CN102930525A - Line matching method based on affine invariant feature and homography - Google Patents
Line matching method based on affine invariant feature and homography Download PDFInfo
- Publication number
- CN102930525A CN102930525A CN2012103425663A CN201210342566A CN102930525A CN 102930525 A CN102930525 A CN 102930525A CN 2012103425663 A CN2012103425663 A CN 2012103425663A CN 201210342566 A CN201210342566 A CN 201210342566A CN 102930525 A CN102930525 A CN 102930525A
- Authority
- CN
- China
- Prior art keywords
- line segment
- line
- doubtful
- same name
- main leaf
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 title claims abstract description 32
- 239000011159 matrix material Substances 0.000 claims description 55
- 230000008878 coupling Effects 0.000 claims description 50
- 238000010168 coupling process Methods 0.000 claims description 48
- 238000005859 coupling reaction Methods 0.000 claims description 48
- 230000008569 process Effects 0.000 claims description 20
- 238000000605 extraction Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 5
- 238000013461 design Methods 0.000 claims description 3
- 239000004744 fabric Substances 0.000 claims description 3
- 230000013011 mating Effects 0.000 claims description 3
- 230000007547 defect Effects 0.000 abstract 1
- 230000000694 effects Effects 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000002441 reversible effect Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 238000011158 quantitative evaluation Methods 0.000 description 1
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention relates to a line matching method based on affine invariant feature and homography. As line matching is lack of an effective geometric constraint of an epipolar line in point matching. A homography constraint is introduced as the geometric constraint of line segment matching to make up the defect that the line segment matching is lack of a strong geometric constraint. The invention additionally discloses a line segment automatic matching method based on the homography constraint. Line segments are transmitted and sleeved among images through the homography constraint, so that the searching difficulty for the line segments of the same name is reduced, and the matching accuracy is improved; the line segments of the same name are backwards searched after primary matching, so that matching errors are removed, and the matching accuracy is further improved. The method achieves the line segment automatic matching for remote-sensing image pairs.
Description
Technical field
The present invention relates to a kind of matching process of remote sensing image, particularly relate to a kind of line matching process based on affine invariant features and homography matrix.
Background technology
Adopt the line feature as the coupling primitive, in some application-specific, have obvious advantage, as in the buildings three-dimensional reconstruction (Habib 1998).This be because: buildings comprises a large amount of straight-line segments on the image; Straight-line segment easily detected and have a clear and definite physical significance more; Line feature corresponding point feature has how describable geometrical constraint, and is more reliable.
Yet, the corresponding point coupling, line match is technical more difficult, and main cause is (Schmid and Zisserman 1997; Baillard and Zisserman 2000): (1) extracts line segment from image, and line segment ruptures mostly, and the topological relation between the line segment is lost; (2) coupling has so very strong restriction relation of nuclear line geometry, and the line coupling lacks such restriction relation.
Present international digital Photogrammetric System is to carry out the same place search by the constraint of nuclear line geometry mostly, yet the constraint of nuclear line can not provide the one-to-one relationship of unique point.Have the researchist to propose to utilize corresponding image points to consist of triangulation network constraint homonymous line hunting zone, the method needs in advance triangle networking, and efficient is not high.
In multiple view geometry, homography matrix represents the reversible homogeneous transformation between two planes, has been widely used in the fields such as vision measurement, camera calibration, three-dimensional reconstruction, Image Mosaics, and has played the part of therein extremely important role.At computer vision field, homography matrix can only be applied to two features convey between the plane in theory, but in the Photogrammetry and Remote Sensing field, for aviation image or satellite image, because the fluctuating of landform or the discrepancy in elevation of atural object are very little with respect to flying height, homography matrix also is (the Schmid and Zisserman 1997) that is suitable for.
Summary of the invention
The technical matters that the present invention solves is: propose a kind of line matching process based on affine invariant features and homography matrix that can improve the line match accuracy rate, realize the right line segment Auto-matching of remote sensing image.
In order to solve the problems of the technologies described above, the technical scheme that the present invention proposes is: a kind of line matching process based on affine invariant features and homography matrix may further comprise the steps:
Step 1, obtain remote sensing image to optimum homography matrix---based on the RANSAC iterative algorithm of affine invariant features coupling, obtain remote sensing image to optimum homography matrix;
Step 2, transmit remote sensing image between line segment---the line segment that extracts respectively remote sensing image centering obtains line chart, and the optimum homography matrix that obtains take step 1 is as constraint, take one of them line chart as main leaf, another line chart is from sheet, to be delivered on the main leaf from the line segment on the sheet, wherein the original line segment on the main leaf is the main leaf line segment, and the line segment that is delivered to main leaf from sheet is from the sheet line segment;
Step 3, determine the sets of line segments doubtful of the same name that the main leaf line segment is corresponding---travel through all main leaf line segments, according to direction, distance and the degree of overlapping between the master and slave line segment, find all corresponding with the main leaf line segment from the sheet line segment doubtful line segments of the same name, consist of the sets of line segments doubtful of the same name of this main leaf line segment;
Step 4, definite from sets of line segments doubtful of the same name corresponding to sheet line segment---travel through all from the sheet line segment, according to direction, distance and the degree of overlapping between the master and slave line segment, find in the main leaf line segment with from all doubtful line segments of the same name corresponding to sheet line segment, consist of this from the sets of line segments doubtful of the same name of sheet line segment;
Innovative point of the present invention is: for the problem that lacks the effective geometrical constraint as a coupling center line geometry in the line coupling, introduce the homography matrix constraint as the geometrical constraint of line match, to remedy the situation that lacks strong geometrical constraint in the line match, and a kind of line segment automatic matching method based on homography matrix constraint proposed, transmission and the fit of line segment between image have been realized by the constraint of homography matrix, reduce line search difficulty of the same name, improved matching accuracy rate; And after preliminary coupling is finished, reverse search line segment of the same name, thus the mistake coupling can be rejected, further improved matching accuracy rate.
The present invention carries out line match take homography matrix as constraint, and homography matrix is the key of subsequent treatment so accurately.Will obtain accurately homography matrix between a pair of remote sensing image, need to guarantee does not have erroneous matching in matching process, and keeps as much as possible correct coupling.ED-MSER algorithm (the Cheng et al. (2008) that utilizes the inventor herein to propose, robust affine invariant feature extraction for image matching, IEEE Geoscience and Remote Sensing Letters, 5 (2)), remote sensing image is carried out affine invariant feature extraction, and then characteristic matching, little for visual angle change, texture is good, overlapping large image can be obtained goodish effect, is matched to power very high.Yet viewpoint angles changes greatly, degree of overlapping is little when facing, during the difficult image of texture, be matched to power still not ideal enough, some still existence of mistake coupling.Therefore the estimation of homography matrix is the key of middle conductor coupling of the present invention between remote sensing image, the present invention has proposed a kind of RANSAC iterative algorithm based on affine invariant features coupling in step 1, realized homography matrix accurately, reliably obtain, concrete grammar is as follows:
I, use MSER describe operator as affine invariant feature extraction operator, SIFT as feature, carry out the feature extraction of two width of cloth images, the generating feature vector;
Do distance operation between II, the proper vector to two subpictures, obtain the SIFT proper vector pair of coupling;
III, use the RANSAC method to the SIFT proper vector of coupling to processing, the input parameter of RANSAC method is that the geometric model of distance threshold, input is the homography matrix of unknown parameters, obtain intra-office SIFT proper vector after processing to reaching the design parameter of homography matrix, when carrying out this step first the distance threshold span of RANSAC method for (0,1];
IV, according to the homography matrix that estimates to remote sensing image to mating, calculate matching accuracy rate, described matching accuracy rate is that correct number of pairs and coupling are to the ratio of sum;
V, progressively increasing the distance threshold of RANSAC method, and repeat the 3rd) step to the 4th step, matching accuracy rate was that 100% homography matrix corresponding to maximal distance threshold is exactly optimum homography matrix until matching accuracy rate descends since 100%.
Wherein, in the step II, during coupling SIFT unique point, when the ratio of minimal characteristic vector distance and time minimal characteristic vector distance greater than 0.6 the time, proper vector apart from that a pair of SIFT unique point of minimum as the SIFT unique point of coupling pair.
In step IV step, if the degree of overlapping of the right characteristic area of the same name of remote sensing image thinks then that greater than 50% both are corresponding one by one, namely this coupling is correct coupling.
The core concept of above-mentioned RANSAC iterative algorithm based on affine invariant features coupling is: carry out at ED-MSER on the basis of affine invariant feature extraction, utilize the affine invariant features coupling of RANSAC algorithm optimization, characteristics according to remote sensing image, with the geometrical constrain model of homography matrix as the RANSAC algorithm, be generally the situation in face territory for affine invariant feature extraction feature that technology is obtained, the feature of the same name of the doubtful coupling after utilizing homography matrix with the RANSAC Robust Estimation correct (inliers) is to carrying out fit, calculate the degree of overlapping of characteristic area of the same name and coupling accuracy of this upper calculated characteristics group in basis again, quantitative evaluation index take the coupling accuracy as the iterative processing effect, automatically adjust distance threshold, instruct the iterative process of RANSAC optimization process, obtain optimum homography matrix.
The line segment that remote sensing image is extracted usually can be the fracture shape, for the accuracy that improves line match, the interference that reduces line match, can completing steps 2 transmit remote sensing images to a line segment after, the line segment of these fractures is carried out pre-service.Therefore the present invention also provides a kind of to the pretreated method of remote sensing image line segment, specific as follows:
When two line segment angles are not more than 5 pixels less than 4 °, vertical range, and two line segments have lap: when line segment length difference surpasses 50%, keep long line segment, remove shorter line segment; Merge otherwise do line segment, get the straight line that distance two line segments equate, and with two line segments farthest two-end-point as the line segment end points after merging.
The present invention provides a kind of method that generates main leaf line segment candidate line-segment sets of the same name in step 3, when satisfying simultaneously following three conditions from the sheet line segment, should be the line segment doubtful of the same name of corresponding main leaf line segment when a certain from the sheet line segment:
A, with main leaf line segment angle less than 5 °;
B, with main leaf line segment overlap length account for more than 20% of main leaf line segment length;
C, be not more than 10 pixels to the vertical range of main leaf line segment.
The present invention also provides the method that sets of line segments doubtful of the same name corresponding to main leaf line segment is optimized, and comprises the similarity of utilizing line segment lap gray scale and utilizes line segment left and right sides grey scale signal ratio.
Wherein, utilize the similarity of line segment lap gray scale to dwindle the concrete grammar of doubtful sets of line segments scope of the same name as follows:
Extend out along vertical main leaf line segment direction left-right symmetric, form and extend out rectangle, calculate the described remote sensing image average gray M that extends out the rectangular foot-print territory; In sets of line segments doubtful of the same name corresponding to this main leaf line segment, calculate the average gray M ＇ that every doubtful line segment of the same name extends out the remote sensing image in rectangular foot-print territory with same method; When M/M ＇ greater than 2 or less than 1/2 the time, should from doubtful sets of line segments of the same name, reject by doubtful line segment of the same name, otherwise keep this doubtful line segment of the same name.
Wherein, utilize line segment left and right sides grey scale signal as follows than the concrete grammar that dwindles doubtful sets of line segments scope of the same name:
Extend out along vertical main leaf line segment direction left-right symmetric, form two rectangles, according to the gray scale of the remote sensing image in two rectangular foot-print territories, record the high side of gray-scale value; In sets of line segments doubtful of the same name corresponding to this main leaf line segment, calculate the remote sensing image left and right sides gray scale in two rectangular foot-print territories of every doubtful line segment of the same name with same method, if the side position that the high side of doubtful line segment gray-scale value of the same name and main leaf line segment gray-scale value are high is different, should from doubtful sets of line segments of the same name, reject by doubtful line segment of the same name, otherwise keep this doubtful line segment of the same name.
The present invention can also utilize the Kmeans clustering algorithm that sets of line segments doubtful of the same name corresponding to main leaf line segment further optimized in step 3, and concrete grammar is:
Calculate the main leaf line segment to the vertical range of its doubtful line segment of the same name, utilize the Kmeans clustering algorithm that doubtful line segment of the same name is divided into two classes according to described vertical range, the line segment doubtful of the same name that vertical range is larger is classified as a class automatically, the line segment doubtful of the same name that the vertical range value is less is classified as another kind of automatically, rejects all line segments doubtful of the same name in the large class of vertical range.
Need to prove: the method for the doubtful sets of line segments of the same name that the optimization main leaf line segment that provides in step 3 is corresponding, all can correspondingly be applied in the step 4, be used for optimizing from sets of line segments doubtful of the same name corresponding to sheet line segment.Doubtful sets of line segments of the same name is optimized, can reduces calculated amount, improve accuracy rate.
The invention has the beneficial effects as follows: (1) the present invention is directed to the problem that lacks the effective geometrical constraint as a coupling center line geometry in the line coupling, introduce the homography matrix constraint as the geometrical constraint of line match, remedied the situation that lacks strong geometrical constraint in the line match;
(2) the present invention carries the line segment automatic matching method that uses based on the homography matrix constraint, transmission and the fit of line segment between image have been realized by the constraint of homography matrix, can realize that line segment strides the accurate transmission of image, reduce line search difficulty of the same name, improve matching accuracy rate;
(3) the present invention by reverse search line segment of the same name, can reject the mistake coupling, thereby further improve matching accuracy rate after preliminary coupling is finished.
Description of drawings
Below in conjunction with accompanying drawing the line matching process based on affine invariant features and homography matrix of the present invention is described further.
Fig. 1 is one of remote sensing image of the embodiment of the invention.
Fig. 2 be the embodiment of the invention remote sensing image two.
Fig. 3 is the line chart that Fig. 1 extracts.
Fig. 4 is the line chart that Fig. 2 extracts.
Fig. 5 is the as a result figure that the line segment of Fig. 4 is delivered to Fig. 3.
Fig. 6 is the line match figure as a result of Fig. 3.
Fig. 7 is the line match figure as a result of Fig. 4.
Embodiment
Embodiment
The experimental data of the present embodiment as depicted in figs. 1 and 2, spatial resolution is higher, coverage is relatively large, has contained the types of ground objects such as building, road, vegetation, bare area.
The line matching process based on affine invariant features and homography matrix of the present embodiment may further comprise the steps:
Step 1, obtain remote sensing image to optimum homography matrix.
The present embodiment use the RANSAC iterative algorithm based on affine invariant features coupling obtain remote sensing image to optimum homography matrix, specific algorithm is as follows:
I, use MSER describe operator as affine invariant feature extraction operator, SIFT as feature, carry out the feature extraction of two width of cloth images, the generating feature vector;
Do distance operation between II, the proper vector to two subpictures, obtain the SIFT proper vector pair of coupling;
During coupling SIFT unique point, when the ratio of minimal characteristic vector distance and time minimal characteristic vector distance greater than 0.6 the time, proper vector apart from that a pair of SIFT unique point of minimum as the SIFT unique point of coupling pair;
III, use the RANSAC method to the SIFT proper vector of coupling to processing, the input parameter of RANSAC method is that the geometric model of distance threshold, input is the homography matrix of unknown parameters, obtain intra-office SIFT proper vector after processing to reaching the design parameter of homography matrix, the distance threshold of RANSAC method is 1E-6 when carrying out this step first;
IV, according to the homography matrix that estimates to remote sensing image to mating, if the degree of overlapping of the right characteristic area of the same name of remote sensing image thinks then that greater than 50% both are corresponding one by one, namely this coupling is correct coupling; Calculate matching accuracy rate, wherein matching accuracy rate is correct number of pairs and the ratio of coupling to sum;
V, according to formula
(n is iterations in the formula) calculates the value of distance threshold, progressively increase the distance threshold of RANSAC method, and repeating Step II I to step IV until matching accuracy rate descends since 100%, matching accuracy rate is that 100% homography matrix corresponding to maximal distance threshold is exactly optimum homography matrix.
Step 2, transmit remote sensing image between line segment.
The present embodiment utilizes the EDISION operator, respectively Fig. 1 and Fig. 2 is carried out line segments extraction, and the result who obtains respectively as shown in Figure 3 and Figure 4.The optimum homography matrix that obtains take step 1 is as constraint, and the image in Fig. 3 is as main leaf, and image is from sheet among Fig. 4, will be delivered on the line chart of main leaf, as shown in Figure 5 from the line segment on the sheet.In Fig. 5, the line segment of Fig. 4 is converted into dotted line, is used for distinguishing the line segment among Fig. 3, and the effect of fit is satisfactory; Wherein the original line segment on the main leaf is the main leaf line segment, and the line segment that is delivered to main leaf from sheet is from the sheet line segment.
After fit, adopt following method to mate pre-service: when two line segment angles are not more than 5 pixels less than 4 °, vertical range, and two line segments are when having lap, if line segment length difference surpasses 50%, keep long line segment, remove shorter line segment; Merge otherwise do line segment, get the straight line that distance two line segments equate, and with two line segments farthest two-end-point as the line segment end points after merging.
Step 3, determine the sets of line segments doubtful of the same name that the main leaf line segment is corresponding.
Traveling through all main leaf line segments, when satisfying simultaneously following three conditions from the sheet line segment, should be the line segment doubtful of the same name of corresponding main leaf line segment when a certain from the sheet line segment:
A, with main leaf line segment angle less than 5 °;
B, with main leaf line segment overlap length account for more than 20% of main leaf line segment length;
C, be not more than 10 pixels to the vertical range of main leaf line segment.
Find all corresponding with the main leaf line segment from the sheet line segment doubtful line segments of the same name, just consisted of the sets of line segments doubtful of the same name of this main leaf line segment.
It is as follows that the present embodiment utilizes the similarity of line segment lap gray scale to dwindle the concrete grammar of doubtful sets of line segments scope of the same name:
Extend out along vertical main leaf line segment direction left-right symmetric, form and extend out rectangle, calculate the described remote sensing image average gray M that extends out the rectangular foot-print territory; In sets of line segments doubtful of the same name corresponding to this main leaf line segment, calculate the average gray M ＇ that every doubtful line segment of the same name extends out the remote sensing image in rectangular foot-print territory with same method; When M/M ＇ greater than 2 or less than 1/2 the time, should from doubtful sets of line segments of the same name, reject by doubtful line segment of the same name, otherwise keep this doubtful line segment of the same name.
The present embodiment utilizes line segment left and right sides grey scale signal than the scope of further dwindling doubtful sets of line segments of the same name, and concrete grammar is as follows:
Extend out along vertical main leaf line segment direction left-right symmetric, form two rectangles, according to the gray scale of the remote sensing image in two rectangular foot-print territories, record the high side of gray-scale value; In sets of line segments doubtful of the same name corresponding to this main leaf line segment, calculate the remote sensing image left and right sides gray scale in two rectangular foot-print territories of every doubtful line segment of the same name with same method, if the side position that the high side of doubtful line segment gray-scale value of the same name and main leaf line segment gray-scale value are high is different, should from doubtful sets of line segments of the same name, reject by doubtful line segment of the same name, otherwise keep this doubtful line segment of the same name.
The present embodiment also utilizes the doubtful of the same name sets of line segments optimization corresponding to the main leaf line segment of Kmeans clustering algorithm, further dwindles the scope of doubtful sets of line segments of the same name, and concrete grammar is:
Calculate the main leaf line segment to the vertical range of its doubtful line segment of the same name, utilize the Kmeans clustering algorithm that doubtful line segment of the same name is divided into two classes according to above-mentioned vertical range, the line segment doubtful of the same name that vertical range is larger is classified as a class automatically, the line segment doubtful of the same name that the vertical range value is less is classified as another kind of automatically, rejects all line segments doubtful of the same name in the large class of vertical range.
Step 4, definite from sets of line segments doubtful of the same name corresponding to sheet line segment.
Refer step 3, travel through all from the sheet line segment, find in the main leaf line segment with from all doubtful line segments of the same name corresponding to sheet line segment, generate this from the sets of line segments doubtful of the same name of sheet line segment, and utilize line segment left and right sides grey scale signal ratio and Kmeans clustering algorithm to optimize doubtful sets of line segments of the same name, thereby dwindle the scope of doubtful sets of line segments of the same name.
Compare from the corresponding relation between line segment and its doubtful line segment of the same name in the main leaf line segment that step 3 is obtained and the corresponding relation between its doubtful line segment of the same name and the step 4, when main leaf line segment L with step 3, step 4, all have corresponding relation from sheet line segment L ＇, then main leaf line segment L is with the match is successful from sheet line segment L ＇, otherwise rejects main leaf line segment L and matching relationship from sheet line segment L ＇.
According to above step, the effect of line segment Auto-matching is numbered identical line segment and is coupling line segment that the present invention obtains pair respectively as shown in Figure 6 and Figure 7.Associating Fig. 6 and Fig. 7, the line segment of coupling are to there being 45 pairs, and wherein correct coupling is 39 pairs, and mistake is mated 6 pairs, and accuracy reaches 87%.Experimental result shows, the present invention can be effectively applied to the line match of general significance and process.
Line matching process based on affine invariant features and homography matrix of the present invention is not limited to the described concrete technical scheme of above-described embodiment, and all employings are equal to the technical scheme of replacing formation and are the protection domain that the present invention requires.
Claims (9)
1. line matching process based on affine invariant features and homography matrix may further comprise the steps:
Step 1, obtain remote sensing image to optimum homography matrix---based on the RANSAC iterative algorithm of affine invariant features coupling, obtain remote sensing image to optimum homography matrix;
Step 2, transmit remote sensing image between line segment---the line segment that extracts respectively remote sensing image centering obtains line chart, and the optimum homography matrix that obtains take step 1 is as constraint, take one of them line chart as main leaf, another line chart is from sheet, to be delivered on the main leaf from the line segment on the sheet, wherein the original line segment on the main leaf is the main leaf line segment, and the line segment that is delivered to main leaf from sheet is from the sheet line segment;
Step 3, determine the sets of line segments doubtful of the same name that the main leaf line segment is corresponding---travel through all main leaf line segments, according to direction, distance and the degree of overlapping between the master and slave line segment, find all corresponding with the main leaf line segment from the sheet line segment doubtful line segments of the same name, consist of the sets of line segments doubtful of the same name of this main leaf line segment;
Step 4, definite from sets of line segments doubtful of the same name corresponding to sheet line segment---travel through all from the sheet line segment, according to direction, distance and the degree of overlapping between the master and slave line segment, find in the main leaf line segment with from all doubtful line segments of the same name corresponding to sheet line segment, consist of this from the sets of line segments doubtful of the same name of sheet line segment;
Step 5, reject the mistake coupling---compare from the corresponding relation between sheet line segment and its doubtful line segment of the same name in main leaf line segment that step 3 is obtained and the corresponding relation between its doubtful line segment of the same name and the step 4, when main leaf line segment L with step 3, step 4, all have corresponding relation from sheet line segment L ＇, then described main leaf line segment L is with the match is successful from sheet line segment L ＇, otherwise rejects main leaf line segment L and matching relationship from sheet line segment L ＇.
2. the line matching process based on affine invariant features and homography matrix according to claim 1 is characterized in that, based on the RANSAC iterative algorithm of affine invariant features coupling, concrete grammar is as follows in the described step 1:
I, use MSER describe operator as affine invariant feature extraction operator, SIFT as feature, carry out the feature extraction of two width of cloth images, the generating feature vector;
Do distance operation between II, the proper vector to two subpictures, obtain the SIFT proper vector pair of coupling;
III, use the RANSAC method to the SIFT proper vector of coupling to processing, the input parameter of RANSAC method is that the geometric model of distance threshold, input is the homography matrix of unknown parameters, obtain intra-office SIFT proper vector after processing to reaching the design parameter of homography matrix, when carrying out this step first the distance threshold span of RANSAC method for (0,1];
IV, according to the homography matrix that estimates to remote sensing image to mating, calculate matching accuracy rate, described matching accuracy rate is that correct number of pairs and coupling are to the ratio of sum;
V, progressively increasing the distance threshold of RANSAC method, and repeat the 3rd) step to the 4th step, matching accuracy rate was that 100% homography matrix corresponding to maximal distance threshold is exactly optimum homography matrix until matching accuracy rate descends since 100%.
3. the line matching process based on affine invariant features and homography matrix according to claim 2, it is characterized in that: in the step IV, if the degree of overlapping of the characteristic area of the same name that remote sensing image is right, thinks then that both are corresponding one by one greater than 50%, namely this coupling is correct coupling.
4. the line matching process based on affine invariant features and homography matrix according to claim 2, it is characterized in that: in the described step II, during coupling SIFT unique point, when the ratio of minimal characteristic vector distance and time minimal characteristic vector distance greater than 0.6 the time, proper vector apart from that a pair of proper vector of minimum as the proper vector of coupling pair.
5. the line matching process based on affine invariant features and homography matrix according to claim 1 is characterized in that, after step 2 is finished, step 3 carries out pre-service to the line segment of remote sensing image before carrying out, concrete grammar is as follows:
When two line segment angles are not more than 5 pixels less than 4 °, vertical range, and two line segments have lap: when line segment length difference surpasses 50%, keep long line segment, remove shorter line segment; Merge otherwise do line segment, get the straight line that distance two line segments equate, and with two line segments farthest two-end-point as the line segment end points after merging.
6. the line matching process based on affine invariant features and homography matrix according to claim 1 is characterized in that, in step 3, when satisfying simultaneously following three conditions from the sheet line segment, should be the line segment doubtful of the same name of corresponding main leaf line segment when a certain from the sheet line segment:
A, with main leaf line segment angle less than 5 °;
B, with main leaf line segment overlap length account for more than 20% of main leaf line segment length;
C, be not more than 10 pixels to the vertical range of main leaf line segment.
7. the line matching process based on affine invariant features and homography matrix according to claim 6 is characterized in that: utilize the similarity of line segment lap gray scale to dwindle the scope of doubtful sets of line segments of the same name corresponding to main leaf line segment, concrete grammar is as follows:
Extend out along vertical main leaf line segment direction left-right symmetric, form and extend out rectangle, calculate the described remote sensing image average gray M that extends out the rectangular foot-print territory; In sets of line segments doubtful of the same name corresponding to this main leaf line segment, calculate the average gray M ＇ that every doubtful line segment of the same name extends out the remote sensing image in rectangular foot-print territory with same method; When M/M ＇ greater than 2 or less than 1/2 the time, should from doubtful sets of line segments of the same name, reject by doubtful line segment of the same name, otherwise keep this doubtful line segment of the same name.
8. the line matching process based on affine invariant features and homography matrix according to claim 6 is characterized in that, utilizes line segment left and right sides grey scale signal than the scope of dwindling doubtful sets of line segments of the same name corresponding to main leaf line segment, and concrete grammar is as follows:
Extend out along vertical main leaf line segment direction left-right symmetric, form two rectangles, according to the gray scale of the remote sensing image in two rectangular foot-print territories, record the high side of gray-scale value; In sets of line segments doubtful of the same name corresponding to this main leaf line segment, calculate the remote sensing image left and right sides gray scale in two rectangular foot-print territories of every doubtful line segment of the same name with same method, if the side position that the high side of doubtful line segment gray-scale value of the same name and main leaf line segment gray-scale value are high is different, should from doubtful sets of line segments of the same name, reject by doubtful line segment of the same name, otherwise keep this doubtful line segment of the same name.
9. according to claim 6,7 or 8 described line matching process based on affine invariant features and homography matrix, it is characterized in that, in step 3, utilize the Kmeans clustering algorithm that sets of line segments doubtful of the same name corresponding to main leaf line segment further optimized, concrete grammar is:
Calculate the main leaf line segment to the vertical range of its doubtful line segment of the same name, utilize the Kmeans clustering algorithm that doubtful line segment of the same name is divided into two classes according to described vertical range, the line segment doubtful of the same name that vertical range is larger is classified as a class automatically, the line segment doubtful of the same name that the vertical range value is less is classified as another kind of automatically, rejects all line segments doubtful of the same name in the large class of vertical range.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210342566.3A CN102930525B (en) | 2012-09-14 | 2012-09-14 | Line matching method based on affine invariant feature and homography |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210342566.3A CN102930525B (en) | 2012-09-14 | 2012-09-14 | Line matching method based on affine invariant feature and homography |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102930525A true CN102930525A (en) | 2013-02-13 |
| CN102930525B CN102930525B (en) | 2015-04-15 |
Family
ID=47645315
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210342566.3A Expired - Fee Related CN102930525B (en) | 2012-09-14 | 2012-09-14 | Line matching method based on affine invariant feature and homography |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102930525B (en) |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103353984A (en) * | 2013-04-03 | 2013-10-16 | 西安交通大学 | Method for matching multiple image segments by using non geometric constraints |
| CN103679676A (en) * | 2013-12-02 | 2014-03-26 | 西北工业大学 | Quick unordered image stitching method based on multi-level word bag clustering |
| CN104392457A (en) * | 2014-12-11 | 2015-03-04 | 中国测绘科学研究院 | Automatic matching method and device for connection points of slanted images |
| CN104616280A (en) * | 2014-11-26 | 2015-05-13 | 西安电子科技大学 | Image registration method based on maximum stable extreme region and phase coherence |
| CN105005962A (en) * | 2015-08-14 | 2015-10-28 | 南京大学 | Island-reef remote sensing image registration method based on hierarchical screening strategy |
| CN105809678A (en) * | 2016-03-04 | 2016-07-27 | 中国民航大学 | Global matching method for line segment characteristics between two views under short baseline condition |
| CN106529591A (en) * | 2016-11-07 | 2017-03-22 | 湖南源信光电科技有限公司 | Improved MSER image matching algorithm |
| CN107399274A (en) * | 2016-05-06 | 2017-11-28 | 财团法人金属工业研究发展中心 | image superposition method |
| CN108830797A (en) * | 2018-05-24 | 2018-11-16 | 桂林航天工业学院 | A kind of matching line segments method based on affine projection matrix model |
| CN109919958A (en) * | 2019-01-14 | 2019-06-21 | 桂林航天工业学院 | A kind of multiple constraint line segments extraction method based on multi-scale image space |
| CN110136159A (en) * | 2019-04-29 | 2019-08-16 | 辽宁工程技术大学 | Line segments extraction method towards high-resolution remote sensing image |
| CN110470295A (en) * | 2018-05-09 | 2019-11-19 | 北京智慧图科技有限责任公司 | A kind of indoor walking navigation and method based on AR positioning |
| CN110874850A (en) * | 2018-09-04 | 2020-03-10 | 湖北智视科技有限公司 | Real-time unilateral grid feature registration method oriented to target positioning |
| CN111582296A (en) * | 2019-12-20 | 2020-08-25 | 珠海大横琴科技发展有限公司 | Remote sensing image comprehensive matching method and device, electronic equipment and storage medium |
| CN112085771A (en) * | 2020-08-06 | 2020-12-15 | 深圳市优必选科技股份有限公司 | Image registration method and device, terminal equipment and computer readable storage medium |
| CN112163622A (en) * | 2020-09-30 | 2021-01-01 | 山东建筑大学 | Overall situation and local fusion constrained line segment feature matching method for aviation wide-baseline stereopair |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464948A (en) * | 2009-01-14 | 2009-06-24 | 北京航空航天大学 | Object identification method for affine constant moment based on key point |
| CN101650778A (en) * | 2009-07-28 | 2010-02-17 | 复旦大学 | Invariance identification method based on characteristic point and homography matching |
-
2012
- 2012-09-14 CN CN201210342566.3A patent/CN102930525B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101464948A (en) * | 2009-01-14 | 2009-06-24 | 北京航空航天大学 | Object identification method for affine constant moment based on key point |
| CN101650778A (en) * | 2009-07-28 | 2010-02-17 | 复旦大学 | Invariance identification method based on characteristic point and homography matching |
Non-Patent Citations (4)
| Title |
|---|
| LIANG CHENG ET AL: "《A New Method for Remote Sensing Image Matching by Integrating Affine Invariant Feature Extraction and RANSAC》", 《3RD INTERNATIONAL CONGRESS ON IMAGE AND SIGNAL PROCESSING》 * |
| LIANG CHENG ET AL: "《Remote sensing image matching by integrating affine invariant feature extraction and RANSAC》", 《COMPUTERS AND ELECTRICAL ENGINEERING》 * |
| 程亮等: "《遥感影像仿射不变特征匹配的自动优化》", 《武汉大学学报信息科学版》 * |
| 程亮等: "《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》", 《测绘学报》 * |
Cited By (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103353984A (en) * | 2013-04-03 | 2013-10-16 | 西安交通大学 | Method for matching multiple image segments by using non geometric constraints |
| CN103679676A (en) * | 2013-12-02 | 2014-03-26 | 西北工业大学 | Quick unordered image stitching method based on multi-level word bag clustering |
| CN104616280A (en) * | 2014-11-26 | 2015-05-13 | 西安电子科技大学 | Image registration method based on maximum stable extreme region and phase coherence |
| CN104616280B (en) * | 2014-11-26 | 2017-07-28 | 西安电子科技大学 | Method for registering images based on maximum stable extremal region and phase equalization |
| CN104392457A (en) * | 2014-12-11 | 2015-03-04 | 中国测绘科学研究院 | Automatic matching method and device for connection points of slanted images |
| CN104392457B (en) * | 2014-12-11 | 2017-07-11 | 中国测绘科学研究院 | Incline the tie point automatic matching method and device of image |
| CN105005962B (en) * | 2015-08-14 | 2018-01-12 | 南京大学 | Islands and reefs Remote Sensing Image Matching method based on hierarchical screening strategy |
| CN105005962A (en) * | 2015-08-14 | 2015-10-28 | 南京大学 | Island-reef remote sensing image registration method based on hierarchical screening strategy |
| CN105809678A (en) * | 2016-03-04 | 2016-07-27 | 中国民航大学 | Global matching method for line segment characteristics between two views under short baseline condition |
| CN105809678B (en) * | 2016-03-04 | 2018-08-31 | 中国民航大学 | A kind of line segment feature global registration method between two views under short base line condition |
| CN107399274A (en) * | 2016-05-06 | 2017-11-28 | 财团法人金属工业研究发展中心 | image superposition method |
| CN106529591A (en) * | 2016-11-07 | 2017-03-22 | 湖南源信光电科技有限公司 | Improved MSER image matching algorithm |
| CN110470295A (en) * | 2018-05-09 | 2019-11-19 | 北京智慧图科技有限责任公司 | A kind of indoor walking navigation and method based on AR positioning |
| CN110470295B (en) * | 2018-05-09 | 2022-09-30 | 北京智慧图科技有限责任公司 | Indoor walking navigation system and method based on AR positioning |
| CN108830797A (en) * | 2018-05-24 | 2018-11-16 | 桂林航天工业学院 | A kind of matching line segments method based on affine projection matrix model |
| CN110874850A (en) * | 2018-09-04 | 2020-03-10 | 湖北智视科技有限公司 | Real-time unilateral grid feature registration method oriented to target positioning |
| CN109919958A (en) * | 2019-01-14 | 2019-06-21 | 桂林航天工业学院 | A kind of multiple constraint line segments extraction method based on multi-scale image space |
| CN109919958B (en) * | 2019-01-14 | 2023-03-28 | 桂林航天工业学院 | Multi-constraint line segment extraction method based on multi-scale image space |
| CN110136159A (en) * | 2019-04-29 | 2019-08-16 | 辽宁工程技术大学 | Line segments extraction method towards high-resolution remote sensing image |
| CN110136159B (en) * | 2019-04-29 | 2023-03-31 | 辽宁工程技术大学 | Line segment extraction method for high-resolution remote sensing image |
| CN111582296A (en) * | 2019-12-20 | 2020-08-25 | 珠海大横琴科技发展有限公司 | Remote sensing image comprehensive matching method and device, electronic equipment and storage medium |
| CN112085771A (en) * | 2020-08-06 | 2020-12-15 | 深圳市优必选科技股份有限公司 | Image registration method and device, terminal equipment and computer readable storage medium |
| CN112085771B (en) * | 2020-08-06 | 2023-12-05 | 深圳市优必选科技股份有限公司 | Image registration method, device, terminal equipment and computer readable storage medium |
| CN112163622A (en) * | 2020-09-30 | 2021-01-01 | 山东建筑大学 | Overall situation and local fusion constrained line segment feature matching method for aviation wide-baseline stereopair |
| CN112163622B (en) * | 2020-09-30 | 2022-07-05 | 山东建筑大学 | Global and local fusion constrained aviation wide-baseline stereopair line segment matching method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102930525B (en) | 2015-04-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102930525B (en) | Line matching method based on affine invariant feature and homography | |
| CN101216895B (en) | An automatic extracting method for ellipse image features in complex background images | |
| CN102136155B (en) | Object elevation vectorization method and system based on three dimensional laser scanning | |
| CN105184801A (en) | Optical and SAR image high-precision registration method based on multilevel strategy | |
| CN103604417B (en) | The multi-view images bi-directional matching strategy that object space is information constrained | |
| CN103727930B (en) | A kind of laser range finder based on edge matching and camera relative pose scaling method | |
| CN104036480B (en) | Quick elimination Mismatching point method based on surf algorithm | |
| CN104809689A (en) | Building point cloud model and base map aligned method based on outline | |
| CN103606170B (en) | Streetscape image feature based on colored Scale invariant detects and matching process | |
| CN105761308A (en) | Ground LiDAR and image data fused occlusion region building facade reconstruction method | |
| CN102750537A (en) | Automatic registering method of high accuracy images | |
| CN104121902A (en) | Implementation method of indoor robot visual odometer based on Xtion camera | |
| CN103617328A (en) | Aircraft three-dimensional attitude calculation method | |
| CN106156752A (en) | A kind of model recognizing method based on inverse projection three-view diagram | |
| CN105513094A (en) | Stereo vision tracking method and stereo vision tracking system based on 3D Delaunay triangulation | |
| CN111612728A (en) | 3D point cloud densification method and device based on binocular RGB image | |
| CN104167000A (en) | Affine-invariant wide-baseline image dense matching method | |
| CN103578105B (en) | A kind of multi-source foreign peoples method for registering images based on provincial characteristics | |
| CN103632376A (en) | Method for suppressing partial occlusion of vehicles by aid of double-level frames | |
| Li et al. | A new 3D LIDAR-based lane markings recognition approach | |
| CN105184786A (en) | Floating-point-based triangle characteristic description method | |
| Zuo et al. | Multimodal localization: Stereo over LiDAR map | |
| CN105654479A (en) | Multispectral image registering method and multispectral image registering device | |
| Liu et al. | A real-time stereo visual-inertial SLAM system based on point-and-line features | |
| CN112116637B (en) | Automatic power tower detection method and system based on unmanned aerial vehicle 3D laser scanning technology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20150415 Termination date: 20150914 |
|
| EXPY | Termination of patent right or utility model |