CN105005962B - Islands and reefs Remote Sensing Image Matching method based on hierarchical screening strategy - Google Patents
Islands and reefs Remote Sensing Image Matching method based on hierarchical screening strategy Download PDFInfo
- Publication number
- CN105005962B CN105005962B CN201510500058.7A CN201510500058A CN105005962B CN 105005962 B CN105005962 B CN 105005962B CN 201510500058 A CN201510500058 A CN 201510500058A CN 105005962 B CN105005962 B CN 105005962B
- Authority
- CN
- China
- Prior art keywords
- islands
- reefs
- mrow
- msub
- characteristic point
- 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.)
- Expired - Fee Related
Links
- 238000012216 screening Methods 0.000 title claims abstract description 57
- 238000000034 method Methods 0.000 title claims abstract description 31
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims abstract description 46
- 230000009466 transformation Effects 0.000 claims abstract description 22
- 239000011159 matrix material Substances 0.000 claims description 24
- 239000013598 vector Substances 0.000 claims description 14
- 238000013519 translation Methods 0.000 claims description 8
- 239000000284 extract Substances 0.000 claims description 7
- 230000003247 decreasing effect Effects 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 2
- 239000006185 dispersion Substances 0.000 claims 1
- 238000004519 manufacturing process Methods 0.000 abstract description 2
- 230000009977 dual effect Effects 0.000 abstract 1
- 238000000605 extraction Methods 0.000 description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- VAYOSLLFUXYJDT-RDTXWAMCSA-N Lysergic acid diethylamide Chemical compound C1=CC(C=2[C@H](N(C)C[C@@H](C=2)C(=O)N(CC)CC)C2)=C3C2=CNC3=C1 VAYOSLLFUXYJDT-RDTXWAMCSA-N 0.000 description 1
- 240000007594 Oryza sativa Species 0.000 description 1
- 235000007164 Oryza sativa Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 239000000428 dust Substances 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000000873 masking effect Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000011017 operating method Methods 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 235000009566 rice Nutrition 0.000 description 1
- 239000004576 sand Substances 0.000 description 1
- 239000007921 spray Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000031068 symbiosis, encompassing mutualism through parasitism Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 239000003643 water by type Substances 0.000 description 1
Classifications
-
- G06T3/147—
Abstract
The present invention relates to a kind of islands and reefs Remote Sensing Image Matching method based on hierarchical screening strategy, step includes：Affine invariants match；Establish geometric transformation model；Under the constraint of geometric transformation model, the first screening of affine invariants matching result is carried out；On the basis of first the selection result, the result of screening is further controlled using islands and reefs face domain degree of overlapping, guarantee screens out the vicious characteristic point pair of institute, is finally completed the registration of islands and reefs.The present invention considers that the distinctive textural characteristics of islands and reefs Remote Sensing Images Matching lack and the unstable dual challenge of textural characteristics, by establishing geometrical constraint model, screen out the obvious characteristic point pair for not meeting spatial distribution, then on the basis of geometric transformation constraint screening, using islands and reefs area degree of overlapping as constraint, to characteristic point to carrying out postsearch screening, only retain correct characteristic point to complete the accuracy registration of islands and reefs.The inventive method strong adaptability, islands and reefs Image registration can be accurately performed, disclosure satisfy that the needs of actual production.
Description
Technical field
The present invention relates to a kind of islands and reefs Remote Sensing Image Matching method, more particularly to a kind of islands and reefs based on hierarchical screening strategy
Remote sensing image affine invariants method for registering.
Background technology
Image registration is by the progress of two width or multiple image that are obtained under different time, different sensors or different condition
The process matched somebody with somebody, the fields such as remotely-sensed data analysis, computer vision, image procossing have been widely used in it.Islands and reefs registration should
Environmental monitoring, resource management and disaster alarm for islands and reefs etc..With the development in earth observation field, can obtain big
The islands and reefs remote sensing image under the same area difference image-forming condition is measured, the registration for these islands and reefs images is that islands and reefs are changed to detect
With the basis of analysis.
The image registration mode of feature based obtains extensive concern.When textural characteristics abundance, image registration is relatively easy；
When image lacks effective textural characteristics, automatic image registration is a difficulty and has the problem of challenge, in most cases
The result for needing artificial correction to match.Local invariant feature, particularly affine invariants, line can be overcome to a certain extent
The problem of reason lacks, because it is to there is masking in image, background is mixed and disorderly and content change has stronger robustness.MSER
(Maximally Stable External Region) operator be considered as optimal feature extraction operator (Mikolajczyk,
2005).SIFT (Scale Invariant Feature Transformation) description show in terms of feature description
The most outstanding (Mikolajczyk, 2005).With reference to both, referred to as affine invariants match, and extremely sane is applied to figure
The autoregistration of picture.In this technical foundation, further from RANSAC (Random Sample Consensus) algorithm to spy
Sign point can obtain more excellent matching effect to screening.But for islands and reefs Image Matching, directly in conjunction with MSER operators
Son is described with SIFT and carries out affine matching, further carries out RANSAC screenings, is often still difficult to obtain satisfied effect.This is
Caused by due to very big technical difficulty existing for islands and reefs Remote Sensing Images Matching, two aspects can be substantially summarized as：
1) islands and reefs image lacks textural characteristics.The waters that large area lacks textural characteristics, and water in islands and reefs image be present
Under reef platform also without obvious textural characteristics so that textural characteristics are concentrated mainly on the dust and sand island portion point of the exposure water surface.Lack
The image of textural characteristics, which is difficult to provide enough information, is used for Image registration.
2) islands and reefs textural characteristics are unstable.On the one hand, the dynamic change of islands and reefs environment can disturb the feature of islands and reefs image
Match somebody with somebody, the spray that the change of the water level of such as islands and reefs, wave strike reef disk are formed.On the other hand, islands and reefs are influenceed by artificial exploitation, by
There is important value in terms of military strategy and tourist resources in islands and reefs, the construction of the heavy construction such as airport and harbour and frequently
Reconstruction so that the islands and reefs remote sensing image information difference of different times is larger.Also, Part Development (such as marine reclamation land engineering) makes
The islands and reefs part shape that exposes the surface also occurs significantly to change with area, these factors show as image on remote sensing image
Textural characteristics it is unstable, cause to generate the characteristic points pair of a large amount of mistakes, the registration of severe jamming image.
The content of the invention
The technical problem to be solved in the present invention is：Overcome prior art shortcoming, propose a kind of based on hierarchical screening strategy
Islands and reefs Remote Sensing Image Matching method.
In order to solve the above technical problems, the islands and reefs Remote Sensing Image Matching side provided by the invention based on hierarchical screening strategy
Method, comprise the following steps：
The first step, affine invariants matching --- affine invariants point is extracted to two width islands and reefs remote sensing images, point by point
Characteristic vector is generated, the characteristic point of two width islands and reefs remote sensing images is matched according to characteristic vector, obtains some characteristic points pair；
Second step, establish geometric transformation model --- selection affine Transform Model is as geometric transformation model, with breaking naturally
Split method and extract islands and reefs border from two width islands and reefs remote sensing images respectively, and resolve to obtain according to two islands and reefs boundary informations of acquisition
The model parameter of affine Transform Model；
3rd step, first screening characteristic point are right --- according to described affine Transform Model by a spy of characteristic point centering
Sign point carries out affine transformation, the Euclidean distance between matched characteristic point is calculated after conversion, the error as characteristic point pair
Value, the characteristic point pair for accounting for total sample number 8% -12% is chosen with error amount from small to large, and error threshold gradually increases, constantly meter
Calculate error amount and be less than the characteristic point centering of the error threshold and belong to the characteristic point of main image and form the spatial cohesion of point set,
When spatial cohesion reaches maximum, corresponding error threshold screens out error amount more than the screening threshold value as screening threshold value
Characteristic point pair；
4th step, postsearch screening characteristic point are right --- using islands and reefs area Duplication as constraints, changed using RANSAC
For algorithm to characteristic point to screening；
5th step, islands and reefs Remote Sensing Image Matching --- after the screening of the 3rd step and the 4th step, according to the characteristic point of reservation
To resolving homography matrix, with the registration for resolving the homography matrix completion islands and reefs remote sensing image obtained.
In order to solve the above technical problems, the present invention also has feature further below：
1st, in the first step, the number of feature points of generation is not less than 4..
2nd, in the second step, affine transformation matrix model obtains according to following formula：
Above parameter can resolve according to the islands and reefs boundary information of two width images, wherein, α is the anglec of rotation, m1And m2Join for ratio
Number, x0And y0For translation parameters.
3rd, in the first step, affine invariants point, Ran Houyong are extracted to two width islands and reefs remote sensing images using MSER operators
Affine invariants point is described SIFT description, the dimensional feature vector of pixel-level generation 128.
4th, in the 3rd step, the characteristic point that preceding n error amount minimal characteristic point centering belongs to main image forms point set
Spatial cohesion SDQnObtained according to following formula：
Wherein,The weighted center point coordinates of n error amount minimal characteristic point pair, (x before expressioni,yi) for the
I error minimal characteristic point centering belongs to the coordinate of the characteristic point of main image.
5th, in the 4th step, using islands and reefs area Duplication as constraints, using RANSAC iterative algorithms to feature
Point is as follows to the specific method screened：
A, first time RANSAC iteration as screening parameter and is started using RANSAC initial thresholds, RANSAC initial thresholds
Value is 0.01；
B, RANSAC threshold values are as screening parameter, using RANSAC algorithms by characteristic point to dividing into intra-office point pair and not in the know
Point pair, screens out point pair not in the know, according to intra-office point to solving homography matrix, and enters line translation to main image using homography matrix, counts
Calculate the area Duplication on the island of image and island in image subject to registration after converting；
C, RANSAC threshold values, and repeat step b are incrementally increased, when islands and reefs area Duplication is decreased obviously, is stopped
RANSAC iteration；
D, a RANSAC threshold value before islands and reefs area Duplication is decreased obviously is optimal RANSAC threshold values, screens out foundation
The point pair not in the know that the optimal RANSAC threshold values are separated out.
6th, in the 4th step, second of iteration and RANSAC threshold values afterwards obtain according to following formula：
Wherein, TRcurrentFor the RANSAC threshold values of current iteration, TRpreviousFor the RANSAC threshold values of last iteration.
7th, in the 4th step, islands and reefs area Duplication Ratio obtains according to following formula：
Wherein, RμaFor with reference to islands and reefs face domain area,To convert islands and reefs face domain area.
Beneficial effects of the present invention are as follows：
1), the present invention is rare and the characteristics of textural characteristics are unstable for islands and reefs remote sensing image textural characteristics, it is affine not
Become on the basis of characteristic matching, it is proposed that from " spatial distribution control " to the screening strategy twice of " control of islands and reefs degree of overlapping ", from
And all error characteristic points pair are removed, complete the registration of islands and reefs remote sensing image.
2) it is, of the invention to propose by establishing the scalping choosing method of geometric transformation model discrimination characteristic point pair, and combine
Characteristic point determines screening threshold value to number and spatial cohesion, and the Space Consistency of characteristic point pair is verified on the whole,
The inconsistent error characteristic point pair of spatial distribution is eliminated, improves the correctness of subsequent match.
3), select permeability of the present invention for RANSAC screening threshold values, it is proposed that according to the control of islands and reefs face domain Duplication certainly
The dynamic optimal screening threshold method of selection, it is ensured that the reliability of the selection result, ensure the precision of final islands and reefs remote sensing image.
4), the inventive method strong adaptability, it is proven, can accurately registering islands and reefs remote sensing using the inventive method
Image, the needs of actual production are disclosure satisfy that, there is stronger practicality compared to traditional Remote Sensing Image Matching method.
Brief description of the drawings
The present invention is further illustrated below in conjunction with the accompanying drawings.
Fig. 1 is method for registering flow chart of the present invention.
Fig. 2 is test block bitmap area and raw video figure in the embodiment of the present invention.
Fig. 3-a are the south of China in 2005 sub-island remote sensing image.
Fig. 3-b are the south of China in 2009 sub-island remote sensing image.
Fig. 3-c are the south of China in 2005 sub-island image islands and reefs Boundary Extraction result.
Fig. 3-d are the south of China in 2009 sub-island image islands and reefs Boundary Extraction result.
Fig. 3-e are that the south of China in 2005 sub-island image affine invariants point extracts result.
Fig. 3-f are that the south of China in 2009 sub-island image affine invariants point extracts result.
Fig. 3-g are China south sub-island image affine invariants matching result.
Fig. 3-h are that China south sub-island image carries out the result after screening for the first time.
Fig. 3-i are that China south sub-island image carries out the result after the postsearch screening that RANSAC threshold values are 0.6.
Fig. 3-j are that China south sub-island image carries out the result after the postsearch screening that RANSAC threshold values are 0.7.
Fig. 4 is islands and reefs overlapping area ratio with RANSAC changes of threshold figures.
Embodiment
The present invention is elaborated below according to accompanying drawing, the technology path and operating procedure for making the present invention become apparent from.
As shown in Fig. 2 southern sub-island of the embodiment selection positioned at China South Sea is east longitude 114 degree 20 as research area, longitude
Point, latitude is 26 points of north latitude 11 degree, about 0.13 square kilometre of area.Category China Hainan province of research area Sansha City, island are ellipse in length
Circle, 3.9 meters of height above sea level average out to.
Experimental data is to cover the 2 scape high-resolution remote sensing images in the region, the first scape (Fig. 2 lower left sides and Fig. 3-a)
For a part for QuickBird remote sensing image panchromatic wave-bands, spatial resolution is 0.6 meter, images on January 27th, 2005；The
Two scapes (on the right side of Fig. 2 bottoms and Fig. 3-b) are a part for WorldView-2 remote sensing image panchromatic wave-bands, spatial resolution 0.5
Rice, imaging time is on December 31st, 2009.
Islands and reefs Remote Sensing Image Matching method (flow chart is shown in Fig. 1) of the invention based on hierarchical screening strategy, including following step
Suddenly：
The first step, affine invariants matching --- two width islands and reefs remote sensing images are extracted using MSER operators affine constant
Characteristic point, then affine invariants point is described with SIFT description, the dimensional feature vector of pixel-level generation 128, according to spy
Sign vector matches to the characteristic point of two width islands and reefs remote sensing images, obtains some characteristic points pair.
Islands and reefs and surrounding body are cut out from original islands and reefs remote sensing image first by ENVI softwares, then using MSER
The affine invariants point (this step is by OpenCV storehouses C++ programming realizations) of operator extraction islands and reefs image, but in the present invention
Involved affine invariants point is not limited to this algorithm and obtained, and two islands and reefs images generate 1562 and 2980 respectively in the present embodiment
Characteristic point, Fig. 3-e are that the south of China in 2005 sub-island image affine invariants point extracts result.Fig. 3-f are the south of China in 2009
Sub-island image affine invariants point extracts result.Then each characteristic point is described with SIFT description, generation 128 is
Characteristic vector (this step is by OpenCV storehouses C++ programming realizations), but involved characteristic vector is not limited to this calculation in the present invention
Method obtains.Calculate the Euclidean distance of the characteristic vector and all characteristic vectors on another image on image.When closest distance
When being less than 0.85 with the ratio of secondary adjacency, it is believed that this two characteristic vector matches, and generates a pair of characteristic points pair, this implementation
Example symbiosis is into 439 characteristic points pair.Fig. 3-g are China south sub-island image affine invariants matching result.
Second step, establish geometric transformation model --- selection affine Transform Model is as geometric transformation model, with breaking naturally
Split method and extract islands and reefs border from two width islands and reefs remote sensing images respectively, and resolve to obtain according to two islands and reefs boundary informations of acquisition
The model parameter of affine Transform Model.
In this step, islands and reefs reef platform is extracted from the islands and reefs remote sensing image of source with natural fracture method first, then using ENVI
Software carries out closing operation of mathematical morphology filtering, reuses ArcGIS softwares and carries out raster data vector quantization, obtains islands and reefs border, and Fig. 3-
C is the south of China in 2005 sub-island image islands and reefs Boundary Extraction result, and Fig. 3-d are that the south of China in 2009 sub-island image islands and reefs border carries
Take result.
Then, affine invariants matrix is progressively resolved according to two islands and reefs boundary informations, this step is overall by OpenCV
Storehouse C++ programming realizations, detailed step are as follows：
A1 central point), is calculated.Center point coordinateObtained according to following formula:
In the present embodiment two islands and reefs border central point plane coordinates be respectively (1168.60,1121.22) and (1138.82,
1135.80)。
A2 principal direction), is determined.The line beam equation of central point (x, y) was determined first, and line beam equation obtains according to following formula
：
Calculate again all straight lines and all boundary points in line beam hang down away from quadratic sum, when quadratic sum minimum, the straight line
Slope is border principal direction.Principal direction slope θ obtains according to following formula：
Two islands and reefs border principal direction slopes are respectively 41.531273 and 40.528446 in the present embodiment.
A3 the anglec of rotation), is calculated.Rotation angle parameter α in affine matrix is calculated, α values can be obtained by being subtracted each other by principal direction slope.This reality
It is -1.002827 to apply two islands and reefs border principal direction rotation angle parameters in example.
A4 scale parameter), is calculated.Principal direction slope is brought into first two straight lines are can obtain in line beam equation, by straight line
With islands and reefs border transversal as major axis, major axis is respectively 1044.132202 and 1043.324707 in the present embodiment.Then it is fixed again
Justice is time direction with principal direction vertical direction, and central point straight line is crossed with islands and reefs border transversal as short axle, this reality using power is upward
It is respectively 527.977173 and 557.712952 to apply a short-and-medium axle, finally according to the ratio of major axis between two islands and reefs borders and short axle
Ratio, try to achieve scale parameter in affine matrix.Major axis and the scale parameter of short axle are respectively 0.999226635 He in the present embodiment
1.056320198。
A5 translation parameters), is calculated.According to the anglec of rotation and scale parameter tried to achieve, after central point is brought into, translation parameters
Obtained according to following formula：
Translation parameters is respectively -7.970558821 and -27.94712967 on primary and secondary direction in the present embodiment.
A6 affine transformation matrix), is tried to achieve.Final to can obtain affine transformation matrix, affine transformation matrix is such as in the present embodiment
Under：
3rd step, first screening characteristic point are right --- according to described affine Transform Model by a spy of characteristic point centering
Sign point carries out affine transformation, the Euclidean distance between matched characteristic point is calculated after conversion, the error as characteristic point pair
Value, the characteristic point pair for accounting for total sample number 8% -12% is chosen with error amount from small to large, and error threshold gradually increases, constantly meter
Calculate error amount and be less than the characteristic point centering of the error threshold and belong to the characteristic point of main image and form the spatial cohesion of point set,
When spatial cohesion reaches maximum, corresponding error threshold screens out error amount more than the screening threshold value as screening threshold value
Characteristic point pair.
In this step, the characteristic point coordinate of characteristic point centering is brought into the affine transformation matrix in second step, tries to achieve change
Change the Euclidean distance of feature point coordinates and feature point coordinates of the same name, the error amount as characteristic point pair.According to the mistake of characteristic point pair
Difference to characteristic point to being ranked up, the spatial cohesion SDQ of preceding n error minimal characteristic point (main image)nObtained according to following formula
：
Wherein,The weighted center point coordinates of n error minimal characteristic point, (x before expressioni,yi) it is i-th
The coordinate of error minimal characteristic point.Keeping characteristics point to the characteristic points pair of the 8%-12% numbers of total sample number, select space from
The maximum characteristic point of divergence, as screening threshold value, characteristic point pair is screened with the threshold value to corresponding error amount.In the present embodiment
It is 100 to screen threshold value, and characteristic point is 35 to number after screening.Fig. 3-h are that China south sub-island image carries out the knot after screening for the first time
Fruit.
4th step, postsearch screening characteristic point are right --- using islands and reefs area Duplication as constraints, changed using RANSAC
For algorithm to characteristic point to screening.
In this step, by the use of islands and reefs area degree of overlapping as constraint, to choose optimal RANSAC screenings threshold value automatically.It is first
First, minimum RANSAC threshold values are chosen, the selection result keeping characteristics point pair or does not only retain the correct characteristic point pair of sub-fraction,
Resolve initial homography matrix, calculate in image islands and reefs face domain after homography matrix converts with islands and reefs face domain in another image
Overlapping area ratio；Then threshold value is stepped up, retains more proper characteristics points pair, overlapping area has substantially when islands and reefs face
Program can terminate during change.RANSAC threshold values now are appropriate threshold.Based on this threshold value, RANSAC is by all characteristic points pair
Screen as intra-office point drawn game exterior point, ensure that intra-office point is correct characteristic point pair, according to these characteristic points to can be accurate
Registering islands and reefs image.This step is overall by OpenCV storehouses C++ programming realizations, comprises the following steps that：
B1 first time RANSAC iteration as screening parameter and), is started using RANSAC initial thresholds, in this example at the beginning of RANSAC
The value of beginning threshold value is 0.01；
B2), RANSAC threshold values are as screening parameter, using RANSAC algorithms by characteristic point to dividing into intra-office point to drawn game
Exterior point pair, point pair not in the know is screened out, according to intra-office point to solving homography matrix；
B3) enter line translation to main image using the homography matrix of solution, calculate the island of image and shadow subject to registration after converting
The area Duplication on island as in；
Islands and reefs area Duplication Ratio obtains according to following formula：
Wherein, RμaFor with reference to islands and reefs face domain area,To convert islands and reefs face domain area；
B4 RANSAC threshold values, and repeat step b2, b3), are incrementally increased, when islands and reefs area Duplication is decreased obviously, is stopped
Only RANSAC iteration；
Second of iteration and RANSAC threshold values afterwards obtain according to following formula：
Wherein, TRcurrentFor the RANSAC threshold values of current iteration, TRpreviousFor the RANSAC threshold values of last iteration；
B5 a RANSAC threshold value before), islands and reefs area Duplication is decreased obviously is optimal RANSAC threshold values, screen out according to
The point pair not in the know being separated out according to the optimal RANSAC threshold values.
As Fig. 4 be islands and reefs overlapping area ratio with RANSAC changes of threshold figures, the upper left corner represents that RANSAC threshold values are in figure
Islands and reefs overlapping area is illustrated when 0.6, and corresponding island area Duplication is 94.52%, and the lower left corner is RANSAC threshold values when being 0.7
Islands and reefs overlapping area is illustrated, it is clear that and when RANSAC threshold values increase to 0.7, islands and reefs area Duplication is dropped near 86%, under
Drop fairly obvious, then 0.6 is optimal RANSAC threshold values.Fig. 3-i are that China south sub-island image progress RANSAC threshold values are 0.6
Result after postsearch screening,
5th step, islands and reefs Remote Sensing Image Matching --- after screening twice, square is singly answered to resolving according to the characteristic point of reservation
Battle array, the registration of islands and reefs remote sensing image is completed with homography matrix.Fig. 3-j are that China south sub-island image progress RANSAC threshold values are 0.7
Postsearch screening after result.
After undergoing above-mentioned steps, the present embodiment finally retains 10 characteristic points pair, and visually distinguishes 10 characteristic points to equal
For proper characteristics point pair, according to this 10 characteristic points to resolving homography matrix, China south sub-island remote sensing shadow is completed with homography matrix
The accuracy registration of picture.
In addition to the implementation, the present invention can also have other embodiment.It is all to use equivalent substitution or equivalent transformation shape
Into technical scheme, all fall within the protection domains of application claims.
Claims (7)
1. a kind of islands and reefs Remote Sensing Image Matching method based on hierarchical screening strategy, comprises the following steps：
The first step, affine invariants matching --- affine invariants point, pixel-level generation are extracted to two width islands and reefs remote sensing images
Characteristic vector, the characteristic point of two width islands and reefs remote sensing images is matched according to characteristic vector, obtains some characteristic points pair；
Second step, establish geometric transformation model --- selection affine Transform Model is as geometric transformation model, with natural fracture method
Extract islands and reefs borders from two width islands and reefs remote sensing images respectively, and according to two islands and reefs boundary informations of acquisition resolve to obtain affine
The model parameter of transformation model；
3rd step, first screening characteristic point are right --- according to described affine Transform Model by a characteristic point of characteristic point centering
Affine transformation is carried out, the Euclidean distance between matched characteristic point is calculated after conversion, as the error amount of characteristic point pair, with
Error amount chooses the characteristic point pair for accounting for total sample number 8% -12% from small to large, and error threshold gradually increases, continuous calculation error
The characteristic point centering that value is less than the error threshold belongs to the characteristic point of main image and forms the spatial cohesion of point set, works as space
Dispersion reaches error threshold corresponding during maximum and is used as screening threshold value, screens out the characteristic point that error amount is more than the screening threshold value
It is right；
4th step, postsearch screening characteristic point are right --- using islands and reefs area Duplication as constraints, calculated using RANSAC iteration
Method is as follows to carrying out screening specific method to characteristic point：
A, first time RANSAC iteration as screening parameter and is started using RANSAC initial thresholds, the value of RANSAC initial thresholds
For 0.01；
B, RANSAC threshold values are as screening parameter, using RANSAC algorithms by characteristic point to dividing into intra-office point to drawn game exterior point
It is right, point pair not in the know is screened out, according to intra-office point to solving homography matrix, and line translation is entered to main image using homography matrix, is calculated
The area Duplication on the island of image and island in image subject to registration after conversion；
C, RANSAC threshold values, and repeat step b are incrementally increased, when islands and reefs area Duplication is decreased obviously, stops RANSAC and changes
Generation；
D, a RANSAC threshold value before islands and reefs area Duplication is decreased obviously is optimal RANSAC threshold values, is screened out according to this most
The point pair not in the know that good RANSAC threshold values are separated out；
5th step, islands and reefs Remote Sensing Image Matching --- after the screening of the 3rd step and the 4th step, according to the characteristic point of reservation to solution
Homography matrix is calculated, with the registration for resolving the homography matrix completion islands and reefs remote sensing image obtained.
2. the islands and reefs Remote Sensing Image Matching method according to claim 1 based on hierarchical screening strategy, it is characterised in that：Institute
State in the first step, the number of feature points of generation is not less than 4.
3. the islands and reefs Remote Sensing Image Matching method according to claim 1 based on hierarchical screening strategy, it is characterised in that：The
In one step, affine invariants point is extracted to two width islands and reefs remote sensing images using MSER operators, then describes son to imitative with SIFT
Penetrate invariant features point to be described, the dimensional feature vector of pixel-level generation 128.
4. the islands and reefs Remote Sensing Image Matching method according to claim 1 based on hierarchical screening strategy, it is characterised in that：Institute
State in second step, affine transformation matrix model obtains according to following formula：
<mrow>
<mi>H</mi>
<mo>=</mo>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mn>0</mn>
</msub>
</mtd>
<mtd>
<mrow>
<msub>
<mi>m</mi>
<mn>1</mn>
</msub>
<mi>c</mi>
<mi>o</mi>
<mi>s</mi>
<mi>&alpha;</mi>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>m</mi>
<mn>1</mn>
</msub>
<mi>s</mi>
<mi>i</mi>
<mi>n</mi>
<mi>&alpha;</mi>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mn>0</mn>
</msub>
</mtd>
<mtd>
<mrow>
<mo>-</mo>
<msub>
<mi>m</mi>
<mn>2</mn>
</msub>
<mi>s</mi>
<mi>i</mi>
<mi>n</mi>
<mi>&alpha;</mi>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>m</mi>
<mn>2</mn>
</msub>
<mi>cos</mi>
<mi>&alpha;</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Above parameter can resolve according to the islands and reefs boundary information of two width images, wherein, α is the anglec of rotation, m1And m2For scale parameter,
x0And y0For translation parameters.
5. the islands and reefs Remote Sensing Image Matching method according to claim 1 based on hierarchical screening strategy, it is characterised in that：The
In three steps, the characteristic point that preceding n error amount minimal characteristic point centering belongs to main image forms the spatial cohesion of point set
SDQnObtained according to following formula：
<mrow>
<msub>
<mi>SDQ</mi>
<mi>n</mi>
</msub>
<mo>=</mo>
<msqrt>
<mfrac>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mover>
<mi>x</mi>
<mo>&OverBar;</mo>
</mover>
<mrow>
<mi>w</mi>
<mi>m</mi>
<mi>c</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
<mo>+</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>n</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>y</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mover>
<mi>y</mi>
<mo>&OverBar;</mo>
</mover>
<mrow>
<mi>w</mi>
<mi>m</mi>
<mi>c</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
<mi>n</mi>
</mfrac>
</msqrt>
</mrow>
Wherein,The weighted center point coordinates of n error amount minimal characteristic point pair, (x before expressioni,yi) missed for i-th
Poor minimal characteristic point centering belongs to the coordinate of the characteristic point of main image.
6. the islands and reefs Remote Sensing Image Matching method according to claim 1 based on hierarchical screening strategy, it is characterised in that：Institute
State in the 4th step, second of iteration and RANSAC threshold values afterwards obtain according to following formula：
Wherein, TRcurrentFor the RANSAC threshold values of current iteration, TRpreviousFor the RANSAC threshold values of last iteration.
7. the islands and reefs Remote Sensing Image Matching method according to claim 1 based on hierarchical screening strategy, it is characterised in that：Institute
State in the 4th step, islands and reefs area Duplication Ratio obtains according to following formula：
<mrow>
<mi>R</mi>
<mi>a</mi>
<mi>t</mi>
<mi>i</mi>
<mi>o</mi>
<mo>=</mo>
<mfrac>
<mrow>
<msub>
<mi>R</mi>
<mrow>
<mi>&mu;</mi>
<mi>a</mi>
</mrow>
</msub>
<mo>&cap;</mo>
<msub>
<mi>R</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>H</mi>
<mi>T</mi>
</msup>
<mi>&mu;</mi>
<mi>b</mi>
<mi>H</mi>
<mo>)</mo>
</mrow>
</msub>
</mrow>
<mrow>
<msub>
<mi>R</mi>
<mrow>
<mi>&mu;</mi>
<mi>a</mi>
</mrow>
</msub>
<mo>&cup;</mo>
<msub>
<mi>R</mi>
<mrow>
<mo>(</mo>
<msup>
<mi>H</mi>
<mi>T</mi>
</msup>
<mi>&mu;</mi>
<mi>b</mi>
<mi>H</mi>
<mo>)</mo>
</mrow>
</msub>
</mrow>
</mfrac>
<mo>&times;</mo>
<mn>100</mn>
<mi>%</mi>
</mrow>
Wherein, RμaFor with reference to islands and reefs face domain area,To convert islands and reefs face domain area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510500058.7A CN105005962B (en) | 2015-08-14 | 2015-08-14 | Islands and reefs Remote Sensing Image Matching method based on hierarchical screening strategy |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510500058.7A CN105005962B (en) | 2015-08-14 | 2015-08-14 | Islands and reefs Remote Sensing Image Matching method based on hierarchical screening strategy |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105005962A CN105005962A (en) | 2015-10-28 |
| CN105005962B true CN105005962B (en) | 2018-01-12 |
Family
ID=54378621
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510500058.7A Expired - Fee Related CN105005962B (en) | 2015-08-14 | 2015-08-14 | Islands and reefs Remote Sensing Image Matching method based on hierarchical screening strategy |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105005962B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108053370B (en) * | 2017-11-29 | 2021-08-06 | 合肥工业大学 | Imaging coordinate correction method based on matching error suppression |
| CN112182125B (en) * | 2020-09-14 | 2022-07-05 | 中国科学院重庆绿色智能技术研究院 | Business gathering area boundary identification system |
| CN116152532A (en) * | 2023-04-14 | 2023-05-23 | 中国地质大学（武汉） | Remote sensing image feature extraction and matching method and device and electronic equipment |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102750696A (en) * | 2012-06-06 | 2012-10-24 | 南京大学 | Affine invariant feature and coastline constraint-based automatic coastal zone remote-sensing image registration method |
| CN102930525A (en) * | 2012-09-14 | 2013-02-13 | 武汉大学 | Line matching method based on affine invariant feature and homography |
-
2015
- 2015-08-14 CN CN201510500058.7A patent/CN105005962B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102750696A (en) * | 2012-06-06 | 2012-10-24 | 南京大学 | Affine invariant feature and coastline constraint-based automatic coastal zone remote-sensing image registration method |
| CN102930525A (en) * | 2012-09-14 | 2013-02-13 | 武汉大学 | Line matching method based on affine invariant feature and homography |
Non-Patent Citations (4)
| Title |
|---|
| Automatic registration of coastal remotely sensed imagery by affine invariant feature matching with shoreline constraint;liang cheng等;《Marine Geodesy》;20140530;第32-46页 * |
| Remote sensing image matching by integrating affine invariant feature extraction and RANSAC;liang cheng等;《Computer and electrical engineering》;20120421;第1023-1032页 * |
| 遥感影像仿射不变特征匹配的自动优化;程亮等;《武汉大学学报》;20090430;第34卷(第4期);第418-422页 * |
| 面向宽基线立体影像匹配的高质量仿射不变特征提取方法;程亮等;《测绘学报》;20080228;第37卷(第1期);第77-82页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105005962A (en) | 2015-10-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104484648B (en) | Robot variable visual angle obstacle detection method based on outline identification | |
| CN101714262B (en) | Method for reconstructing three-dimensional scene of single image | |
| CN103020966B (en) | A kind of aviation based on contour of building constraint and ground LiDAR data autoegistration method | |
| CN108007355A (en) | Distress in concrete detection method based on Image distance transform | |
| CN105005962B (en) | Islands and reefs Remote Sensing Image Matching method based on hierarchical screening strategy | |
| CN110033411A (en) | The efficient joining method of highway construction scene panoramic picture based on unmanned plane | |
| CN104217459B (en) | A kind of spheroid character extracting method | |
| CN108596892A (en) | A kind of identification of Weld Defects based on improvement LeNet-5 models | |
| CN107945217B (en) | Image characteristic point pair rapid screening method and system suitable for automatic assembly | |
| CN107967474A (en) | A kind of sea-surface target conspicuousness detection method based on convolutional neural networks | |
| CN114219819A (en) | Oblique photography model unitization method based on orthoscopic image boundary detection | |
| CN105447274A (en) | Method of performing coastal wetland drawing for medium-resolution remote sensing image by utilizing object-oriented classification technology | |
| CN109712112A (en) | Taking photo by plane based on local feature is insulated the localization method of subgraph | |
| CN108764132A (en) | A kind of lake and marshland remote sensing images error detection method | |
| CN110990780B (en) | Gradient extraction method based on SRTM data | |
| CN106407973A (en) | Robust recognition method for AR code marked on cylinder | |
| CN106845343A (en) | A kind of remote sensing image offshore platform automatic testing method | |
| CN104751451B (en) | Point off density cloud extracting method based on unmanned plane low latitude high resolution image | |
| CN116228539A (en) | Unmanned aerial vehicle remote sensing image stitching method | |
| CN104180794B (en) | The disposal route in digital orthoimage garland region | |
| CN108109208A (en) | A kind of marine wind electric field augmented reality method | |
| CN107169412A (en) | Remote sensing image harbor-berthing ship detection method based on mixed model decision | |
| CN110532865A (en) | Spacecraft structure recognition methods based on visible light and laser fusion | |
| CN103886289B (en) | Direction self-adaptive method and system for identifying on-water bridge targets | |
| CN111914848B (en) | Remote sensing image semantic segmentation method and system fusing GIS data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20180112 Termination date: 20180814 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |