camera
Angle constructs affine transformation matrix；The affine transformation matrix
Wherein, λ indicates that fisheye camera focal length, λ > 0, γ indicate that fisheye camera rotates angle, and θ indicates that angle of latitude, φ are warp
Spend angle.
Matching unit is used for according to the affine transformation matrix, using ASIFT algorithm to the triangular mesh fish-eye image
As being matched, flake matching figure is obtained.
As an alternative embodiment, the match point rejects module 806, specifically include:
First match point culling unit, for determining the triangular mesh flake matching figure using the method for dynamic statistics
As the number of the match point inside intermediate cam shape grid, the number of matching internal point is compared with preset threshold, deletes institute
The error matching points pair in triangular mesh flake matching image are stated, the first matching result is obtained.
Grid rotary unit, it is default for rotating the corresponding triangular mesh of the first triangular mesh fish eye images
Angle obtains postrotational first triangular mesh fish eye images, so that in rotation anteposition on triangular mesh edge
It is fallen into inside triangular mesh with point.
Second match point culling unit, for determining postrotational first triangular mesh fish using the method for dynamic statistics
The number of match point inside the corresponding triangular mesh flake matching image intermediate cam shape grid of eye image, will be postrotational interior
The number of portion's match point is compared with preset threshold, deletes the error matching points in the triangular mesh flake matching image
It is right, obtain the second matching result.
It is dilute to obtain fish eye images for merging first matching result and second matching result for integrated unit
Dredge matching result.
As an alternative embodiment, the first match point culling unit, specifically includes:
First determines subelement, for determining preferred triangular mesh；The preferred triangular mesh is the described 1st
Match point number is greater than or equal to the triangular mesh of the first preset threshold in hexagonal lattice fish eye images.
Second determines subelement, for determining the first corresponding region；First corresponding region is second triangle
The first corresponding triangular mesh in grid fish eye images；Described first corresponding triangular mesh is second triangular mesh
The triangular mesh most comprising match point number corresponding with the preferred triangular mesh in fish eye images.
Third determines subelement, if for the matching double points in the preferred triangular mesh and first corresponding region
Number be greater than or equal to the second preset threshold, then the preferred triangular mesh and first corresponding region are determined as the
One matching result.
As an alternative embodiment, the second match point culling unit, specifically includes:
4th determines subelement, for preferred triangular mesh after determining rotation；Preferred triangular mesh after the rotation
It is greater than or equal to the triangle of first preset threshold for match point number in postrotational first triangular mesh fish eye images
Shape grid；
5th determines subelement, for determining the second corresponding region；Second corresponding region is second triangle
The second corresponding triangular mesh in grid fish eye images；Described second corresponding triangular mesh is second triangular mesh
The triangular mesh most comprising match point number corresponding with triangular mesh preferred after the rotation in fish eye images；
6th determines subelement, if in triangular mesh preferred after the rotation and second corresponding region
Number with point pair is greater than or equal to the second preset threshold, then triangular mesh preferred after the rotation is corresponding with described second
Region is determined as the second matching result.
The fish eye images matching system of the present embodiment realizes the quick and precisely matching of fish eye images.
For the system disclosed in the embodiment, since it is corresponded to the methods disclosed in the examples, so the ratio of description
Relatively simple, reference may be made to the description of the method.
Used herein a specific example illustrates the principle and implementation of the invention, and above embodiments are said
It is bright to be merely used to help understand method and its core concept of the invention；At the same time, for those skilled in the art, foundation
Thought of the invention, there will be changes in the specific implementation manner and application range.In conclusion the content of the present specification is not
It is interpreted as limitation of the present invention.
Claims (10)
1. a kind of fish eye images matching process characterized by comprising
Obtain fish eye images to be matched；The fish eye images to be matched shoot to obtain using fisheye camera；It is described to be matched
Fish eye images include the first fish eye images and the second fish eye images；
It is imaged in half spherical model in flake and constructs inscribe triangular grid polyhedron；The triangular grid polyhedron is by multiple triangles
What shape grid was constituted；
Feature point extraction is carried out to the fish eye images to be matched, obtains fish eye images characteristic point；
By on the fish eye images projecting characteristic points to the triangular grid polyhedron, and by triangular mesh normal and fish-eye image
As the deflection of normal is as the longitude and latitude angle in ASIFT algorithm, using ASIFT algorithm to first fish eye images and described
Two fish eye images are matched, and flake matching figure is obtained；
The triangular grid polyhedron is projected to respectively in the first flake matching image and the second flake matching image, obtains three
Hexagonal lattice flake matching image；The triangular mesh flake matching image is the fish-eye image after being covered by triangular mesh
Picture；The triangular mesh is that the triangular grid polyhedric projection obtains；The triangular mesh flake matching image packet
Include the first triangular mesh fish eye images and the second triangular mesh fish eye images；The first triangular mesh fish eye images are
Corresponding first fish eye images in the flake matching figure；The second triangular mesh fish eye images are flake matching figure
In corresponding second fish eye images；
It is obtained using the method for dynamic statistics by the error matching points in the triangular mesh flake matching image to deletion
The sparse matching result of fish eye images；The error matching points are to the match point for wrong triangular mesh centering；Described wrong three
Hexagonal lattice is to the triangular mesh pair for being less than preset threshold for match point number in triangular mesh.
