CN101667293A - Method for conducting high-precision and steady registration on diversified sensor remote sensing images - Google Patents

Method for conducting high-precision and steady registration on diversified sensor remote sensing images Download PDF

Info

Publication number
CN101667293A
CN101667293A CN200910307634A CN200910307634A CN101667293A CN 101667293 A CN101667293 A CN 101667293A CN 200910307634 A CN200910307634 A CN 200910307634A CN 200910307634 A CN200910307634 A CN 200910307634A CN 101667293 A CN101667293 A CN 101667293A
Authority
CN
China
Prior art keywords
registration
reference mark
gray scale
point
precision
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN200910307634A
Other languages
Chinese (zh)
Inventor
郭琰
张晔
谷延锋
仲伟志
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Harbin Institute of Technology
Original Assignee
Harbin Institute of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Harbin Institute of Technology filed Critical Harbin Institute of Technology
Priority to CN200910307634A priority Critical patent/CN101667293A/en
Publication of CN101667293A publication Critical patent/CN101667293A/en
Pending legal-status Critical Current

Links

Images

Abstract

The invention discloses a method for conducting high-precision and steady registration on diversified sensor remote sensing images, which relates to the field of remote sensing image processing and adopts the following steps aiming at the problem that the control points are unevenly distributed in the registration process: step 1: crude registration of combined point characteristics and regional characteristics and elimination of larger scale, rotation and translation differences between a reference image and an input image, which are realized by matching the combined point characteristics andregional characteristics; step 2: scale space feature extraction and matching of a great amount of registration control point pairs, which aim at extracting a great amount of control point pairs forfine registration; and step 3: screening and fine registration based on control point information amount, which aims at screening the control points according to self-contained information amount andfinishing high-precision fine registration. The invention provides the method for conducting high-precision and steady registration on diversified sensor remote sensing images.

Description

The multiple sensors remote sensing images are carried out the method for high-precision and steady registration
Technical field
The present invention relates to the remote sensing image processing field, be specifically related to a kind of method of the multiple sensors remote sensing images being carried out high-precision and steady registration.
Background technology
Along with the development of remote sensor technology, the complementarity between the data that different sensors is obtained is more and more stronger.How to unite the view data of these dissimilar sensors and carry out associated treatment, image registration is its vital precondition.
Method for registering images mainly can be divided into two big classes based on global technique and local techniques, and global technique need be utilized all Pixel Information of image, and calculated amount is bigger, and the local feature that local techniques is then extracted image carries out registration.The pixel quantity of remote sensing images is huge, mainly adopts the registration of local techniques, and the key of local techniques is extraction and coupling to registration control points in the reference picture of registration and the input picture.For the remote sensing images that multiple sensors obtains, the difference of its imaging causes extracts the difficulty increasing that correct registration control points is right in reference picture and the input picture, thereby is difficult to realize high-precision registration.
Existing multisensor remote sensing images mainly adopt certain Feature Extraction Technology to extract local feature (point in reference picture and the input picture based on the method for registering of local techniques, line, provincial characteristics), the constant descriptor of utilization local feature is set up the matching relationship of local feature in reference picture and the input picture, and then it is right to obtain the registration registration control points.These methods adopt a certain local feature to extract mostly and the method for coupling is obtained registration control points, this single feature extraction and matching process are poor to the robustness of image pixel gray-scale value nonlinearities change, thereby have limited the application in having the multi-sensor image registration of the non-linear difference of gray scale.In the image registration based on local feature, the distribution situation of registration control points will influence final registration accuracy in reference picture and the input picture, and it is even more to distribute, and the root-mean-square error of registration is more little.
Summary of the invention
The present invention is directed to reference mark problem pockety in the registration process, and propose the multiple sensors remote sensing images are carried out the method for high-precision and steady registration.
The present invention is as follows to the step that the multiple sensors remote sensing images carry out high-precision and steady registration:
Step 1: the thick registration of uniting a feature and provincial characteristics:
Choosing of some feature: the key point for coupling is extracted in the conversion of utilization yardstick invariant features respectively from reference picture and input picture; Key point in the utilization Euclidean distance calculating reference picture and the distance between the key point in the input picture, it is right as the thick registration point of candidate to choose three pairs of minimum in described distance key points;
Choosing of provincial characteristics: the utilization maximum stable extremal region extracts the zone in reference picture and the input picture respectively, adopt invariant moments that two-way coupling is carried out in zone in the reference picture and the zone in the input picture and obtain three pairs of the most similar zones, with the center of gravity in zone as the thick registration control points of candidate;
Thick registration: unite the thick registration point that obtains to carrying out the transformation model parameter estimation, and input picture is carried out coordinate transform and resample finishing thick registration with thick registration control points;
Step 2: metric space feature extraction and a large amount of right coupling of registration control points:
The metric space feature extraction: the input picture and the reference picture that obtain behind the thick registration are set up its multilayer metric space respectively, extract a some feature in the Corner Detection Algorithm of each layer of described metric space utilization Harris, with described as candidate point;
A large amount of right couplings of registration control points: adopting mutual information as similarity measure candidate point in input picture and the reference picture to be carried out the window coupling, to obtain a large amount of reference mark right;
Step 3: based on the screening and the meticulous registration of control point information amount:
Screening based on the control point information amount: at first respectively the public domain of the input picture that comprises the reference mark behind the thick registration and the public domain that comprises the reference picture at reference mark are divided into 3 * 3 subregion, respectively the reference mark in each subregion is screened again afterwards, thus the reference picture that input picture that controlled respectively point is evenly distributed and reference mark are evenly distributed;
Meticulous registration: to laggard line translation model parameter estimation, and the right input picture in reference mark that comprises behind the thick registration carried out coordinate transform and resample finishing meticulous registration at the reference mark that obtains to be evenly distributed.
The present invention a kind ofly carries out the high-precision and steady registration method to the multiple sensors remote sensing images, be data set-up procedure indispensable in the multisensor remote sensing application disposal system, this invention has solved that a kind of feature of traditional simple employing mates can not be correctly (panchromatic from the multisensor remote sensing images, infrared, multispectral, high spectrum, synthetic aperture radar image-forming etc.) extract the right problem in reference mark, and based on the reference mark quantity of information has been proposed the equally distributed strategy in a kind of reference mark, thereby utilized the high registration accuracy that realizes the multisensor remote sensing images to smart mode by thick.
Description of drawings
Fig. 1 is the schematic flow sheet of thick registration of the present invention; Fig. 2 is the search synoptic diagram of similarity search corresponding point with the mutual information for what adopt, and A represents reference picture, and B represents input picture, and a represents search window, and b represents to estimate window; Fig. 3 is based on the screening of control point information amount and meticulous registration schematic flow sheet; Fig. 4 is screening and the meticulous registration schematic flow sheet based on the control point information amount that has screening process.
Embodiment
Embodiment one: in conjunction with Fig. 1 to Fig. 3 present embodiment is described, the step of present embodiment is as follows:
Step 1: the thick registration of uniting a feature and provincial characteristics:
Choosing of some feature: (Scale Invariant Feature Transform SIFT) extracts from reference picture and input picture respectively for the key point of mating the conversion of utilization yardstick invariant features; Key point in the utilization Euclidean distance calculating reference picture and the distance between the key point in the input picture, it is right as the thick registration point of candidate to choose three pairs of minimum in described distance key points;
Choosing of provincial characteristics: utilization maximum stable extremal region (Maximally Stable Exremal Region, MSER) extract zone in reference picture and the input picture respectively, adopt invariant moments that two-way coupling is carried out in zone in the reference picture and the zone in the input picture and obtain three pairs of the most similar zones, with the center of gravity in zone as the thick registration control points of candidate;
Thick registration: unite the thick registration point that obtains to carrying out the transformation model parameter estimation, and input picture is carried out coordinate transform and resample finishing thick registration with thick registration control points;
Unite the thick registration of a feature and provincial characteristics, purpose is to eliminate the difference of yardstick bigger between reference picture and the input picture, rotation and translation, for the smart registration of back is prepared; The advantage of associating SIFT and MSER is and can keeps good robustness to combine to the image pixel nonlinearities change provincial characteristics of accurate match point feature of SIFT and MSER extraction, finishes thick registration process.
Step 2: metric space feature extraction and a large amount of right coupling of registration control points:
The metric space feature extraction: the input picture and the reference picture that obtain behind the thick registration are set up its multilayer metric space respectively, extract a some feature in the Corner Detection Algorithm of each layer of described metric space utilization Harris, with described as candidate point;
A large amount of right couplings of registration control points: adopt mutual information (Mutual Information, MI) as similarity measure candidate point in input picture and the reference picture being carried out the window coupling, to obtain a large amount of reference mark right;
Thick registration adopts registration control points limited to number, and what finish is rough registration, and in order to reach high-precision registration, it is right to need to extract more control point, and the distribution at reference mark is effectively controlled.It is right that the coupling that metric space feature extraction and registration control points are right, purpose are to extract the required a large amount of reference mark of meticulous registration;
Step 3: based on the screening and the meticulous registration of control point information amount:
Screening based on the control point information amount: at first respectively the public domain of the input picture that comprises the reference mark behind the thick registration and the public domain that comprises the reference picture at reference mark are divided into 33 subregion, respectively the reference mark in each subregion is screened again afterwards, thus the reference picture that input picture that controlled respectively point is evenly distributed and reference mark are evenly distributed;
The process of screening is:
Whether the distributed mass of judging all reference mark in the subregion satisfies threshold value;
Do not satisfy threshold value, then delete the reference mark of the quantity of information minimum at reference mark in this zone, rejudge this regional distributed mass again and whether satisfy threshold value, until satisfying threshold value; Satisfy threshold value, then all reference mark in the current region is evenly distributed.
With each subregion in described reference picture and the input picture all judge finish after, respectively equally distributed reference picture of controlled point and the equally distributed reference picture in reference mark;
Judgement is respectively the size of the quantity of information at a large amount of reference mark in each subregion in input picture and the reference picture to be screened, so that the distributed mass at each subregion reference mark satisfies the threshold value requirement in input picture and the reference picture, thus controlled input picture and reference picture to being evenly distributed; Just the distributed mass at reference mark in each subregion is calculated,, deleted the little reference mark of quantity of information one by one to meeting the requirements if do not meet setting threshold;
Meticulous registration: to laggard line translation model parameter estimation, and the right input picture in reference mark that comprises behind the thick registration carried out coordinate transform and resample finishing meticulous registration at the reference mark that obtains to be evenly distributed.
Based on the screening and the meticulous registration of control point information amount, purpose is the reference mark is screened by each self-contained quantity of information, finishes high-precision meticulous registration.The reference mark is to make the reference mark to evenly distributing in reference picture and input picture to the screening purpose, improves the precision of registration.
Embodiment two: when present embodiment and embodiment one difference were to adopt in the step 1 Euclidean distance to calculate three pairs of minimum key points, the setting threshold scope was 0.5~0.7; Wherein the Euclidean distance optimal threshold is 0.58,0.6 or 0.62.Other step is identical with embodiment one.
Embodiment three: when present embodiment and embodiment one difference were in the step 1 to adopt invariant moments to judge similar area, the setting threshold scope was 0.7~0.9, and wherein the optimal threshold of She Dinging is 0.78,0.8 or 0.82.Other step is identical with embodiment one.
Embodiment four: present embodiment and embodiment one difference are that the described transformation model in the step 1 adopts the affined transformation model, and with least square method solving model parameter; The bilinearity difference is adopted in described resampling; Other step is identical with embodiment one.
Embodiment five: in conjunction with Fig. 2 present embodiment is described, present embodiment and embodiment one to four difference are that the matching process that a large amount of registration control points are right in the step 2 is as follows:
Step a: the setting search window, in reference picture, find corresponding position for a candidate point to be matched in the input imagery, as the search window center, the setting size is 51 * 51 search window with described position;
Step b: set estimating window, is that the center is set one 41 * 41 and estimated window with candidate point to be matched in input picture, is that the center is set a plurality of 41 * 41 and estimated windows with each candidate point in the search window in reference picture;
Step c: the gray scale similarity is calculated, with the estimation window in the input picture one by one with reference picture in each estimate that window carries out the gray scale calculation of similarity degree;
Steps d: the gray scale similarity of acquisition and the gray scale similarity threshold of setting are compared;
If the gray scale similarity that obtains during less than the gray scale similarity threshold set, is then deleted the candidate point of the estimation window center in the corresponding reference image;
If, then expanding the search window in the reference picture to 101 * 101 sizes all less than the gray scale similarity threshold of setting, all gray scale similarities that obtain continue execution in step b and step c;
If all gray scale similarities that expand 101 * 101 size back acquisitions to are then abandoned the coupling of the candidate point to be matched in the input imagery also all less than the gray scale similarity threshold of setting;
Otherwise, to select the match point of the candidate point of the estimation window center in the wherein maximum gray scale similarity corresponding reference image as the candidate point to be matched in the input picture in all gray scale similarities greater than the gray scale similarity threshold that obtain, it is right to obtain a reference mark.
Other step is identical with embodiment one to four.
Embodiment six: in conjunction with Fig. 4 present embodiment is described, the difference of present embodiment and embodiment one is step 3, in the step 3 in each subregion the distributed mass at reference mark be based on the reference mark quantity of information calculated:
The following computing method that provide for each selected control point information amount:
For a selected reference mark, the descriptor that defines this point is the one group gray scale rotational invariants of this point for the local window at center, adopts two invariant of order P Ps to represent descriptor:
Formula one: υ → [ 0 . . 3 ] = L x L x + L y L y L xx L x L x + 2 L xy L x L y + L yy L y L y L xx + L yy L xx L xx + 2 L xy L xy + L yy L yy - - - ( 1 )
Vector space wherein
Figure A20091030763400102
First be shade of gray square, the 3rd is Laplce's gradient;
The quantity of information at reference mark is represented with the entropy of described descriptor:
The calculating of entropy need be to vector space
Figure A20091030763400103
Cut apart,
Cut apart distance with Manhalanobis, and
Figure A20091030763400104
Because Λ is decomposed into Λ -1=P TDP,
Wherein D is a diagonal matrix, and P is an orthogonal matrix,
So the Manhalanobis distance is converted into again
Figure A20091030763400105
Apart from d MBe the mean value of standardized vector:
Figure A20091030763400106
The descriptor vector just uses the grid cell with size to come the compute vector space in all directions through after the standardization Entropy;
Obtain the quantity of information of representing the reference mark of entropy by aforementioned calculation, calculate reference mark distributed mass in each subregion, with the mode of following distributed mass (Distribution Quality, DQ) calculate:
Formula two: DQ = Σ i = 1 n ( x i - x ‾ ) 2 + Σ i = 1 n ( y i - y ‾ ) 2 n / ( M + N ) - - - ( 2 )
Formula three: ( x ‾ , y ‾ ) = ( Σ i = 1 n w i x i Σ i = 1 n w i , Σ i = 1 n w i y i Σ i = 1 n w i ) - - - ( 3 )
M in the formula two, N are the row and column of zonule, w in the formula three iGet the quantity of information of reference mark i, and the quantity of information that comprises with reference mark i is as weight, thereby calculates reference mark distributed mass in each subregion;
Reference mark distributed mass and preset threshold in described each subregion are compared,, then enter next step if meet threshold value; If do not meet threshold value, then delete the quantity of information at reference mark is little in the current region reference mark to recomputating distributed mass until meeting threshold value; The setting threshold scope is 2.0~0.3, and wherein the optimal threshold of She Dinging is 0.22,0.24 or 0.26.Other step is identical with embodiment one.
Content of the present invention is not limited only to the content of the respective embodiments described above, and the combination of one of them or several embodiments equally also can realize the purpose of inventing.

Claims (9)

1. the multiple sensors remote sensing images are carried out the method for high-precision and steady registration, it is characterized in that its step is as follows:
Step 1: the thick registration of uniting a feature and provincial characteristics:
Choosing of some feature: the key point for coupling is extracted in the conversion of utilization yardstick invariant features respectively from reference picture and input picture; Key point in the utilization Euclidean distance calculating reference picture and the distance between the key point in the input picture, it is right as the thick registration point of candidate to choose three pairs of minimum in described distance key points;
Choosing of provincial characteristics: the utilization maximum stable extremal region extracts the zone in reference picture and the input picture respectively, adopt invariant moments that two-way coupling is carried out in zone in the reference picture and the zone in the input picture and obtain three pairs of the most similar zones, with the center of gravity in zone as the thick registration control points of candidate;
Thick registration: unite the thick registration point that obtains to carrying out the transformation model parameter estimation, and input picture is carried out coordinate transform and resample finishing thick registration with thick registration control points;
Step 2: metric space feature extraction and a large amount of right coupling of registration control points:
The metric space feature extraction: the input picture and the reference picture that obtain behind the thick registration are set up its multilayer metric space respectively, extract a some feature in the Corner Detection Algorithm of each layer of described metric space utilization Harris, with described as candidate point;
A large amount of right couplings of registration control points: adopting mutual information as similarity measure candidate point in input picture and the reference picture to be carried out the window coupling, to obtain a large amount of reference mark right;
Step 3: based on the screening and the meticulous registration of control point information amount:
Screening based on the control point information amount: at first respectively the public domain of the input picture that comprises the reference mark behind the thick registration and the public domain that comprises the reference picture at reference mark are divided into 3 * 3 subregion, respectively the reference mark in each subregion is screened again afterwards, thus the reference picture that input picture that controlled respectively point is evenly distributed and reference mark are evenly distributed;
Meticulous registration: to laggard line translation model parameter estimation, and the right input picture in reference mark that comprises behind the thick registration carried out coordinate transform and resample finishing meticulous registration at the reference mark that obtains to be evenly distributed.
2. according to claim 1 the multiple sensors remote sensing images are carried out the method for high-precision and steady registration, it is characterized in that in the right matching process of a large amount of registration control points described in the step 2, obtain a right process in reference mark and be:
Step a: the setting search window, in reference picture, find corresponding position for a candidate point to be matched in the input imagery, as the search window center, the setting size is 51 * 51 search window with described position;
Step b: set estimating window, is that the center is set one 41 * 41 and estimated window with candidate point to be matched in input picture, is that the center is set a plurality of 41 * 41 and estimated windows with each candidate point in the search window in reference picture;
Step c: gray scale similarity ρ calculates, with the estimation window in the input picture one by one with reference picture in each estimate that window carries out the calculating of gray scale similarity ρ;
Steps d: with the gray scale similarity ρ of acquisition and the gray scale similarity threshold ρ of setting ThCompare;
If the gray scale similarity ρ that obtains is less than the gray scale similarity threshold ρ that sets ThThe time, then delete the candidate point of the estimation window center in the corresponding reference image;
If all gray scale similarity ρ that obtain are less than the gray scale similarity threshold ρ that sets Th, then the search window in the reference picture is expanded to 101 * 101 sizes and continues execution in step b and step c;
If all gray scale similarity ρ that expand 101 * 101 size back acquisitions to are also less than the gray scale similarity threshold ρ that sets Th, then abandon the coupling of the candidate point to be matched in the input imagery;
Otherwise, with obtain all greater than gray scale similarity threshold ρ ThGray scale similarity ρ in select the match point of the candidate point of the estimation window center in the wherein maximum gray scale similarity ρ corresponding reference image as the candidate point to be matched in the input picture, it is right to obtain a reference mark.
3. according to claim 1 and 2 the multiple sensors remote sensing images are carried out the method for high-precision and steady registration, it is characterized in that the process of in the step 3 reference mark in each subregion being screened is: whether the distributed mass of judging all reference mark in the subregion satisfies threshold value; Do not satisfy threshold value, then delete the reference mark of the quantity of information minimum at reference mark in this zone, rejudge this regional distributed mass again and whether satisfy threshold value, until satisfying threshold value; Satisfy threshold value, then all reference mark in the current region is evenly distributed.
4. according to claim 3 the multiple sensors remote sensing images are carried out the method for high-precision and steady registration, it is characterized in that the computing method of the quantity of information at described reference mark:
For a selected reference mark, the descriptor that defines this point is the one group gray scale rotational invariants of this point for the local window at center, adopts two invariant of order P Ps to represent descriptor:
Formula one: υ → [ 0 . . 3 ] = L x L x + L y L y L xx L x L x + 2 L xy L x L y + L yy L y L y L xx + L yy L xx L xx + 2 L xy L xy + L yy L yy - - - ( 1 )
Vector space wherein
Figure A2009103076340004C2
First be shade of gray square, the 3rd is Laplce's gradient;
The quantity of information at reference mark is represented with the entropy of described descriptor:
The calculating of entropy need be to vector space
Figure A2009103076340004C3
Cut apart,
Cut apart distance with Manhalanobis, and d M ( υ → 1 , υ → 2 ) = ( υ → 1 - υ → 2 ) T Λ - 1 ( υ → 1 - υ → 2 ) ,
Because Λ is decomposed into Λ -1=P TDP, wherein D is a diagonal matrix, P is an orthogonal matrix,
So the Manhalanobis distance is converted into again d M ( υ → 1 , υ → 2 ) = | | D 1 / 2 P ( υ → 1 - υ → 2 ) | | ,
Apart from d MBe the mean value of standardized vector: υ → norm = D 1 / 2 P υ → ;
The descriptor vector just uses the grid cell with size to come the compute vector space in all directions through after the standardization
Figure A2009103076340004C7
Entropy;
Obtain the quantity of information of representing the reference mark of entropy by aforementioned calculation.
5. according to claim 3 the multiple sensors remote sensing images are carried out the method for high-precision and steady registration, it is characterized in that the distributed mass discussed with as the mode of getting off calculate:
Formula two: DQ = Σ i = 1 n ( x i - x ‾ ) 2 + Σ i = 1 n ( y i - y ‾ ) 2 n / ( M + N ) - - - ( 2 )
Formula three: ( x ‾ , y ‾ ) = ( Σ i = 1 n w i x i Σ i = 1 n w i , Σ i = 1 n w i y i Σ i = 1 n w i ) - - - ( 3 )
M in the formula two, N are the row and column of subregion, w in the formula three iGet the quantity of information of reference mark i, and the quantity of information that comprises with reference mark i is as weight, thereby calculates the distributed mass at all reference mark in the subregion.
6. according to claim 1 or 5 the multiple sensors remote sensing images are carried out the method for high-precision and steady registration, when it is characterized in that adopting in the step 1 Euclidean distance to calculate three pairs of minimum key points, the setting threshold scope is 0.5~0.7; Wherein the Euclidean distance optimal threshold is 0.58,0.6 or 0.62.
7. according to claim 1 or 5 the multiple sensors remote sensing images are carried out the method for high-precision and steady registration, when it is characterized in that in the step 1 adopting invariant moments to judge similar area, the setting threshold scope is 0.7~0.9, and wherein the optimal threshold of She Dinging is 0.78,0.8 or 0.82.
8. according to claim 1 or 5 the multiple sensors remote sensing images are carried out the method for high-precision and steady registration, it is characterized in that the described transformation model in the step 1 adopts the affined transformation model, and with least square method solving model parameter; The bilinearity difference is adopted in described resampling.
9. according to claim 3 or 5 described methods of the multiple sensors remote sensing images being carried out high-precision and steady registration, when it is characterized in that judging the distributed mass at all reference mark in the subregion in the step 3, the setting threshold scope is 2.0~0.3, and wherein the optimal threshold of She Dinging is 0.22,0.24 or 0.26.
CN200910307634A 2009-09-24 2009-09-24 Method for conducting high-precision and steady registration on diversified sensor remote sensing images Pending CN101667293A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN200910307634A CN101667293A (en) 2009-09-24 2009-09-24 Method for conducting high-precision and steady registration on diversified sensor remote sensing images

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN200910307634A CN101667293A (en) 2009-09-24 2009-09-24 Method for conducting high-precision and steady registration on diversified sensor remote sensing images

Publications (1)

Publication Number Publication Date
CN101667293A true CN101667293A (en) 2010-03-10

Family

ID=41803905

Family Applications (1)

Application Number Title Priority Date Filing Date
CN200910307634A Pending CN101667293A (en) 2009-09-24 2009-09-24 Method for conducting high-precision and steady registration on diversified sensor remote sensing images

Country Status (1)

Country Link
CN (1) CN101667293A (en)

Cited By (30)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101922930A (en) * 2010-07-08 2010-12-22 西北工业大学 Aviation polarization multi-spectrum image registration method
CN102005047A (en) * 2010-11-15 2011-04-06 无锡中星微电子有限公司 Image registration system and method thereof
CN102194225A (en) * 2010-03-17 2011-09-21 中国科学院电子学研究所 Automatic registering method for coarse-to-fine space-borne synthetic aperture radar image
CN102208038A (en) * 2011-06-27 2011-10-05 清华大学 Image classification method based on visual dictionary
CN102495407A (en) * 2011-11-14 2012-06-13 电子科技大学 Characterization method of similarity of polarized synthetic aperture radar image
CN102609960A (en) * 2012-02-16 2012-07-25 四川大学 Method for extracting target area
CN102693533A (en) * 2012-03-12 2012-09-26 清华大学 Medical digital image mosaicing method
CN102842137A (en) * 2012-08-14 2012-12-26 中山大学 Automatic registration method for multi-temporal empty spectrum remote sensing image based on space comprehensive mutual information
CN103020945A (en) * 2011-09-21 2013-04-03 中国科学院电子学研究所 Remote sensing image registration method of multi-source sensor
CN103198483A (en) * 2013-04-07 2013-07-10 西安电子科技大学 Multiple time phase remote sensing image registration method based on edge and spectral reflectivity curve
CN103295222A (en) * 2013-02-21 2013-09-11 南京理工大学 Implement method and device for infrared image registration
CN103310443A (en) * 2013-05-20 2013-09-18 华浩博达(北京)技术有限公司 Quick processing method and system for high-resolution remote sensing image
CN103390272A (en) * 2013-07-16 2013-11-13 西安应用光学研究所 Method for achieving registration and fusion of multi-spectral pseudo color images
CN103400384A (en) * 2013-07-22 2013-11-20 西安电子科技大学 Large viewing angle image matching method capable of combining region matching and point matching
CN103473765A (en) * 2013-09-03 2013-12-25 深圳大学 Method and system for registering and rectifying multispectral remote sensing images
CN103489156A (en) * 2012-06-09 2014-01-01 北京国药恒瑞美联信息技术有限公司 Digital X-ray image processing method
CN103646399A (en) * 2013-12-04 2014-03-19 中国资源卫星应用中心 Registering method for optical and radar images
CN104331899A (en) * 2014-11-24 2015-02-04 中国科学院电子学研究所 Registration method and device for SAR image
CN103473765B (en) * 2013-09-03 2016-11-30 深圳大学 The registration of a kind of multi-spectral remote sensing image and bearing calibration, system
CN106548491A (en) * 2016-09-30 2017-03-29 深圳大学 A kind of method for registering images, its image interfusion method and its device
CN106960449A (en) * 2017-03-14 2017-07-18 西安电子科技大学 The heterologous method for registering constrained based on multiple features
CN107004267A (en) * 2014-12-15 2017-08-01 皇家飞利浦有限公司 Quality of image registration is controlled
CN107240130A (en) * 2017-06-06 2017-10-10 苍穹数码技术股份有限公司 Remote Sensing Image Matching method, apparatus and system
CN107610166A (en) * 2017-09-26 2018-01-19 上海海事大学 A kind of plane map method for registering images based on linear character region segmentation
CN108335320A (en) * 2017-01-20 2018-07-27 中电科海洋信息技术研究院有限公司 The spatial registration method and spatial registration system of multi-source Remote Sensing Images
CN110058233A (en) * 2019-04-28 2019-07-26 电子科技大学 A kind of anti-duplicity interference method of multistatic SARS system
CN110136083A (en) * 2019-05-14 2019-08-16 深圳大学 A kind of the base map update method and device of combination interactive mode
CN110210511A (en) * 2019-04-19 2019-09-06 哈尔滨工业大学 A kind of improvement PCA-SIFT method for registering images based on cosine measure
CN111060910A (en) * 2019-12-11 2020-04-24 西安电子科技大学 InSAR carrier reverse positioning based on terrain-image matching
CN112288009A (en) * 2020-10-29 2021-01-29 西安电子科技大学 R-SIFT chip hardware Trojan horse image registration method based on template matching

Cited By (47)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102194225A (en) * 2010-03-17 2011-09-21 中国科学院电子学研究所 Automatic registering method for coarse-to-fine space-borne synthetic aperture radar image
CN101922930B (en) * 2010-07-08 2013-11-06 西北工业大学 Aviation polarization multi-spectrum image registration method
CN101922930A (en) * 2010-07-08 2010-12-22 西北工业大学 Aviation polarization multi-spectrum image registration method
CN102005047A (en) * 2010-11-15 2011-04-06 无锡中星微电子有限公司 Image registration system and method thereof
CN102005047B (en) * 2010-11-15 2012-09-26 无锡中星微电子有限公司 Image registration system and method thereof
CN102208038A (en) * 2011-06-27 2011-10-05 清华大学 Image classification method based on visual dictionary
CN103020945B (en) * 2011-09-21 2016-02-10 中国科学院电子学研究所 A kind of remote sensing image registration method of Multiple Source Sensor
CN103020945A (en) * 2011-09-21 2013-04-03 中国科学院电子学研究所 Remote sensing image registration method of multi-source sensor
CN102495407B (en) * 2011-11-14 2013-03-13 电子科技大学 Characterization method of similarity of polarized synthetic aperture radar image
CN102495407A (en) * 2011-11-14 2012-06-13 电子科技大学 Characterization method of similarity of polarized synthetic aperture radar image
CN102609960A (en) * 2012-02-16 2012-07-25 四川大学 Method for extracting target area
CN102609960B (en) * 2012-02-16 2014-01-08 四川大学 Method for extracting target area
CN102693533A (en) * 2012-03-12 2012-09-26 清华大学 Medical digital image mosaicing method
CN103489156A (en) * 2012-06-09 2014-01-01 北京国药恒瑞美联信息技术有限公司 Digital X-ray image processing method
CN103489156B (en) * 2012-06-09 2016-08-17 北京国药恒瑞美联信息技术有限公司 A kind of processing method of digital X-ray
CN102842137A (en) * 2012-08-14 2012-12-26 中山大学 Automatic registration method for multi-temporal empty spectrum remote sensing image based on space comprehensive mutual information
CN102842137B (en) * 2012-08-14 2015-04-22 中山大学 Automatic registration method for multi-temporal empty spectrum remote sensing image based on space comprehensive mutual information
CN103295222A (en) * 2013-02-21 2013-09-11 南京理工大学 Implement method and device for infrared image registration
CN103198483A (en) * 2013-04-07 2013-07-10 西安电子科技大学 Multiple time phase remote sensing image registration method based on edge and spectral reflectivity curve
CN103198483B (en) * 2013-04-07 2015-09-30 西安电子科技大学 Based on the Multitemporal Remote Sensing Images method for registering of edge and spectral reflectance rate curve
CN103310443A (en) * 2013-05-20 2013-09-18 华浩博达(北京)技术有限公司 Quick processing method and system for high-resolution remote sensing image
CN103310443B (en) * 2013-05-20 2016-04-27 华浩博达(北京)科技股份有限公司 High-resolution remote sensing image immediate processing method and system
CN103390272A (en) * 2013-07-16 2013-11-13 西安应用光学研究所 Method for achieving registration and fusion of multi-spectral pseudo color images
CN103390272B (en) * 2013-07-16 2016-01-13 西安应用光学研究所 Realize the registration fusion method of multispectral pseudo color image
CN103400384A (en) * 2013-07-22 2013-11-20 西安电子科技大学 Large viewing angle image matching method capable of combining region matching and point matching
CN103400384B (en) * 2013-07-22 2016-05-25 西安电子科技大学 The wide-angle image matching process of calmodulin binding domain CaM coupling and some coupling
CN103473765B (en) * 2013-09-03 2016-11-30 深圳大学 The registration of a kind of multi-spectral remote sensing image and bearing calibration, system
CN103473765A (en) * 2013-09-03 2013-12-25 深圳大学 Method and system for registering and rectifying multispectral remote sensing images
CN103646399A (en) * 2013-12-04 2014-03-19 中国资源卫星应用中心 Registering method for optical and radar images
CN104331899B (en) * 2014-11-24 2018-06-19 中国科学院电子学研究所 A kind of SAR image registration method and device
CN104331899A (en) * 2014-11-24 2015-02-04 中国科学院电子学研究所 Registration method and device for SAR image
CN107004267B (en) * 2014-12-15 2021-02-09 皇家飞利浦有限公司 Image registration quality control
CN107004267A (en) * 2014-12-15 2017-08-01 皇家飞利浦有限公司 Quality of image registration is controlled
CN106548491B (en) * 2016-09-30 2019-09-27 深圳大学 A kind of method for registering images, its image interfusion method and its device
CN106548491A (en) * 2016-09-30 2017-03-29 深圳大学 A kind of method for registering images, its image interfusion method and its device
CN108335320A (en) * 2017-01-20 2018-07-27 中电科海洋信息技术研究院有限公司 The spatial registration method and spatial registration system of multi-source Remote Sensing Images
CN106960449B (en) * 2017-03-14 2020-02-14 西安电子科技大学 Heterogeneous registration method based on multi-feature constraint
CN106960449A (en) * 2017-03-14 2017-07-18 西安电子科技大学 The heterologous method for registering constrained based on multiple features
CN107240130A (en) * 2017-06-06 2017-10-10 苍穹数码技术股份有限公司 Remote Sensing Image Matching method, apparatus and system
CN107610166A (en) * 2017-09-26 2018-01-19 上海海事大学 A kind of plane map method for registering images based on linear character region segmentation
CN110210511A (en) * 2019-04-19 2019-09-06 哈尔滨工业大学 A kind of improvement PCA-SIFT method for registering images based on cosine measure
CN110058233A (en) * 2019-04-28 2019-07-26 电子科技大学 A kind of anti-duplicity interference method of multistatic SARS system
CN110058233B (en) * 2019-04-28 2021-09-14 电子科技大学 Deception jamming resisting method of multi-base synthetic aperture radar system
CN110136083A (en) * 2019-05-14 2019-08-16 深圳大学 A kind of the base map update method and device of combination interactive mode
CN111060910A (en) * 2019-12-11 2020-04-24 西安电子科技大学 InSAR carrier reverse positioning based on terrain-image matching
CN111060910B (en) * 2019-12-11 2023-08-29 西安电子科技大学 InSAR carrier reverse positioning based on topography-image matching
CN112288009A (en) * 2020-10-29 2021-01-29 西安电子科技大学 R-SIFT chip hardware Trojan horse image registration method based on template matching

Similar Documents

Publication Publication Date Title
CN101667293A (en) Method for conducting high-precision and steady registration on diversified sensor remote sensing images
CN101251926B (en) Remote sensing image registration method based on local configuration covariance matrix
CN104574421B (en) Large-format small-overlap-area high-precision multispectral image registration method and device
CN100587518C (en) Method for automatically selecting remote sensing image high-precision control point
CN101980293B (en) Method for detecting MTF of hyperspectral remote sensing system based on edge image
CN107563438A (en) The multi-modal Remote Sensing Images Matching Method and system of a kind of fast robust
CN104751465A (en) ORB (oriented brief) image feature registration method based on LK (Lucas-Kanade) optical flow constraint
CN101826206B (en) Camera self-calibration method
CN105354841B (en) A kind of rapid remote sensing image matching method and system
CN102750696B (en) Affine invariant feature and coastline constraint-based automatic coastal zone remote-sensing image registration method
CN102254319A (en) Method for carrying out change detection on multi-level segmented remote sensing image
CN103400388A (en) Method for eliminating Brisk (binary robust invariant scale keypoint) error matching point pair by utilizing RANSAC (random sampling consensus)
CN105139375B (en) Combining global DEM and stereoscopic vision a kind of satellite image cloud detection method of optic
CN101556695A (en) Image matching method
CN103440659B (en) Based on starry sky image Distortion Detect and the method for estimation of star pattern matching
CN105654423B (en) Remote sensing image registration method based on region
CN106251362B (en) A kind of sliding window method for tracking target and system based on fast correlation neighborhood characteristics point
CN102236798A (en) Image matching method and device
CN103489193B (en) OO high-resolution remote sensing image change detecting method based on convergence strategy
CN101833767A (en) Gradient and color characteristics-based automatic straight line matching method in digital image
CN104077782A (en) Satellite-borne remote sense image matching method
CN104134208A (en) Coarse-to-fine infrared and visible light image registration method by adopting geometric construction characteristics
CN104680151B (en) A kind of panchromatic remote sensing image variation detection method of high-resolution for taking snow covering influence into account
CN103679720A (en) Fast image registration method based on wavelet decomposition and Harris corner detection
CN104517124B (en) SAR image change detection based on SIFT feature

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Open date: 20100310