CN106327423B - Remote sensing image registration method and system based on directed line segment - Google Patents
Remote sensing image registration method and system based on directed line segment Download PDFInfo
- Publication number
- CN106327423B CN106327423B CN201610709894.0A CN201610709894A CN106327423B CN 106327423 B CN106327423 B CN 106327423B CN 201610709894 A CN201610709894 A CN 201610709894A CN 106327423 B CN106327423 B CN 106327423B
- Authority
- CN
- China
- Prior art keywords
- line segment
- registration
- directed line
- image
- matching
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
Landscapes
- Image Analysis (AREA)
Abstract
The present invention relates to technical field of image processing, specifically disclose a kind of remote sensing image registration method based on directed line segment.This method uses BRISK algorithm to extract thick matching double points first, then oriented line-segment sets are constructed in reference picture and image subject to registration respectively using two groups of matching point sets, it is used as the neighbouring matched method of line segment to carry out directed line segment matching using F- norm criterion, the accurate positioning of match point is finally carried out based on probability statistics model, solve registration parameter, so obtain registration reference picture and image subject to registration be registrated.In addition, the present invention also provides a kind of remote sensing image registration system based on directed line segment.The present invention carries out Feature Points Matching, has reached effective rejecting Mismatching point by fully considering the texture information of image when constructing oriented line segment feature, hence it is evident that improves the technical effect of matching accuracy.
Description
Technical field
The present invention relates to technical field of image processing, in particular to a kind of remote sensing image registration method based on directed line segment
And system.
Background technique
Image registration is the key technology in the fields such as remote sensing image processing, target identification, image reconstruction, robot vision
One of, it is the basis of image mosaic technology.
Remote sensing image registration is broadly divided into two classes at present: the method for registering based on area grayscale and the registration side based on feature
Method.
Wherein, the common method for registering images based on area grayscale has: cross-correlation method, the phase correlation method based on FFT and
Mutual information method etc..
The technological means generally taken are as follows: the method based on characteristics of image extracts edge, angle point, profile first from image
With the features such as regional center, then characteristic point is described and finds corresponding relationship between them.
In the prior art, BRISK algorithm (Binary Robust Invariant Scalable is generallyd use
Keypoints it) is randomly selected in local image neighborhood a little pair, establishes Feature Descriptor with the gray-scale relation of point pair.But tradition
Although BRISK algorithm have speed advantage, its poor robustness, error hiding rate are higher, tend not to meet image registration
Required precision.
Summary of the invention
The present invention is directed to overcome the higher technological deficiency of existing BRISK algorithm error hiding rate, provide a kind of based on directed line
The remote sensing image registration method of section, comprising the following steps:
S1, input reference picture I and image I ' subject to registration；
S2, the reference picture I and the image I ' subject to registration are slightly matched using BRISK algorithm, and is obtained thick
Matching double points form two groups of matching point set V={ a1,a2…,anAnd V'={ b1,b2…,bn}；
S3, oriented line-segment sets are constructed in reference picture I and image I ' subject to registration according to two groups of matching point sets respectively,
And it matches to obtain smart matching double points using the directed line segment collection；
S4, the obtained smart matching double points, solution registration parameter are utilized；Using the registration parameter to image subject to registration
Geometric transformation is carried out, registration result is obtained.
In some embodiments, the step S2 the following steps are included:
S21, in the reference picture I and the image I ' subject to registration characteristic point is extracted respectively；
S22, BRISK feature descriptor is established for extracted characteristic point；
S23, pass through Hamming distance metrics match degree, obtain thick matching double points.
In some embodiments, the step S3 the following steps are included:
S31, according to the characteristic point of reference picture I and image I' subject to registration, construct two groups of directed graph G=(V, E) respectively
With G'=(V', E')；
S32, m directed line segment is constructed respectively from two groups of directed graphs G and G', and calculate every directed line segment
Feature difference matrix；
S33, F- norm criterion is used to carry out directed line segment matching as the neighbouring matched method of line segment, acquisition is matched to be had
To line segment；
S34, for the matched directed line segment, accurate matching double points are obtained according to probability statistics model.