2. a kind of fish eye images matching process according to claim 1, which is characterized in that described that the fish eye images are special
Sign point projects on the triangular grid polyhedron, and using the deflection of triangular mesh normal and fish eye images normal as
Longitude and latitude angle in ASIFT algorithm, using ASIFT algorithm to first fish eye images and second fish eye images progress
Match, obtain flake matching figure, specifically include:
By on the fish eye images projecting characteristic points to the triangular grid polyhedron, by the triangular mesh normal and flake
The inclination angle of image normal is determined as angle of latitude, the rotation angle of the triangular mesh normal and fish eye images normal be determined as through
Spend angle；
Angle is rotated according to the angle of latitude, the longitude angle, fisheye camera focal length and fisheye camera, constructs affine transformation square
Battle array；The affine transformation matrix
Wherein, λ indicates that fisheye camera focal length, λ > 0, γ indicate that fisheye camera rotates angle, and θ indicates angle of latitude, and φ is longitude
Angle；
According to the affine transformation matrix, the triangular mesh fish eye images are matched using ASIFT algorithm, obtain fish
Eye matching figure.
3. a kind of fish eye images matching process according to claim 1, which is characterized in that the side using dynamic statistics
Method obtains the sparse matching result of fish eye images by the error matching points in the triangular mesh flake matching image to deletion,
It specifically includes:
Match point inside the triangular mesh flake matching image intermediate cam shape grid is determined using the method for dynamic statistics
Number, the number of matching internal point is compared with preset threshold, is deleted in the triangular mesh flake matching image
Error matching points pair, obtain the first matching result；
The corresponding triangular mesh of the first triangular mesh fish eye images is rotated into predetermined angle, obtains postrotational first
Triangular mesh fish eye images, so that being fallen into triangular mesh in rotation anteposition in the match point on triangular mesh edge
Portion；
The corresponding triangular mesh flake of postrotational first triangular mesh fish eye images is determined using the method for dynamic statistics
The number of match point inside matching image intermediate cam shape grid, by the number of postrotational matching internal point and preset threshold into
Row compares, and deletes the error matching points pair in the triangular mesh flake matching image, obtains the second matching result；
First matching result and second matching result are merged, the sparse matching result of fish eye images is obtained.
4. a kind of fish eye images matching process according to claim 3, which is characterized in that the side using dynamic statistics
Method determines the number of the match point inside the triangular mesh flake matching image intermediate cam shape grid, by matching internal point
Number is compared with preset threshold, is deleted the error matching points pair of the triangular mesh flake matching image, is obtained first
Matching result specifically includes:
Determine preferred triangular mesh；The preferred triangular mesh is match point in the first triangular mesh fish eye images
Number is greater than or equal to the triangular mesh of the first preset threshold；
Determine the first corresponding region；First corresponding region is corresponding for first in the second triangular mesh fish eye images
Triangular mesh；The first corresponding triangular mesh be in the second triangular mesh fish eye images with the preferred triangle
The corresponding triangular mesh most comprising match point number of shape grid；
If it is pre- that the number of the matching double points in the preferred triangular mesh and first corresponding region is greater than or equal to second
If threshold value, then the preferred triangular mesh and first corresponding region are determined as the first matching result.
5. a kind of fish eye images matching process according to claim 3, which is characterized in that the side using dynamic statistics
Method determines the corresponding triangular mesh flake matching image intermediate cam shape grid of postrotational first triangular mesh fish eye images
The number of internal match point, the number of postrotational matching internal point is compared with preset threshold, deletes the triangle
Error matching points pair in shape grid flake matching image, obtain the second matching result, specifically include:
Determine preferred triangular mesh after rotating；Preferred triangular mesh is postrotational first triangular mesh after the rotation
Match point number is greater than or equal to the triangular mesh of first preset threshold in fish eye images；
Determine the second corresponding region；Second corresponding region is corresponding for second in the second triangular mesh fish eye images
Triangular mesh；The second corresponding triangular mesh be in the second triangular mesh fish eye images with it is excellent after the rotation
Selecting triangular mesh corresponding includes the most triangular mesh of match point number；
If the number of preferred triangular mesh and the matching double points in second corresponding region is greater than or equal to after the rotation
Triangular mesh preferred after the rotation and second corresponding region are then determined as the second matching knot by the second preset threshold
Fruit.
6. a kind of fish eye images matching system characterized by comprising
Image collection module, for obtaining fish eye images to be matched；The fish eye images to be matched are shot using fisheye camera
It obtains；The fish eye images to be matched include the first fish eye images and the second fish eye images；
Polyhedron constructs module, constructs inscribe triangular grid polyhedron for being imaged in half spherical model in flake；The trigonal lattice
Net polyhedron is made of multiple triangular mesh；
Feature point extraction module obtains fish eye images characteristic point for carrying out feature point extraction to the fish eye images to be matched；
Matching module, for by the fish eye images projecting characteristic points to the triangular grid polyhedron, and by triangle lattice
The deflection of net normal and fish eye images normal is as the longitude and latitude angle in ASIFT algorithm, using ASIFT algorithm to first fish
Eye image and second fish eye images are matched, and flake matching figure is obtained；
Projection module, for the triangular grid polyhedron to be projected to the first flake matching image and the matching of the second flake respectively
In image, triangular mesh flake matching image is obtained；The triangular mesh flake matching image is to be covered by triangular mesh
Fish eye images after lid；The triangular mesh is that the triangular grid polyhedric projection obtains；The triangular mesh fish
Eye matching image includes the first triangular mesh fish eye images and the second triangular mesh fish eye images；First network of triangle
Lattice fish eye images are that the flake matches corresponding first fish eye images in figure；The second triangular mesh fish eye images are institute
State corresponding second fish eye images in flake matching figure；
Match point rejects module, for the method using dynamic statistics, by the mistake in the triangular mesh flake matching image
Mismatching point obtains the sparse matching result of fish eye images to deletion；The error matching points are to for wrong triangular mesh centering
Match point；The mistake triangular mesh is to the triangular mesh for being less than preset threshold for match point number in triangular mesh
It is right.
7. a kind of fish eye images matching system according to claim 6, which is characterized in that the matching module, it is specific to wrap
It includes:
Angle of latitude determination unit is used for the fish eye images projecting characteristic points to the triangular grid polyhedron, will be described
The inclination angle of triangular mesh normal and fish eye images normal is determined as angle of latitude, the triangular mesh normal and fish eye images
The rotation angle of normal is determined as longitude angle；
Matrix construction unit, for according to the angle of latitude, the longitude angle, fisheye camera focal length and fisheye camera rotation angle
Degree constructs affine transformation matrix；The affine transformation matrix
Wherein, λ indicates that fisheye camera focal length, λ > 0, γ indicate that fisheye camera rotates angle, and θ indicates angle of latitude, and φ is longitude
Angle；
Matching unit, for according to the affine transformation matrix, using ASIFT algorithm to the triangular mesh fish eye images into
Row matching, obtains flake matching figure.
8. a kind of fish eye images matching system according to claim 6, which is characterized in that the match point rejects module,
It specifically includes:
First match point culling unit, for being determined in the triangular mesh flake matching image using the method for dynamic statistics
The number of matching internal point is compared by the number of the match point inside triangular mesh with preset threshold, deletes described three
Error matching points pair in hexagonal lattice flake matching image, obtain the first matching result；
Grid rotary unit, for the corresponding triangular mesh of the first triangular mesh fish eye images to be rotated preset angle
Degree, obtains postrotational first triangular mesh fish eye images, so that in rotation anteposition in the matching on triangular mesh edge
Point is fallen into inside triangular mesh；
Second match point culling unit, for determining postrotational first triangular mesh fish-eye image using the method for dynamic statistics
As the number of the match point inside corresponding triangular mesh flake matching image intermediate cam shape grid, by postrotational inside
Number with point is compared with preset threshold, deletes the error matching points pair in the triangular mesh flake matching image,
Obtain the second matching result；
Integrated unit obtains fish eye images sparse for merging first matching result and second matching result
With result.
9. a kind of fish eye images matching system according to claim 8, which is characterized in that first match point is rejected single
Member specifically includes:
First determines subelement, for determining preferred triangular mesh；The preferred triangular mesh is first triangle
Match point number is greater than or equal to the triangular mesh of the first preset threshold in grid fish eye images；
Second determines subelement, for determining the first corresponding region；First corresponding region is second triangular mesh
The first corresponding triangular mesh in fish eye images；Described first corresponding triangular mesh is the second triangular mesh flake
The triangular mesh most comprising match point number corresponding with the preferred triangular mesh in image；
Third determines subelement, if the number for the matching double points in the preferred triangular mesh and first corresponding region
Mesh is greater than or equal to the second preset threshold, then the preferred triangular mesh and first corresponding region is determined as first
With result.
10. a kind of fish eye images matching system according to claim 8, which is characterized in that second match point is rejected
Unit specifically includes:
4th determines subelement, for preferred triangular mesh after determining rotation；Preferred triangular mesh is rotation after the rotation
Match point number is greater than or equal to the network of triangle of first preset threshold in the first triangular mesh fish eye images after turning
Lattice；
5th determines subelement, for determining the second corresponding region；Second corresponding region is second triangular mesh
The second corresponding triangular mesh in fish eye images；Described second corresponding triangular mesh is the second triangular mesh flake
The triangular mesh most comprising match point number corresponding with triangular mesh preferred after the rotation in image；
6th determines subelement, if for the match point in triangular mesh preferred after the rotation and second corresponding region
Pair number be greater than or equal to the second preset threshold, then by triangular mesh preferred after the rotation and second corresponding region
It is determined as the second matching result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910693447.4A CN110399931B (en) | 2019-07-30 | 2019-07-30 | Fisheye image matching method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910693447.4A CN110399931B (en) | 2019-07-30 | 2019-07-30 | Fisheye image matching method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110399931A true CN110399931A (en) | 2019-11-01 |
| CN110399931B CN110399931B (en) | 2021-07-06 |
Family
ID=68326603
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910693447.4A Active CN110399931B (en) | 2019-07-30 | 2019-07-30 | Fisheye image matching method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110399931B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115690607A (en) * | 2023-01-04 | 2023-02-03 | 西湖大学 | Rainfall inversion method and device based on infrared remote sensing and deep learning |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012146199A (en) * | 2011-01-13 | 2012-08-02 | Nippon Steel Corp | Sight line position detection device, sight line position detection method, and computer program |
| CN102651069A (en) * | 2012-03-31 | 2012-08-29 | 重庆大学 | Contour-based local invariant region detection method |
| CN104392457A (en) * | 2014-12-11 | 2015-03-04 | 中国测绘科学研究院 | Automatic matching method and device for connection points of slanted images |
| CN105513094A (en) * | 2015-12-17 | 2016-04-20 | 上海交通大学 | Stereo vision tracking method and stereo vision tracking system based on 3D Delaunay triangulation |
| CN107067370A (en) * | 2017-04-12 | 2017-08-18 | 长沙全度影像科技有限公司 | A kind of image split-joint method based on distortion of the mesh |
| CN108765292A (en) * | 2018-05-30 | 2018-11-06 | 中国人民解放军军事科学院国防科技创新研究院 | Image split-joint method based on the fitting of space triangular dough sheet |
| CN109388455A (en) * | 2018-09-14 | 2019-02-26 | 浩云科技股份有限公司 | It is a kind of to support multi-platform fish eye images that monitoring method is unfolded based on Opengl es |
| CN109816706A (en) * | 2019-02-01 | 2019-05-28 | 辽宁工程技术大学 | A kind of smoothness constraint and triangulation network equal proportion subdivision picture are to dense matching method |
| CN111126418A (en) * | 2018-10-30 | 2020-05-08 | 国家海洋局第一海洋研究所 | Oblique image matching method based on planar perspective projection |
-
2019
- 2019-07-30 CN CN201910693447.4A patent/CN110399931B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012146199A (en) * | 2011-01-13 | 2012-08-02 | Nippon Steel Corp | Sight line position detection device, sight line position detection method, and computer program |
| CN102651069A (en) * | 2012-03-31 | 2012-08-29 | 重庆大学 | Contour-based local invariant region detection method |
| CN104392457A (en) * | 2014-12-11 | 2015-03-04 | 中国测绘科学研究院 | Automatic matching method and device for connection points of slanted images |
| CN105513094A (en) * | 2015-12-17 | 2016-04-20 | 上海交通大学 | Stereo vision tracking method and stereo vision tracking system based on 3D Delaunay triangulation |
| CN107067370A (en) * | 2017-04-12 | 2017-08-18 | 长沙全度影像科技有限公司 | A kind of image split-joint method based on distortion of the mesh |
| CN108765292A (en) * | 2018-05-30 | 2018-11-06 | 中国人民解放军军事科学院国防科技创新研究院 | Image split-joint method based on the fitting of space triangular dough sheet |
| CN109388455A (en) * | 2018-09-14 | 2019-02-26 | 浩云科技股份有限公司 | It is a kind of to support multi-platform fish eye images that monitoring method is unfolded based on Opengl es |
| CN111126418A (en) * | 2018-10-30 | 2020-05-08 | 国家海洋局第一海洋研究所 | Oblique image matching method based on planar perspective projection |
| CN109816706A (en) * | 2019-02-01 | 2019-05-28 | 辽宁工程技术大学 | A kind of smoothness constraint and triangulation network equal proportion subdivision picture are to dense matching method |
Non-Patent Citations (6)
| Title |
|---|
| ENDE WANG 等: "Tri-SIFT: A Triangulation-Based Detection and Matching Algorithm for Fish-Eye Images", 《INFORMATION》 * |
| JEAN-MICHEL MOREL 等: "ASIFT: A NEW FRAMEWORK FOR FULLY AFFINE INVARIANT IMAGE COMPARISON", 《SIAM JOURNAL ON IMAGING SCIENCES》 * |
| LIU FENG 等: "Affine invariant image watermarking scheme based on ASIFT and Delaunay tessellation", 《MULTIMEDIA TOOLS AND APPLICATIONS》 * |
| 李爽 等: "结合局部单应矩阵的ZY-3遥感影像稠密匹配", 《测绘与空间地理信息 》 * |
| 王向军 等: "Delaunay三角剖分和仿射约束的特征相同", 《光学学报》 * |
| 赵鹏图 等: "基于局部特征的大视角图像匹配", 《光学学报》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115690607A (en) * | 2023-01-04 | 2023-02-03 | 西湖大学 | Rainfall inversion method and device based on infrared remote sensing and deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110399931B (en) | 2021-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110211043B (en) | Registration method based on grid optimization for panoramic image stitching | |
| CN108564617A (en) | Three-dimensional rebuilding method, device, VR cameras and the panorama camera of more mesh cameras | |
| CN107169924A (en) | The method for building up and system of three-dimensional panoramic image | |
| WO2018014601A1 (en) | Method and relevant apparatus for orientational tracking, method and device for realizing augmented reality | |
| CN110349219A (en) | A kind of Camera extrinsic scaling method and device | |
| CN106875451A (en) | Camera calibration method, device and electronic equipment | |
| CN104217459B (en) | A kind of spheroid character extracting method | |
| CN110580720B (en) | Panorama-based camera pose estimation method | |
| CN106898051A (en) | The visual field elimination method and server of a kind of virtual role | |
| CN111415420B (en) | Spatial information determining method and device and electronic equipment | |
| WO2004111799A2 (en) | Systems and methods for using visual hulls to determine the number of people in a crowd | |
| CN108550166A (en) | A kind of spatial target images matching process | |
| CN114119607B (en) | Wine bottle defect sample generation method and system based on deep neural network | |
| CN110070567A (en) | A kind of ground laser point cloud method for registering | |
| CN106651956A (en) | Method for calibrating parabolic refraction and reflection camera via single ball and parallel circle properties | |
| CN103714530A (en) | Vanishing point detection and image correction method | |
| CN109117851A (en) | A kind of video image matching process based on lattice statistical constraint | |
| CN110399931A (en) | A kind of fish eye images matching process and system | |
| Yan et al. | A novel data augmentation method for detection of specific aircraft in remote sensing RGB images | |
| CN108596032A (en) | It fights in a kind of video behavioral value method, apparatus, equipment and medium | |
| CN116402976A (en) | Training method and device for three-dimensional target detection model | |
| CN114998773A (en) | Characteristic mismatching elimination method and system suitable for aerial image of unmanned aerial vehicle system | |
| Bellavia et al. | Image orientation with a hybrid pipeline robust to rotations and wide-baselines | |
| CN107067441A (en) | Camera marking method and device | |
| CN110727817A (en) | Three-dimensional model retrieval method based on t-CNN, terminal equipment and storage medium |
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 |