In some embodiments, in the step S31, for G=(V, E) and G'=(V', E'), V={ a is defined1,a2…,
anAnd V'={ b1,b2…,bnIt is reference picture I and the matched characteristic point of image I' subject to registration respectively, E and E' are directed graphs
The endpoint of G and G', here E={ (ai,aj), i < j }, E'={ (bi,bj), i < j }, wherein j≤n.
In some embodiments, the step S32 the following steps are included:
M directed line segment is respectively constructed in the reference picture I and the image I' subject to registration, is expressed as L=
[l1,l2,…lm] and L'=[l'1,l'2,…l'm]；
The feature description for calculating every directed line segment, remembers directed line segment liFeature be described as Si, directed line segment l'iSpy
Sign is described as S'i, wherein i≤m.
In some embodiments, directed line segment is constructed in the following ways, for directed line segment eij, line segment starting point ai, terminal
aj, take three sampled point (p1,p2,p3), calculation formula is as follows:
p1=ai；
p2=(ai+aj)/2；
p3=aj；
Extract the BRISK feature difference matrix S=(s of three sampled points1,s2,s3), the feature as directed line segment
Difference matrix, each in S are classified as BRISK description.
In some embodiments, the step S33 the following steps are included:
S331 calculates F- norm described in the directed line segment feature difference matrix:
d(li,l'i)=| | Si-S'i||F,
Obtaining d is m dimensional vector；
D is normalized in S332, when d is less than given threshold value TFWhen, l' is the proximity matching line segment of l, two neighbours
The beginning and end of the directed line segment of near match is two pairs of match points respectively.
In some embodiments, the step S34 counts each pair of matching double points by traversing the proximity matching line segment.
In addition, the invention also provides a kind of remote sensing image registration system based on directed line segment, comprises the following modules:
Input module, for inputting reference picture I and image I ' subject to registration；
Thick matching module, for carrying out thick to the reference picture I and the image I ' subject to registration using BRISK algorithm
Match, and obtain thick matching double points, forms two groups of matching point sets；
Smart matching module, for according to two groups of matching point sets structure in reference picture I and image I ' subject to registration respectively
Directed line segment collection is made, and matches to obtain smart matching double points using the directed line segment collection；
Output module, for solving registration parameter, and utilize the registration parameter using the obtained smart matching double points
Geometric transformation is carried out to image subject to registration, exports registration result.
In some embodiments, the essence matching module further includes with lower module:
Directed graph constructing module constructs two groups for the characteristic point according to reference picture I and image I' subject to registration respectively
Directed graph G=(V, E) and G'=(V', E')；
Directed line segment constructing module, for constructing m directed line segment respectively from two groups of directed graphs G and G', and
Calculate the feature difference matrix of every directed line segment；
Directed line segment matching module, for using F- norm criterion to carry out directed line segment as the neighbouring matched method of line segment
Matching, obtains matched directed line segment；
Accurate matching double points obtain module, for obtaining accurate matching double points according to probability statistics model.
The beneficial effects of the present invention are: the present invention is based on directed line segments to carry out images match, constructs oriented line segment feature
When, it fully considers the texture information of image, has reached and made the matching double points of acquisition more accurate, what error hiding rate substantially reduced has
Beneficial effect.
More, due to the description using BRISK feature as directed line segment, hence it is evident that shorten runing time, there is speed
Degree advantage.
Detailed description of the invention
Fig. 1 shows the flow chart of the remote sensing image registration method the present invention is based on directed line segment；
Fig. 2 shows the present invention is based on the flow charts of one specific embodiment of remote sensing image registration method of directed line segment；
Fig. 3 shows feature of present invention point detecting method；
Fig. 4 shows 60 point sampling model of BRISK Feature Descriptor of the present invention；
Fig. 5 is that the present invention is based on the specific embodiments that the remote sensing image registration method of directed line segment constructs directed line segment
Schematic diagram；
The present invention is based on the remote sensing image registration methods of directed line segment to carry out emulation experiment to remote sensing images according to Fig. 6
Registration result figure.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, below in conjunction with attached drawing and specific implementation
Example, the present invention will be described in further detail.It should be appreciated that specific embodiment described herein is only to explain this hair
It is bright, but not to limit the present invention.
Central scope of the invention are as follows: by designing a kind of remote sensing image registration method based on directed line segment, to reference
Image and image subject to registration carry out thick matching and essence matching, using existing BRISK algorithm to reference picture and image subject to registration
It is slightly matched, then smart matching is carried out using directed line segment.
Specific embodiments of the present invention is described in detail with reference to the accompanying drawing.
Fig. 1 and Fig. 2 are please referred to, for the present invention is based on the flow charts of the remote sensing image registration method of directed line segment and one to have
The flow chart of body embodiment, comprising the following steps:
Execute step S1: input reference picture I and image I' subject to registration.
It executes step S2: slightly being matched using BRISK algorithm, obtain thick matching double points, form two groups and slightly match point set,
The feature point set for remembering reference picture I is V={ a1,a2…,an, the matched feature point set of image I' subject to registration is V'={ b1,
b2…,bn}。
Specifically, step S2 is realized by following steps:
S21, in the reference picture I and the image I ' subject to registration characteristic point is extracted respectively.
Referring to Fig. 3, showing image characteristic point detection process of the present invention.
Below to the feature point extraction process of reference picture I, it is described.Extraction for characteristic point, uses with lower section
Formula carries out:
S211, scale space pyramid is constructed for the reference picture I of input.Pyramid includes n surface layer ciWith n internal layer
di, wherein n is a setup parameter, and n takes 4, i={ 0,1 ..., n-1 } in the present invention.Wherein, c0For reference picture I.To original image
As c0Continuous half sampling obtains all surface layers, internal layer d0By to original image c0Down-sampled 1.5 times obtain, other internal layers continuous half
Sampling obtains di.To one 2n layers of pyramid of building；
S212, using FAST/AGAST Corner Detection Algorithm in pyramidal each layer of progress characteristic point detection, the present invention
Middle Corner Detection threshold parameter value is 30, and carries out the inhibition of 3D maximum value to gained characteristic point；
S213, least square fitting and one-dimensional Parabolic Fit are carried out to each angle point by interpolation method, to be had
There is the characteristic point of Pixel-level positioning accuracy and exact scale.
By step S211-S213 to obtain the spy that reference picture I has Pixel-level positioning accuracy and exact scale
Sign point；Similarly, by the corresponding characteristic point of the available image I ' to be processed of same mode, details are not described herein again.
The characteristic point of S22, respectively acquired reference picture I and the image I ' subject to registration are established BRISK feature and are retouched
State symbol.The process for establishing BRISK feature descriptor is described below:
Referring to Fig. 4, showing feature of the present invention describes 60 point sampling models.By following steps come real
It is existing:
S221, centered on characteristic point, around it obtain n sampled point, to construct multiple concentric circles, sampled point etc.
Spacing is distributed on discretization Breaenham concentric circles.Gaussian filtering (standard deviation δ is carried out to each sampled pointi), δiWith
The distance of characteristic point to sampled point is directly proportional.
S222, Ω (sampling point set) is calculated, Ω is defined by Euclidean distanceL(point is to subset over long distances) and ΩsIt is (short
Range points are to subset), calculation formula is as follows:
Ω={ (pi,pj)∈R2×R2| j < i < N ∧ i, j ∈ N }, (1)；
Wherein, (pi,pj) it is sampled point pair；δmin、δmaxIt is distance threshold, takes 13.67t and 9.75t respectively；
S223, the partial gradient for calculating every bit pair, calculation formula are as follows:
S224, by being put over long distances to subset ΩLPoint the direction of characteristic point is calculated, calculation formula is as follows:
S225, by image rotation to principal direction, i.e., centered on characteristic point will sample 2 (g of rotation alpha=arctany,gx) angle
Degree, in ΩsComparison sampled point is to (p in (short distance point is to subset)α i,pα j) intensity, eventually form 512bit Feature Descriptor,
Each bit definitions are shown below:
S23, final handy Hamming distance metrics match degree, obtain slightly matching point set V={ a1,a2…,anAnd V'=
{b1,b2…,bn}。
It executes step S3: being configured in reference picture I and image I ' subject to registration respectively according to two groups of matching point sets
It matches to obtain smart matching double points to line-segment sets, and using the directed line segment collection.
Specifically, step S3 is realized by following steps:
S31: according to the characteristic point of reference picture I and image I' subject to registration, two groups of directed graph G=(V, E) are constructed respectively
With G'=(V', E').
Wherein, V={ a1,a2…,an, (7)
V'={ b1,b2…,bn, (8)
E and E' is the endpoint of directed graph G and G', it may be assumed that
E={ (ai,aj), i < j }, (9)
E'={ (bi,bj),i<j}。 (10)
Wherein, j≤n.
S32: m directed line segment is constructed from directed graph G and G' respectively, and calculates the feature difference of every directed line segment
Matrix.
Referring to Fig. 5, showing the construction directed line segment e from directed graph GijProcess.Specifically, for directed line segment
eij(directed line segment eijStarting point is ai, terminal aj), take three sampled point (p1,p2,p3), wherein p1=ai, p2=(ai+aj)/2,
p3=aj, namely starting point, midpoint and the terminal of G are chosen respectively as three sampled points.
Then the BRISK feature difference matrix for extracting these three sampled points, is denoted as S=(s1,s2,s3), as directed line segment
Feature difference matrix, each BRISK description for being classified as the point in S, i.e. s1、s2、s3Respectively p1、p2、p3BRISK retouch
State son.
Directed line segment e is constructed from directed graph G ' using above-mentioned same methodij', and obtain directed line segment eij'
Feature difference matrix S '.
By executing step S32, m directed line segment is respectively constructed in reference picture I and image I' subject to registration, is respectively indicated
For L=[l1,l2,…lm] and L'=[l'1,l'2,…l'm]；Directed line segment l is remembered simultaneouslyiFeature be described as Si, directed line segment
l'iFeature be described as S'i, wherein i≤m.
S33: F- norm criterion is used to carry out directed line segment matching as the neighbouring matched method of line segment, acquisition is matched to be had
To line segment.
Specifically, step S331 is executed, the F- norm of directed line segment feature difference matrix is calculated, formula is as follows:
d(li,l'i)=| | Si-S'i||F, (11)
Obtaining d is m dimensional vector, executes step S332, d is normalized, when d is less than given threshold value TFWhen, l' is
For the proximity matching line segment of l, and the beginning and end of two matched directed line segments is two pairs of match points respectively.Preferably, originally
Threshold value T in inventionFValue is 0.6.
S34, for the matched directed line segment, accurate matching double points are obtained according to probability statistics model.Specifically, first
First, it is as follows to define element value calculating method in the vector K, K that a length is m:
Secondly, establishing an empty statistical matrix G ∈ R2×m, for counting the matching times of corresponding points, the calculating process of G
It is as follows:
Initialize empty matrix G ∈ R2×m；
I=1,2 ..., m,
If K (i)=1, directed line segment liIt is ap→aq, corresponding neighbouring directed line segment l'iIt is bp→bq, then G (1, i) is indicated
Corresponding points (ap,bp), G (2, i) indicates corresponding points (aq,bq),
G (1, i)=G (1, i)+1, G (2, i)=G (2, i)+1；
Output matrix G；
Element is the frequency that each pair of match point occurs in matrix G, thus obtains the frequency matrix F of each pair of match point, is calculated
Method such as following formula:
In formula, p=1,2, q=1,2..., m.
Given frequency threshold value TPIf certain F > T to matching double pointsP, that is, the matching double points are chosen as smart matching double points.
TPIt is smaller to be worth value, then it is more to retain match point logarithm, while Mismatching point pair may be retained；TPIt is bigger to be worth value
Mismatching point pair more can be effectively rejected, but match point logarithm opposite can be reduced.
Preferably, the present invention uses TP=0.0025.This is because TPWhen=0.0025, error hiding can be effectively rejected
Point, while enough match points can be retained.
It executes step S4: using obtained accurate matching double points, transformation square can be solved by following formula image transformation relation
Battle array H, is overlapped after image I ' subject to registration is converted by H with reference picture I, so that it may obtain stitching image.
In formula, (x, y) and (x ', y ') is the essence matching of reference picture I and image I ' subject to registration acquired in step S3
The coordinate of point pair.
The remote sensing image registration system based on directed line segment that the invention also provides a kind of is based on directed line segment using above-mentioned
Remote sensing image registration method carry out the registrations of remote sensing images.
Specifically, the present invention is based on the remote sensing image registration systems of directed line segment to comprise the following modules:
Input module, for inputting reference picture I and image I ' subject to registration.
Thick matching module, for carrying out thick to the reference picture I and the image I ' subject to registration using BRISK algorithm
Match, and obtain thick matching double points, forms two groups of matching point sets.
Smart matching module, for according to two groups of matching point sets structure in reference picture I and image I ' subject to registration respectively
Directed line segment collection is made, and matches to obtain smart matching double points using the directed line segment collection.
Output module, for solving registration parameter, and utilize the registration parameter using the obtained smart matching double points
Geometric transformation is carried out to image subject to registration, exports registration result.
Preferably, smart matching module further includes with lower module:
Directed graph constructing module constructs two groups for the characteristic point according to reference picture I and image I' subject to registration respectively
Directed graph G=(V, E) and G'=(V', E')；
Directed line segment constructing module, for constructing m directed line segment respectively from two groups of directed graphs G and G', and
Calculate the feature difference matrix of every directed line segment；
Directed line segment matching module, for using F- norm criterion to carry out directed line segment as the neighbouring matched method of line segment
Matching, obtains matched directed line segment；
Accurate matching double points obtain module, for obtaining accurate matching double points according to probability statistics model.
The present invention is based on directed line segments to carry out images match, when constructing oriented line segment feature, fully considers the texture of image
Information has reached and has made the matching double points of acquisition more accurate, the beneficial effect that error hiding rate substantially reduces.
Effect of the invention is described further below with reference to analogous diagram.
1. simulated conditions and parameter
Hardware platform are as follows: Intel (R) Core (TM) i5-4570CPU 3.20GHz；
Software platform are as follows: Windows 7.0, Matlab 2012b.
Emulation experiment parameter: BRISK algorithm Corner Detection threshold value is 60, neighbouring line segment matching threshold TFIt is 0.6, probability system
Count the threshold value T in modelPIt is 0.0025.
2. emulation experiment content:
Fig. 6 (a) is reference picture, and size is 306 × 386 pixels；
Fig. 6 (b) is special registration image, and size is 335 × 472 pixels；
Fig. 6 (c) is the image after registration.
It from Fig. 6 (c) as it can be seen that the image after registration is perfectly aligned, does not shift, illustrates that the present invention can be very to remote sensing images
Good registration.
3. the simulation experiment result compares:
The present invention is compared with the matching accuracy and runing time of existing BRISK algorithm and SIFT algorithm, is tied
Fruit such as table 1:
Table 1 shows the present invention and compares with existing algorithm:
| Method | CMR | Time(s) |
| BRISK | 71.42% | 0.1023 |
| SIFT | 81.58% | 1.7813 |
| The present invention | 100% | 0.1146 |
Wherein, CMR indicates matching accuracy, and Time is runing time.
As seen from Table 1, the more existing BRISK algorithm of the present invention and SIFT algorithmic match accuracy are all improved, same to phase
Compared with SIFT algorithm, runing time of the invention is obviously shortened.
In conclusion the present invention is taking into account matching accuracy and two aspect superior performance of runing time, it can be to remote sensing figure
As carrying out effective registration in real time.
The above described specific embodiments of the present invention are not intended to limit the scope of the present invention..Any basis
Any other various changes and modifications made by technical concept of the invention should be included in the guarantor of the claims in the present invention
It protects in range.
Claims (8)
1. a kind of remote sensing image registration method based on directed line segment, which comprises the following steps:
S1, input reference picture I and image I ' subject to registration；
S2, the reference picture I and the image I ' subject to registration are slightly matched using BRISK algorithm, and obtains thick matching
Point pair forms two groups of matching point sets；
S3, oriented line-segment sets, and benefit are constructed in reference picture I and image I ' subject to registration respectively according to two groups of matching point sets
It matches to obtain smart matching double points with the directed line segment collection；
S4, the obtained smart matching double points, solution registration parameter are utilized；Image subject to registration is carried out using the registration parameter
Geometric transformation obtains registration result；
Wherein, the step S3 the following steps are included:
S31, according to the characteristic point of reference picture I and image I' subject to registration, construct two groups of directed graph G=(V, E) and G' respectively
=(V', E')；
S32, m directed line segment is constructed respectively from two groups of directed graphs G and G', and calculate the feature of every directed line segment
Difference matrix；
S33, use F- norm criterion to carry out directed line segment matching as the neighbouring matched method of line segment, obtain matched directed line
Section；
S34, for the matched directed line segment, accurate matching double points are obtained according to probability statistics model.
2. as described in claim 1 based on the remote sensing image registration method of directed line segment, which is characterized in that the step S2 packet
Include following steps:
S21, in the reference picture I and the image I ' subject to registration characteristic point is extracted respectively；
S22, BRISK feature descriptor is established for extracted characteristic point；
S23, pass through Hamming distance metrics match degree, obtain thick matching double points.
3. as claimed in claim 2 based on the remote sensing image registration method of directed line segment, which is characterized in that the step S31
In, for G=(V, E) and G'=(V', E'), define V={ a1,a2…,anAnd V'={ b1,b2…,bnIt is with reference to figure respectively
Picture I and the matched characteristic point of image I' subject to registration, E and E' are the endpoints of directed graph G and G', here E={ (ai,aj), i < j },
E'={ (bi,bj), i < j }, wherein j≤n.
4. as claimed in claim 3 based on the remote sensing image registration method of directed line segment, which is characterized in that the step S32
The following steps are included:
M directed line segment is respectively constructed in the reference picture I and the image I' subject to registration, is expressed as L=[l1,
l2,…lm] and L'=[l'1,l'2,…l'm]；
The feature description for calculating every directed line segment, remembers directed line segment liFeature be described as Si, directed line segment l'iFeature retouch
It states as S'i, wherein i≤m.
5. as claimed in claim 4 based on the remote sensing image registration method of directed line segment, which is characterized in that in the following ways
Directed line segment is constructed, for directed line segment eij, line segment starting point ai, terminal aj, take three sampled point (p1,p2,p3), calculation formula
It is as follows:
p1=ai；
p2=(ai+aj)/2；
p3=aj；
Extract the BRISK feature difference matrix S=(s of three sampled points1,s2,s3), the feature difference as directed line segment
Matrix, each in S are classified as BRISK description.
6. as described in claim 1 based on the remote sensing image registration method of directed line segment, which is characterized in that the step S33
The following steps are included:
S331 calculates the F- norm of the directed line segment feature difference matrix:
d(li,l'i)=| | Si-S'i||F,
Obtaining d is m dimensional vector；
D is normalized in S332, when d is less than given threshold value TFWhen, l' is the proximity matching line segment of l, and two neighbouring
The beginning and end for the directed line segment matched is two pairs of match points respectively.
7. as claimed in claim 6 based on the remote sensing image registration method of directed line segment, which is characterized in that the step S34
By traversing the proximity matching line segment, each pair of matching double points are counted.
8. a kind of remote sensing image registration system based on directed line segment, which is characterized in that comprise the following modules:
Input module, for inputting reference picture I and image I ' subject to registration；
Thick matching module, for slightly being matched using BRISK algorithm to the reference picture I and the image I ' subject to registration,
And thick matching double points are obtained, form two groups of matching point sets；
Smart matching module, for being configured in reference picture I and image I ' subject to registration respectively according to two groups of matching point sets
It matches to obtain smart matching double points to line-segment sets, and using the directed line segment collection；
Output module, for solving registration parameter, and treat using the registration parameter using the obtained smart matching double points
It is registrated image and carries out geometric transformation, export registration result；
Wherein, the smart matching module includes:
Directed graph constructing module constructs two groups of orientations for the characteristic point according to reference picture I and image I' subject to registration respectively
Figure G=(V, E) and G'=(V', E')；
Directed line segment constructing module for constructing m directed line segment respectively from two groups of directed graphs G and G', and calculates
The feature difference matrix of every directed line segment；
Directed line segment matching module, for using F- norm criterion to carry out directed line segment as the neighbouring matched method of line segment
Match, obtains matched directed line segment；
Accurate matching double points obtain module, for obtaining accurate matching double points according to probability statistics model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610709894.0A CN106327423B (en) | 2016-08-23 | 2016-08-23 | Remote sensing image registration method and system based on directed line segment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610709894.0A CN106327423B (en) | 2016-08-23 | 2016-08-23 | Remote sensing image registration method and system based on directed line segment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106327423A CN106327423A (en) | 2017-01-11 |
| CN106327423B true CN106327423B (en) | 2019-08-23 |
Family
ID=57742423
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610709894.0A Active CN106327423B (en) | 2016-08-23 | 2016-08-23 | Remote sensing image registration method and system based on directed line segment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106327423B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107526772A (en) * | 2017-07-12 | 2017-12-29 | 湖州师范学院 | Image indexing system based on SURF BIT algorithms under Spark platforms |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103310453B (en) * | 2013-06-17 | 2015-12-09 | 北京理工大学 | A kind of fast image registration method based on subimage Corner Feature |
-
2016
- 2016-08-23 CN CN201610709894.0A patent/CN106327423B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN106327423A (en) | 2017-01-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ma et al. | Robust feature matching for remote sensing image registration via locally linear transforming | |
| CN101980250B (en) | Method for identifying target based on dimension reduction local feature descriptor and hidden conditional random field | |
| CN104167003A (en) | Method for fast registering remote-sensing image | |
| CN101833765B (en) | Characteristic matching method based on bilateral matching and trilateral restraining | |
| CN104751465A (en) | ORB (oriented brief) image feature registration method based on LK (Lucas-Kanade) optical flow constraint | |
| CN105427298A (en) | Remote sensing image registration method based on anisotropic gradient dimension space | |
| CN106991695A (en) | A kind of method for registering images and device | |
| CN102592281B (en) | Image matching method | |
| CN103400384A (en) | Large viewing angle image matching method capable of combining region matching and point matching | |
| CN101556692A (en) | Image mosaic method based on neighborhood Zernike pseudo-matrix of characteristic points | |
| CN104616247B (en) | A kind of method for map splicing of being taken photo by plane based on super-pixel SIFT | |
| CN108550166B (en) | Spatial target image matching method | |
| Chen et al. | Robust affine-invariant line matching for high resolution remote sensing images | |
| CN112396643A (en) | Multi-mode high-resolution image registration method with scale-invariant features and geometric features fused | |
| CN103632142A (en) | Local coordinate system feature description based image matching method | |
| CN104504723A (en) | Image registration method based on remarkable visual features | |
| CN108550165A (en) | A kind of image matching method based on local invariant feature | |
| CN104616280A (en) | Image registration method based on maximum stable extreme region and phase coherence | |
| Tao et al. | Robust point sets matching by fusing feature and spatial information using nonuniform Gaussian mixture models | |
| US20200005078A1 (en) | Content aware forensic detection of image manipulations | |
| Kabbai et al. | Image matching based on LBP and SIFT descriptor | |
| Liu et al. | Iris recognition in visible spectrum based on multi-layer analogous convolution and collaborative representation | |
| CN108961164A (en) | Image registration joining method and device and computer equipment based on geometrical invariants | |
| CN103914690A (en) | Shape matching method based on projective invariant | |
| Selvaraj et al. | Enhanced copy–paste forgery detection in digital images using scale‐invariant feature transform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |