CN105551012A - Method and system for reducing wrong matching pair in computer image registration - Google Patents
Method and system for reducing wrong matching pair in computer image registration Download PDFInfo
- Publication number
- CN105551012A CN105551012A CN201410613520.XA CN201410613520A CN105551012A CN 105551012 A CN105551012 A CN 105551012A CN 201410613520 A CN201410613520 A CN 201410613520A CN 105551012 A CN105551012 A CN 105551012A
- Authority
- CN
- China
- Prior art keywords
- coupling
- primary area
- local feature
- module
- feature region
- 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
Landscapes
- Image Analysis (AREA)
Abstract
The invention, which relates to the technology of the computer image registration, discloses a method and system for reducing a wrong matching pair in computer image registration. During the computer image registration, purification is carried out on matching pairs of an obtained local feature point by using main direction information of the local feature point, so that detection and removing of wrong matching pairs can be realized efficiently; and the method can be applied to the system with the high real-time requirement, so that the practical value is substantially improved. In addition, purification is carried out on matching pairs of the obtained local feature point by using scale information of the local feature point or coordinate information of the local feature point, so that the wrong matching pair can be reduced further and thus the system accuracy is improved.
Description
Technical field
The present invention relates to computer picture registration technology, particularly reduce the right method of erroneous matching and system thereof in computer picture registration.
Background technology
Image registration is detect two width that different channel obtains or whether multiple image mates.As shown in Figure 1, general conventional flow process is roughly: from every width figure, detect local feature; Then between these two local feature collection, Match of elemental composition is carried out.In general, be not the coupling that each local feature is corresponding be correct, so to need these couplings carrying out primary purification filtering, finally just to obtain between two images required registration relation (whether mate, the anglec of rotation, scaling, displacement etc.).
Conventional image ratio to and image search system in, conventional method generally all comprises coupling two steps between local shape factor and proper vector.After characteristic matching completes, often add coupling and erroneous matching rate is reduced to the such process of purification.
Local feature coupling can be summed up as the problem of being carried out similarity retrieval by distance function between high dimension vector in essence.Roughly have two class solutions, the first is by the method for exhaustion (linear scanning method), and point and query point by data centralization are carried out distance one by one and compared; The second sets up index to carry out Rapid matching, such as conventional kd tree and kd tree query mode (BBF, Best-Bin-First) etc. improved.And show due to experiment and analysis, the arest neighbors obtained based on local feature region matching way can not ensure that coupling is correct, also needs further filtration, the purification problem that coupling that Here it is is right.The algorithm of general normal employing is ratio method of purification and consistance method of purification.
The logic of ratio method of purification can be summarized as follows: for each unique point of target tightening, its arest neighbors unique point and time neighbour's unique point is obtained at benchmark Integrated query, if meet: nearest neighbor distance >=secondary nearest neighbor distance * Ratio, what then retain that this unique point forms with its arest neighbors mates, otherwise it is right to reject this coupling.General Ratio gets 0.7.
Consistance method of purification, for when ignoring image deformation, having one-to-one relationship, calculating try to achieve by perspective transformation matrix between different visual angles hypograph under same scene.In general this matrix has 8 independent variables, and therefore the minimum unique point (totally 8 coordinate figures) by 4 couplings obtains.Usual use RANSAC algorithm (RandomSampleConsensusAlgorithm is called for short " RANSAC algorithm ") is tried to achieve.In general, still there is the erroneous matching (as shown in Figure 12 and Figure 13) of significant proportion in the match point that ratio method of purification obtains, therefore under the scene that accuracy requirement is higher, after ratio method of purification, a step consistance method of purification is often added again, further filtering Mismatching point.
But, the present inventor finds, consistance method of purification (RANSAC etc.) is although have very high accuracy, may be used for the scene that accuracy requirement is higher, but (consuming time more than 100 milliseconds to the single image of routine) very consuming time, cannot use in real-time system and comparing in enormous quantities.
Summary of the invention
The object of the present invention is to provide in a kind of computer picture registration and reduce the right method of erroneous matching and system thereof, realize the right detection of erroneous matching and removal efficiently, substantially increase practical value.
For solving the problems of the technologies described above, embodiments of the present invention disclose in a kind of computer picture registration and reduce the right method of erroneous matching, and the method comprises the following steps:
Obtain the coupling of the local feature region of target image and the local feature region of reference picture to set;
Obtain the principal direction of this coupling to each coupling centering two local feature region in set, and the principal direction calculated between each coupling centering two local feature region is poor;
By quantizing for the principal direction between coupling centering two local feature region to multiple interval, the burst length in each interval is the value preset;
Add up coupling that each interval comprises to quantity, and choose coupling to the maximum interval of quantity as between the first primary area;
Judge whether the coupling comprised between the first primary area is greater than first threshold to quantity, if the coupling comprised between the first primary area is greater than first threshold to quantity, then export the coupling comprised between the first primary area to gather as first, other coupling abandoned outside this first set is right.
Embodiments of the present invention also disclose in a kind of computer picture registration and reduce the right system of erroneous matching, and system comprises:
Acquisition module, for the coupling of the local feature region of the local feature region and reference picture that obtain target image to set;
First computing module, for obtaining the coupling of acquisition module acquisition to the principal direction of each coupling centering two local feature region in set, and the principal direction calculated between each coupling centering two local feature region is poor;
First quantization modules, quantizing to multiple interval for the principal direction between coupling centering two local feature region of being calculated by the first computing module, the burst length in each interval is the value preset;
First chooses module, for adding up the coupling that comprises of each interval that the first quantization modules quantizes to quantity, and chooses coupling to the maximum interval of quantity as between the first primary area;
First judge module, for judging that first chooses the coupling comprised between the first primary area that module chooses and whether be greater than first threshold to quantity; And
First output module, if confirm that the coupling comprised between the first primary area is greater than first threshold to quantity for the first judge module, export the coupling comprised between the first primary area and gather as first, other coupling abandoned outside this first set is right.
Compared with prior art, the key distinction and effect thereof are embodiment of the present invention:
In computer picture registration of the present invention, utilize the principal direction information of local feature region to the coupling of the local feature region obtained to purifying, the right detection of erroneous matching and removal can be realized efficiently, can be used for the system that requirement of real-time is higher, substantially increase practical value.
Further, utilizing the dimensional information of local feature region to the coupling of the local feature region obtained to purifying, erroneous matching pair can be reduced further, improve system accuracies.
Further, utilizing the coordinate information of local feature region to the coupling of the local feature region obtained to purifying according to registration differential seat angle and registration zoom ratio, erroneous matching pair can be reduced further, improve system accuracies.
Accompanying drawing explanation
Fig. 1 is the schematic flow sheet of existing a kind of image registration.
Fig. 2 is the schematic flow sheet reducing the right method of erroneous matching in first embodiment of the invention in a kind of computer picture registration;
Fig. 3 is the schematic flow sheet reducing the right method of erroneous matching in second embodiment of the invention in a kind of computer picture registration;
Fig. 4 is the schematic flow sheet reducing the right method of erroneous matching in third embodiment of the invention in a kind of computer picture registration;
Fig. 5 is the schematic flow sheet reducing the right method of erroneous matching in third embodiment of the invention in a kind of computer picture registration;
Fig. 6 is the quantification histogram mating right principal direction anglec of rotation difference in third embodiment of the invention;
Fig. 7 is the quantification histogram mating right scaling ratio in third embodiment of the invention;
Fig. 8 is the structural representation reducing the right system of erroneous matching in four embodiment of the invention in a kind of computer picture registration;
Fig. 9 is the structural representation reducing the right system of erroneous matching in fifth embodiment of the invention in a kind of computer picture registration;
Figure 10 is the structural representation reducing the right system of erroneous matching in sixth embodiment of the invention in a kind of computer picture registration;
Figure 11 is the structural representation reducing the right system of erroneous matching in sixth embodiment of the invention in a kind of computer picture registration.
Figure 12 is the matching result of existing employing Feature Points Matching and ratio method of purification.
Figure 13 is the matching result of existing employing Feature Points Matching and ratio method of purification.
Embodiment
In the following description, many ins and outs are proposed in order to make reader understand the application better.But, persons of ordinary skill in the art may appreciate that even without these ins and outs with based on the many variations of following embodiment and amendment, also can realize each claim of the application technical scheme required for protection.
For making the object, technical solutions and advantages of the present invention clearly, below in conjunction with accompanying drawing, embodiments of the present invention are described in further detail.
First embodiment of the invention relates in a kind of computer picture registration and reduces the right method of erroneous matching.Fig. 2 is the schematic flow sheet reducing the right method of erroneous matching in this computer picture registration.As shown in Figure 2, the method comprises the following steps:
In step 201, the coupling of the local feature region of target image and the local feature region of reference picture is obtained to set.Preferably, this coupling to set be through ratio method of purification filter after the coupling that obtains to set.Be appreciated that in other embodiments of the present invention, this coupling also can be that initial matching after local feature coupling is to set or the coupling that obtains is to set after other method of purifications filter to set.
In addition, be appreciated that global characteristics is used to describe the macrofeature of whole image, such as color histogram, its shortcoming is be not suitable for image aliasing and have the situation of blocking.And local feature generally comprises the segment space scope in image, a good local feature needs to possess following character: 1) repeatable; 2) unique; 3) locality; 4) quantitative; 5) accuracy; 6) high efficiency is most important with repeatability again in above-mentioned character.Preferably, above-mentioned local feature region can be the most frequently used SIFT (Scale-invariantfeaturetransform) and SURF (SpeededUpRobustFeatures) feature.For SIFT feature, comprise the X-coordinate of 1 dimension, the Y-coordinate of 1 dimension, the dimensional information of 1 dimension, the principal direction information of 1 dimension and the Feature Descriptor information of 128 dimensions.Certainly, in other embodiments of the present invention, above-mentioned local feature region also can be the local feature of other types, as long as comprise principal direction parameter.
After this enter step 202, obtain the principal direction of this coupling to each coupling centering two local feature region in set, and the principal direction calculated between each coupling centering two local feature region is poor.Be appreciated that, the principal direction of the poor local feature region for target image of the principal direction between each coupling centering two local feature region deducts the principal direction of the local feature region of reference picture, or the principal direction of the local feature region of reference picture deducts the principal direction of the local feature region of target image.
After this enter step 203, by quantizing for the principal direction between coupling centering two local feature region to multiple interval, the burst length in each interval is the value preset.
After this enter step 204, add up coupling that each interval comprises to quantity, and choose coupling to the maximum interval of quantity as between the first primary area.
After this enter step 205, judge whether the coupling comprised between the first primary area is greater than first threshold to quantity, if the coupling comprised between the first primary area is greater than first threshold to quantity, then enters step 206, otherwise process ends.
In step 206, export the coupling comprised between the first primary area and gather as first, other coupling abandoned outside this first set is right.
After this process ends.
In addition, be appreciated that burst length can rule of thumb set.Burst length arranges larger, then the coupling comprised between the first primary area is comparatively large to quantity, and burst length arranges less, then the coupling comprised between the first primary area is less to quantity.Therefore, in various embodiments of the present invention, when screening coupling pair, the setting of burst length can also be coordinated, carry out expanding the coupling obtaining registration correct centered by by between the first primary area right.Specifically:
Further comprising the steps of after step 205:
If the coupling comprised between the first primary area is less than first threshold to quantity, then choose between the first primary area and adjacent two intervals as between the 7th primary area, such as, 0.628,0.942 and 1.256 3 interval shown in Fig. 6;
Judge whether the coupling comprised between the 7th primary area is greater than first threshold to quantity, if the coupling comprised between the 7th primary area is greater than first threshold to quantity, then export the coupling that comprises between the 7th primary area to as the 7th set, other coupling abandoned outside the 7th set is right.
In the computer picture registration of present embodiment, utilize the principal direction information of local feature region to the coupling of the local feature region obtained to purifying, the right detection of erroneous matching and removal can be realized efficiently, can be used for the system that requirement of real-time is higher, substantially increase practical value.
Second embodiment of the invention relates in a kind of computer picture registration and reduces the right method of erroneous matching.Fig. 3 is the schematic flow sheet reducing the right method of erroneous matching in this computer picture registration.
Second embodiment improves on the basis of the first embodiment, main improvements are: utilize the dimensional information of local feature region to the coupling of the local feature region obtained to purifying, erroneous matching pair can be reduced further, improve system accuracies.Specifically:
As shown in Figure 3, further comprising the steps of after step 206:
In step 301, obtain the yardstick of coupling to each coupling centering two local feature region in set, and calculate the scaling ratio between each coupling centering two local feature region.Being appreciated that local feature region can be the most frequently used SIFT, SURF feature etc., also can be the local feature of other types, as long as comprise scale parameter.
In addition, be appreciated that when analyzing unknown scene by a Vision Builder for Automated Inspection, computing machine is had no idea the object yardstick known in advance in image, therefore often needs to consider the description of image under multiple dimensioned simultaneously, knows the best scale of attention object.The such local feature region of similar SIFT, often with dimensional information (scale), represents unique point and responds the strongest under this yardstick.For the unique point that a pair coupling in two width images is correct, the ratio of both yardsticks represents the zoom ratio between image.That is, scaling ratio between each coupling centering two local feature region is the yardstick of yardstick divided by the local feature region of reference picture of the local feature region of target image, or the yardstick of the local feature region of reference picture is divided by the yardstick of the local feature region of target image.
After this step 302 is entered, according to 1/2
nscaling ratio between coupling centering two local feature region is quantized to multiple interval, and wherein N is predefined integer.Preferably, in the present embodiment, the yardstick according to local feature region describes, and selects according to 1/2
nscaling ratio is quantized.In other embodiments of the present invention, also can describe according to the particular dimensions of local feature region, correspondingly carry out adjusting and being arranged to according to such as 1/3
n, (2/3)
netc. quantize.
After this enter step 303, add up coupling that each interval comprises to quantity, and choose coupling to the maximum interval of quantity as between the second primary area.
After this enter step 304, judge whether the coupling comprised between the second primary area is greater than Second Threshold to quantity, if the coupling comprised between the second primary area is greater than Second Threshold to quantity, then enters step 305, otherwise process ends.
In step 305, export the coupling that comprises between the second primary area to as the second set, and export the first set with the second intersection of sets collection as the 3rd set, abandon the 3rd gather outside other coupling right.
After this process ends.
Similarly, when the coupling comprised between the second primary area is less than Second Threshold to quantity, can considers to expand centered by by between the second primary area, namely choose the adjacent interval between the second primary area, right with the coupling obtaining registration correct.In view of similar in concrete mode of operation and the first embodiment, do not repeat them here.
In addition, be appreciated that, in other embodiments of the present invention, also the dimensional information of local feature region can first be utilized to the coupling of the local feature region obtained to purifying, recycling local feature region principal direction information to the coupling of the local feature region obtained to purifying, or simultaneously to the coupling of the local feature region obtained to purifying, then export the first set and the second intersection of sets collection as the 3rd set, other coupling abandoned outside the 3rd set is right.Be not limited to above-mentioned order.
Third embodiment of the invention relates in a kind of computer picture registration and reduces the right method of erroneous matching.Fig. 4 is the schematic flow sheet reducing the right method of erroneous matching in this computer picture registration.
3rd embodiment improves on the basis of the second embodiment, main improvements are: utilize the coordinate information of local feature region to the coupling of the local feature region obtained to purifying according to registration differential seat angle and registration zoom ratio, erroneous matching pair can be reduced further, improve system accuracies.Specifically:
As shown in Figure 4, further comprising the steps of after step 305:
In step 401, the mean value calculating the principal direction difference between all coupling centering two local feature region in the first set is as the registration differential seat angle θ of target image and reference picture, and the mean value calculating the scaling ratio between all coupling centerings two local feature region in the second set is as the registration zoom ratio S of target image and reference picture.
After this step 402 is entered, obtain X-coordinate and the Y-coordinate <pi.x of each coupling centering two local feature region in the 3rd set, pi.y, qj.x, qj.y>, and calculate coupling to displacement in x and y direction according to registration differential seat angle θ and registration zoom ratio S
Being appreciated that local feature region can be the most frequently used SIFT, SURF feature etc., also can be the local feature of other types, as long as comprise displacement parameter.In addition, in other embodiments of the present invention, also the coordinate information of local feature region can be utilized to purify to set to the coupling obtained in step 201 according to registration differential seat angle and registration zoom ratio.
In addition, be appreciated that, each coupling is the coordinate that the coordinate of the local feature region of target image deducts the local feature region of reference picture to displacement in x and y direction, or the coordinate of the local feature region of reference picture deducts the coordinate of the local feature region of target image, and calculate again after corresponding conversion being carried out to coordinate according to registration differential seat angle θ and registration zoom ratio S.
After this enter step 403, coupling is quantized to multiple interval respectively to displacement in x and y direction, in X and Y-direction, the burst length in each interval is the value preset.
After this enter step 404, in statistics X and Y-direction, the coupling that comprises of each interval is to quantity, and chooses and X and Y-direction mate to the maximum interval of quantity respectively as between the 4th primary area and between the 5th primary area.
After this step 405 is entered, judge whether the coupling comprised between the 4th primary area is greater than the 4th threshold value to quantity and whether the coupling comprised between the 5th primary area is greater than the 5th threshold value to quantity, if the coupling comprised between the 4th primary area is greater than the 4th threshold value to quantity and the coupling comprised between the 5th primary area is greater than the 5th threshold value to quantity, then enter step 406, otherwise process ends.
In a step 406, export between the 4th primary area and the coupling comprised between the 5th primary area to respectively as the 4th set and the 5th set, and export the 4th set with the 5th intersection of sets collection as the 6th set, other coupling abandoned outside the 6th set is right.
After this process ends.
Preferably, further comprising the steps of after step 406:
Calculate respectively the 4th set and the 5th gather in all couplings to the registration displacement T of the mean value of displacement in x and y direction as target image and reference picture
xand T
y.
Be appreciated that in other embodiments of the present invention, also can not calculate registration displacement Tx and Ty.
Similarly, the coupling comprised between the 4th primary area is less than the coupling comprised between the 4th threshold value or the 5th primary area when being less than the 5th threshold value to quantity to quantity, can consider to expand centered by by between the 4th primary area or between the 5th primary area, namely the adjacent interval between the 4th primary area or between the 5th primary area is chosen, right with the coupling obtaining registration correct.In view of similar in concrete mode of operation and the first embodiment, do not repeat them here.
In a preference of the present invention, have employed the description of SIFT feature, as shown in the table:
Table 1 symbol
| Title | Represent |
| Input picture A and B subject to registration | Figure A, figure B |
| N the unique point of input picture A | {p i}i＝1，2，……，n |
| M the unique point of input picture B | {q j}j＝1，2，……，m |
| Unique point p iX-coordinate | p i.x |
| Unique point p iY-coordinate | p i.y |
| Unique point p iYardstick | p i.scale |
| Unique point p iPrincipal direction | p i.orientation |
| Unique point p iDescriptor | p i.desc j j＝1、2、……、128 |
As shown in Figure 5:
1. input figure A and figure B coupling of obtaining after ratio method of purification filters right: <p
i, q
j>, j=1,2 ..., m.Certainly, in other embodiments of the invention, mate to also can be initial matching after local feature coupling to or after other method of purifications filter, the coupling that obtains is right.
2. mate right principal direction in obtaining step 1: <p
i.orientation, q
j.orientation>, j=1,2 ..., m.
3. mate the right principal direction anglec of rotation in calculation procedure 2 poor: q
j.rotation=p
i.orientation-q
j.orientation, j=1,2 ..., m.
4. a series of values that pair step 3 exports carry out quantification and statistic histogram (as shown in Figure 6), and concrete grammar is:
1) burst length that setting quantizes is L
1.
2) it is poor that each group exported step 3 mates the right principal direction anglec of rotation, calculates the interval K dropped on after it quantizes
j=q
j.rotation/L
1, j=1,2 ..., m.
3) add up the number of pairs that each interval comprises, choose to comprise and mate at most right interval as between primary area; The coupling comprised in interval, to alternatively gathering, is designated as U
1; By U
1the number that interior coupling is right is designated as C
1; Calculate U
1the mean value of the principal direction anglec of rotation difference that interior each coupling is right, as figure A and the registration differential seat angle of scheming B, is designated as θ.
5. mate right yardstick in obtaining step 1: <p
i.scale, q
j.scale>, j=1,2 ..., m.
6. mate right scaling ratio in calculation procedure 5: q
j.ratio=q
j.scale/p
i.scale.
7. a series of values that pair step 6 exports carry out quantification and statistic histogram (as shown in Figure 7), and concrete grammar is:
1) according to 1/2
ngenerate a series of quantification space: such as 1/8,1/4,1/2,1,2,4,8 ... Deng.Certainly, in other embodiments of the invention, also can according to such as 1/3
n, (2/3)
netc. quantize.
2) to the right scaling ratio of each group coupling that step 6 exports, calculate drop on previous step after it quantizes designation area between.
3) add up the number of pairs that each interval comprises, choose to comprise and mate at most right interval as between primary area; The coupling comprised in interval, to alternatively gathering, is designated as U
2; By U
2the number that interior coupling is right is designated as C
2; Calculate U
2the mean value of the scaling ratio that interior each coupling is right, as figure A and the registration zoom ratio scheming B, is designated as S.
8. the C will obtained in step 4 and step 7
1and C
2with the threshold value T preset
1and T
2compare, if met: C
1> T
1aMP.AMp.Amp C
2> T
2, then U is exported
1and U
2common factor alternatively characteristic set, be designated as U
3; U
3the number of pairs comprised, is designated as k; θ and S is respectively registration differential seat angle and registration zoom ratio, and enters step 9; Otherwise exporting registration is false, and exits flow process.
9. obtain U
3in (or in step 1) mate right X-coordinate and Y-coordinate: <p
i.x, p
i.y, q
j.x, q
j.y>, j=1,2 ..., k.
10. according to projective transformation matrix
J=1,2 ..., k, the displacement that each coupling is right can be calculated:
j＝1、2、……、k。
11. use the method for similar step 4, displacement quantization previous step obtained statistic histogram, and concrete grammar is:
1) q is set
j.T
xand q
j.T
ythe burst length quantized is respectively L
xand L
y
2) right to each coupling in the set of step 10 output, calculate q
j.T
xand q
j.T
yquantized value, according to following formula:
K
x,j＝q
j.T
x/L
x
j＝1，2，……，k
K
y,j＝q
j.T
y/L
y
3) add up the number of pairs that each interval comprises, choose to comprise and mate at most right interval as between primary area; By the coupling comprised in interval to alternatively gathering, be designated as U
4and U
5(displacement in corresponding X and Y both direction); Respectively by U
4and U
5mate right number and be designated as C
4and C
5; Calculate U respectively
4and U
5the mean value of the displacement of the right X-direction of interior each coupling and Y-direction, as figure A and the registration displacement scheming B, is designated as T
xand T
y.
12. C that step 11 is exported
4and C
5with the threshold value T preset
4and T
5compare, if met: C
4> T
4aMP.AMp.Amp C
5> T
5, then U is exported
4and U
5common factor U
6the coupling correct as registration is right; U
6inside comprise the number C that coupling is right
6as similarity; θ, S, T
xand T
ybe respectively registration differential seat angle, registration zoom ratio, registration X-direction top offset and registration Y-direction top offset.
Can see from above, the preference of said method based on cardinal principle be: if the series of features points of two input pictures (figure A and figure B) are to ({ p
ii=1,2 ..., n and { q
jj=1,2 ..., be m) correct coupling, should following characteristics be had between the feature that so this series of points is right: 1, there is identical or close scaling ratio; 2, there is the identical or close principal direction anglec of rotation; 3, there is identical or close coordinate displacement.Thus propose a kind of completely newly, utilize coupling between the method for purification of the information such as yardstick, the anglec of rotation, geometry distribution, can under the prerequisite ensureing the precision suitable with RANSAC, efficiently (consuming time large about about 1 millisecond, reduce 2 orders of magnitude than the former) realize the right detection of error matching points and removal, substantially increase accuracy rate and the practical value of system.
Image registration is widely used in the fields such as remotely-sensed data analysis, computer vision, image procossing, such as, map match, brand recognition etc. in scenery coupling, Aerial vehicle position system.Figure 12 and Figure 13 respectively illustrates the example adopting Feature Points Matching and the ratio method of purification trade mark to clothes and wrist-watch to carry out similarity identification.Can see, although the most of matching result in Figure 12 is correct, the not erroneous matching of exclusive segment, and in Figure 13, be a typical erroneous matching result, then in identified similar brand, be mixed into some complete incoherent trade marks.Utilize the coupling mentioned in the present embodiment between yardstick, the anglec of rotation and coordinate information the matching result in Figure 12 and Figure 13 is purified further, the right detection of error matching points and removal will be realized efficiently, namely detect and remove the error matching points of mating with hair in Figure 12, and detect and remove the erroneous matching in Figure 13, thus improve the accuracy rate of brand recognition.
Be appreciated that and these are only preference of the present invention.In other embodiments of the invention, also can adopt the local feature of other types, as long as comprise principal direction, yardstick and displacement parameter.Further, only can utilizing principal direction information or only utilize dimensional information the coupling of local feature region purifying, can reach and reduce the right effect of erroneous matching.Be not limited to above-mentioned optimum configurations and order.
Each method embodiment of the present invention all can realize in modes such as software, hardware, firmwares.No matter the present invention realizes with software, hardware or firmware mode, instruction code can be stored in the addressable storer of computing machine of any type (such as permanent or revisable, volatibility or non-volatile, solid-state or non-solid, fixing or removable medium etc.).Equally, storer can be such as programmable logic array (ProgrammableArrayLogic, be called for short " PAL "), random access memory (RandomAccessMemory, be called for short " RAM "), programmable read only memory (ProgrammableReadOnlyMemory, be called for short " PROM "), ROM (read-only memory) (Read-OnlyMemory, be called for short " ROM "), Electrically Erasable Read Only Memory (ElectricallyErasableProgrammableROM, be called for short " EEPROM "), disk, CD, digital versatile disc (DigitalVersatileDisc, be called for short " DVD ") etc.
Four embodiment of the invention relates in a kind of computer picture registration and reduces the right system of erroneous matching.Fig. 8 is the structural representation reducing the right system of erroneous matching in this computer picture registration.As shown in Figure 8, this system comprises:
Acquisition module, for the coupling of the local feature region of the local feature region and reference picture that obtain target image to set.
First computing module, for obtaining the coupling of acquisition module acquisition to the principal direction of each coupling centering two local feature region in set, and the principal direction calculated between each coupling centering two local feature region is poor.
First quantization modules, quantizing to multiple interval for the principal direction between coupling centering two local feature region of being calculated by the first computing module, the burst length in each interval is the value preset.
First chooses module, for adding up the coupling that comprises of each interval that the first quantization modules quantizes to quantity, and chooses coupling to the maximum interval of quantity as between the first primary area.
First judge module, for judging that first chooses the coupling comprised between the first primary area that module chooses and whether be greater than first threshold to quantity.And
First output module, if confirm that the coupling comprised between the first primary area is greater than first threshold to quantity for the first judge module, export the coupling comprised between the first primary area and gather as first, other coupling abandoned outside this first set is right.
In addition, be appreciated that burst length can rule of thumb set.Burst length arranges larger, then the coupling comprised between the first primary area is comparatively large to quantity, and burst length arranges less, then the coupling comprised between the first primary area is less to quantity.Therefore, in various embodiments of the present invention, when screening coupling pair, the setting of burst length can also be coordinated, carry out expanding the coupling obtaining registration correct centered by by between the first primary area right.Specifically:
Said system also comprises:
4th chooses module, if confirm that the coupling comprised between the first primary area is less than first threshold to quantity for the first judge module, choose between the first primary area and adjacent two intervals as between the 7th primary area.
4th judge module, for judging that the 4th chooses the coupling comprised between the 7th primary area that module chooses and whether be greater than first threshold to quantity.And
4th output module, if confirm that the coupling comprised between the 7th primary area is greater than first threshold to quantity for the 4th judge module, export the coupling that comprises between the 7th primary area to as the 7th set, other coupling abandoned outside the 7th set is right.
In the system of present embodiment, each module utilizes the principal direction information of local feature region to the coupling of the local feature region obtained to purifying, the right detection of erroneous matching and removal can be realized efficiently, can be used for the system that requirement of real-time is higher, substantially increase practical value.
First embodiment is the method embodiment corresponding with present embodiment, and present embodiment can be worked in coordination with the first embodiment and be implemented.The relevant technical details mentioned in first embodiment is still effective in the present embodiment, in order to reduce repetition, repeats no more here.Correspondingly, the relevant technical details mentioned in present embodiment also can be applicable in the first embodiment.
Fifth embodiment of the invention relates in a kind of computer picture registration and reduces the right system of erroneous matching.Fig. 9 is the structural representation reducing the right system of erroneous matching in this computer picture registration.
5th embodiment improves on the basis of the 4th embodiment, main improvements are: utilize the dimensional information of local feature region to the coupling of the local feature region obtained to purifying, erroneous matching pair can be reduced further, improve system accuracies.Specifically:
As shown in Figure 9, this system also comprises:
Second computing module, for obtaining the coupling of acquisition module acquisition to the yardstick of each coupling centering two local feature region in set, and calculates the scaling ratio between each coupling centering two local feature region.
Second quantization modules, for according to 1/2
nscaling ratio between coupling centering two local feature region calculate the second computing module is quantized to multiple interval, and wherein N is predefined integer.Preferably, in the present embodiment, the yardstick according to local feature region describes, and selects according to 1/2
nscaling ratio is quantized.In other embodiments of the present invention, also can describe according to the particular dimensions of local feature region, correspondingly carry out adjusting and being arranged to according to such as 1/3
n, (2/3)
netc. quantize.
Second chooses module, for adding up the coupling that comprises of each interval that the second quantization modules quantizes to quantity, and chooses coupling to the maximum interval of quantity as between the second primary area.
Second judge module, for judging that second chooses the coupling comprised between the second primary area that module chooses and whether be greater than Second Threshold to quantity.And
Second output module, exports the coupling that comprises between the second primary area to as the second set, and exports the first set that above-mentioned first output module exports with the second intersection of sets collection as the 3rd set, and other coupling abandoned outside the 3rd set is right.
Similarly, when the coupling comprised between the second primary area is less than Second Threshold to quantity, can considers to expand centered by interpolation module is by between the second primary area, namely choose the adjacent interval between the second primary area, right with the coupling obtaining registration correct.In view of similar in concrete structure and the 4th embodiment, do not repeat them here.
Second embodiment is the method embodiment corresponding with present embodiment, and present embodiment can be worked in coordination with the second embodiment and be implemented.The relevant technical details mentioned in second embodiment is still effective in the present embodiment, in order to reduce repetition, repeats no more here.Correspondingly, the relevant technical details mentioned in present embodiment also can be applicable in the second embodiment.
Sixth embodiment of the invention relates in a kind of computer picture registration and reduces the right system of erroneous matching.Figure 10 is the structural representation reducing the right system of erroneous matching in this computer picture registration.
6th embodiment improves on the basis of the 5th embodiment, main improvements are: utilize the coordinate information of local feature region to the coupling of the local feature region obtained to purifying based on registration differential seat angle and registration zoom ratio, erroneous matching pair can be reduced further, improve system accuracies.Specifically:
As shown in Figure 10, said system also comprises:
4th computing module, for the mean value that calculates the principal direction difference between all coupling centering two local feature region in the first set that above-mentioned first output module the exports registration differential seat angle θ as target image and reference picture, and the mean value calculating the scaling ratio between all coupling centerings two local feature region in the second set that above-mentioned second output module exports is as the registration zoom ratio S of target image and reference picture.
3rd computing module, for obtaining X-coordinate and the Y-coordinate <pi.x of each coupling centering two local feature region in the 3rd set that the second output module exports, pi.y, qj.x, qj.y>, and according to the 4th computing module calculate registration differential seat angle θ and registration zoom ratio S calculate coupling to displacement in x and y direction
Be appreciated that in other embodiments of the present invention, also the coordinate information of local feature region can be utilized to purify to set to the coupling that above-mentioned acquisition module obtains according to registration differential seat angle and registration zoom ratio.
3rd quantization modules, is quantized to multiple interval for the coupling calculated by the 3rd computing module respectively to displacement in x and y direction, and in X and Y-direction, the burst length in each interval is the value preset.
3rd chooses module, for adding up on the X and Y-direction that the 3rd quantization modules quantizes the coupling that comprises of each interval to quantity, and chooses and X and Y-direction mates to the maximum interval of quantity respectively as between the 4th primary area and between the 5th primary area.
3rd judge module, for judging that the 3rd chooses the coupling that comprises between the 4th primary area that module chooses and whether be greater than the 4th threshold value to quantity and the 3rd choose the coupling comprised between the 5th primary area that module chooses and whether be greater than the 5th threshold value to quantity.And
3rd output module, if confirm that the coupling that comprises between the 4th primary area is greater than the 4th threshold value to quantity and the coupling comprised between the 5th primary area is greater than the 5th threshold value to quantity for the 3rd judge module, export the coupling comprised between the 4th primary area and between the 5th primary area to gather and the 5th set respectively as the 4th, and exporting the 4th set with the 5th intersection of sets collection as the 6th set, other coupling abandoned outside the 6th set is right.
Preferably, said system also comprises:
5th computing module, the 4th set and the 5th exported for calculating above-mentioned 3rd output module respectively gather in all couplings to registration displacement Tx and Ty of the mean value of displacement in x and y direction as target image and reference picture.
Be appreciated that in other embodiments of the present invention, also can not calculate registration displacement T
xand T
y.
Similarly, the coupling comprised between the 4th primary area is less than the coupling comprised between the 4th threshold value or the 5th primary area when being less than the 5th threshold value to quantity to quantity, can consider to expand centered by interpolation module is by between the 4th primary area or between the 5th primary area, namely the adjacent interval between the 4th primary area or between the 5th primary area is chosen, right with the coupling obtaining registration correct.In view of similar in concrete structure and the 4th embodiment, do not repeat them here.
In a preference of the present invention, as shown in figure 11, said system comprises principal direction unit, multi-scale and coordinate unit, comprises each module shown in Fig. 8-10 respectively, thus can optionally utilize each information to purify to set to coupling.
3rd embodiment is the method embodiment corresponding with present embodiment, and present embodiment can be worked in coordination with the 3rd embodiment and be implemented.The relevant technical details mentioned in 3rd embodiment is still effective in the present embodiment, in order to reduce repetition, repeats no more here.Correspondingly, the relevant technical details mentioned in present embodiment also can be applicable in the 3rd embodiment.
It should be noted that, the each module mentioned in the present invention's each equipment embodiment or unit are all logic module or logical block, physically, a logic module or logical block can be a physical module or physical location, also can be a part for a physical module or physical location, can also realize with the combination of multiple physical module or physical location, the Physical realization of these logic modules or logical block itself is not most important, the combination of the function that these logic modules or logical block realize is only the key solving technical matters proposed by the invention.In addition, in order to outstanding innovative part of the present invention, the module not too close with solving technical matters relation proposed by the invention or unit are not introduced by the above-mentioned each equipment embodiment of the present invention, and this does not show that the said equipment embodiment does not exist other module or unit.
It should be noted that, in the claim and instructions of this patent, the such as relational terms of first and second grades and so on is only used for an entity or operation to separate with another entity or operational zone, and not necessarily requires or imply the relation that there is any this reality between these entities or operation or sequentially.And, term " comprises ", " comprising " or its any other variant are intended to contain comprising of nonexcludability, thus make to comprise the process of a series of key element, method, article or equipment and not only comprise those key elements, but also comprise other key elements clearly do not listed, or also comprise by the intrinsic key element of this process, method, article or equipment.When not more restrictions, the key element " being comprised " limited by statement, and be not precluded within process, method, article or the equipment comprising described key element and also there is other identical element.
Although by referring to some of the preferred embodiment of the invention, to invention has been diagram and describing, but those of ordinary skill in the art should be understood that and can do various change to it in the form and details, and without departing from the spirit and scope of the present invention.
Claims (10)
1. reduce the right method of erroneous matching in computer picture registration, it is characterized in that, the method comprises the following steps:
Obtain the coupling of the local feature region of target image and the local feature region of reference picture to set;
Obtain the principal direction of this coupling to each coupling centering two local feature region in set, and the principal direction calculated between each coupling centering two local feature region is poor;
By quantizing for the principal direction between described coupling centering two local feature region to multiple interval, the burst length in each interval is the value preset;
Add up coupling that each interval comprises to quantity, and choose coupling to the maximum interval of quantity as between the first primary area;
Judge whether the coupling comprised between described first primary area is greater than first threshold to quantity, if the coupling comprised between described first primary area is greater than first threshold to quantity, then export the coupling comprised between described first primary area to gather as first, other coupling abandoned outside this first set is right.
2. reduce the right method of erroneous matching in computer picture registration according to claim 1, it is characterized in that, exporting the coupling that comprises between described first primary area to as the first set, abandon outside this first set other mate right step after further comprising the steps of:
Obtain the yardstick of described coupling to each coupling centering two local feature region in set, and calculate the scaling ratio between each coupling centering two local feature region;
According to 1/2
nscaling ratio between described coupling centering two local feature region is quantized to multiple interval, and wherein N is predefined integer;
Add up coupling that each interval comprises to quantity, and choose coupling to the maximum interval of quantity as between the second primary area;
Judge whether the coupling comprised between described second primary area is greater than Second Threshold to quantity, if the coupling comprised between described second primary area is greater than Second Threshold to quantity, then export the coupling comprised between described second primary area to gather as second, and export described first set with described second intersection of sets collection as the 3rd set, other coupling abandoned outside the 3rd set is right.
3. reduce the right method of erroneous matching in computer picture registration according to claim 2, it is characterized in that, gather as the 3rd with described second intersection of sets collection in described first set of output, abandon the 3rd gather outside other mate right step after further comprising the steps of:
The mean value calculating the principal direction difference between all coupling centering two local feature region in described first set is as the registration differential seat angle θ of described target image and described reference picture, and the mean value calculating the scaling ratio between all coupling centerings two local feature region in described second set is as the registration zoom ratio S of described target image and described reference picture;
Obtain X-coordinate and the Y-coordinate <pi.x of each coupling centering two local feature region in described 3rd set, pi.y, qj.x, qj.y>, and calculate described coupling to displacement in x and y direction according to described registration differential seat angle θ and described registration zoom ratio S
Described coupling is quantized to multiple interval respectively to displacement in x and y direction, and in X and Y-direction, the burst length in each interval is the value preset;
In statistics X and Y-direction, the coupling that comprises of each interval is to quantity, and chooses and X and Y-direction mate to the maximum interval of quantity respectively as between the 4th primary area and between the 5th primary area;
Judge whether the coupling comprised between described 4th primary area is greater than the 4th threshold value to quantity and whether the coupling comprised between described 5th primary area is greater than the 5th threshold value to quantity, if the coupling comprised between described 4th primary area is greater than the 4th threshold value to quantity and the coupling comprised between described 5th primary area is greater than the 5th threshold value to quantity, then export the coupling comprised between described 4th primary area and between described 5th primary area to gather and the 5th set respectively as the 4th, and export the 4th set with the 5th intersection of sets collection as the 6th set, other coupling abandoned outside the 6th set is right.
4. reduce the right method of erroneous matching in computer picture registration according to claim 1, it is characterized in that, further comprising the steps of after in the coupling judging to comprise between described first primary area whether quantity being greater than to the step of first threshold:
If the coupling comprised between described first primary area is less than first threshold to quantity, then choose between described first primary area and adjacent two intervals as between the 7th primary area;
Judge whether the coupling comprised between described 7th primary area is greater than first threshold to quantity, if the coupling comprised between described 7th primary area is greater than first threshold to quantity, then export the coupling that comprises between described 7th primary area to as the 7th set, other coupling abandoned outside the 7th set is right.
5. reduce the right method of erroneous matching in computer picture registration according to claim 3, it is characterized in that, gather as the 6th with the 5th intersection of sets collection in output the 4th set, abandon the 6th gather outside other mate right step after further comprising the steps of:
Calculate respectively described 4th set and the described 5th gather in all couplings to the registration displacement T of the mean value of displacement in x and y direction as described target image and described reference picture
xand T
y.
6. reduce the right system of erroneous matching in computer picture registration, it is characterized in that, described system comprises:
Acquisition module, for the coupling of the local feature region of the local feature region and reference picture that obtain target image to set;
First computing module, for obtaining the coupling of described acquisition module acquisition to the principal direction of each coupling centering two local feature region in set, and the principal direction calculated between each coupling centering two local feature region is poor;
First quantization modules, quantizing to multiple interval for the described principal direction of mating between centering two local feature region calculated by described first computing module, the burst length in each interval is the value preset;
First chooses module, for adding up the coupling that comprises of each interval that described first quantization modules quantizes to quantity, and chooses coupling to the maximum interval of quantity as between the first primary area;
First judge module, for judging that described first chooses the coupling comprised between the first primary area that module chooses and whether be greater than first threshold to quantity; And
First output module, if confirm that the coupling comprised between described first primary area is greater than first threshold to quantity for described first judge module, export the coupling comprised between described first primary area to gather as first, other coupling abandoned outside this first set is right.
7. reduce the right system of erroneous matching in computer picture registration according to claim 6, it is characterized in that, described system also comprises:
Second computing module, for obtaining the coupling of described acquisition module acquisition to the yardstick of each coupling centering two local feature region in set, and calculates the scaling ratio between each coupling centering two local feature region;
Second quantization modules, for according to 1/2
nscaling ratio between described coupling centering two local feature region calculated by described second computing module is quantized to multiple interval, and wherein N is predefined integer;
Second chooses module, for adding up the coupling that comprises of each interval that described second quantization modules quantizes to quantity, and chooses coupling to the maximum interval of quantity as between the second primary area;
Second judge module, for judging that described second chooses the coupling comprised between the second primary area that module chooses and whether be greater than Second Threshold to quantity; And
Second output module, if confirm that the coupling comprised between described second primary area is greater than Second Threshold to quantity for described second judge module, export the coupling comprised between described second primary area to gather as second, and export described first output module output first gathers with described second intersection of sets collection as the 3rd set, and other coupling abandoned outside the 3rd set is right.
8. reduce the right system of erroneous matching in computer picture registration according to claim 7, it is characterized in that, described system also comprises:
4th computing module, the mean value calculating the principal direction difference between all coupling centering two local feature region in the first set that described first output module exports is as the registration differential seat angle θ of described target image and described reference picture, and the mean value calculating the scaling ratio between all coupling centerings two local feature region in the second set that described second output module exports is as the registration zoom ratio S of described target image and described reference picture;
3rd computing module, for obtaining X-coordinate and the Y-coordinate <pi.x of each coupling centering two local feature region in the 3rd set that described second output module exports, pi.y, qj.x, qj.y>, and according to described 4th computing module calculate registration differential seat angle θ and registration zoom ratio S calculate described coupling to displacement in x and y direction
3rd quantization modules, be quantized to multiple interval for the described coupling calculated by described 3rd computing module respectively to displacement in x and y direction, in X and Y-direction, the burst length in each interval is the value preset;
3rd chooses module, for adding up on the X and Y-direction that described 3rd quantization modules quantizes the coupling that comprises of each interval to quantity, and chooses and X and Y-direction mates to the maximum interval of quantity respectively as between the 4th primary area and between the 5th primary area;
3rd judge module, for judging that the described 3rd chooses the coupling that comprises between the 4th primary area that module chooses and whether be greater than the 4th threshold value to quantity and the described 3rd choose the coupling comprised between the 5th primary area that module chooses and whether be greater than the 5th threshold value to quantity; And
3rd output module, if confirm that the coupling that comprises between described 4th primary area is greater than the 4th threshold value to quantity and the coupling comprised between described 5th primary area is greater than the 5th threshold value to quantity for described 3rd judge module, export the coupling comprised between described 4th primary area and between described 5th primary area to gather and the 5th set respectively as the 4th, and exporting the 4th set with the 5th intersection of sets collection as the 6th set, other coupling abandoned outside the 6th set is right.
9. reduce the right system of erroneous matching in computer picture registration according to claim 6, it is characterized in that, described system also comprises:
4th chooses module, if confirm that the coupling comprised between described first primary area is less than first threshold to quantity for described first judge module, choose between described first primary area and adjacent two intervals as between the 7th primary area;
4th judge module, for judging that the described 4th chooses the coupling comprised between the 7th primary area that module chooses and whether be greater than described first threshold to quantity; And
4th output module, if confirm that the coupling comprised between described 7th primary area is greater than first threshold to quantity for described 4th judge module, export the coupling that comprises between described 7th primary area to as the 7th set, other coupling abandoned outside the 7th set is right.
10. reduce the right system of erroneous matching in computer picture registration according to claim 8, it is characterized in that, described system also comprises:
5th computing module, the 4th set and the 5th exported for calculating described 3rd output module respectively gather in all couplings to the registration displacement T of the mean value of displacement in x and y direction as described target image and described reference picture
xand T
y.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410613520.XA CN105551012B (en) | 2014-11-04 | 2014-11-04 | The method and its system of erroneous matching pair are reduced in computer picture registration |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410613520.XA CN105551012B (en) | 2014-11-04 | 2014-11-04 | The method and its system of erroneous matching pair are reduced in computer picture registration |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105551012A true CN105551012A (en) | 2016-05-04 |
| CN105551012B CN105551012B (en) | 2019-04-05 |
Family
ID=55830188
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410613520.XA Active CN105551012B (en) | 2014-11-04 | 2014-11-04 | The method and its system of erroneous matching pair are reduced in computer picture registration |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105551012B (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107980140A (en) * | 2017-10-16 | 2018-05-01 | 厦门中控智慧信息技术有限公司 | A kind of recognition methods of vena metacarpea and device |
| CN108171738A (en) * | 2018-01-25 | 2018-06-15 | 北京雅森科技发展有限公司 | Multimodal medical image registration method based on brain function template |
| CN112116622A (en) * | 2020-09-25 | 2020-12-22 | 贝壳技术有限公司 | Contour extraction method, search library establishment method, contour matching method and system |
| CN112990231A (en) * | 2021-03-12 | 2021-06-18 | 河北经贸大学 | Method, device and terminal for removing image mismatching pairs |
| CN113409372A (en) * | 2021-06-25 | 2021-09-17 | 浙江商汤科技开发有限公司 | Image registration method, related device, equipment and storage medium |
| CN113727095A (en) * | 2021-08-27 | 2021-11-30 | 杭州萤石软件有限公司 | Method, device, equipment and system for detecting movement of camera and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011013862A1 (en) * | 2009-07-28 | 2011-02-03 | 주식회사 유진로봇 | Control method for localization and navigation of mobile robot and mobile robot using same |
| CN103823887A (en) * | 2014-03-10 | 2014-05-28 | 北京大学 | Based on low-order overall situation geometry consistency check error match detection method |
-
2014
- 2014-11-04 CN CN201410613520.XA patent/CN105551012B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2011013862A1 (en) * | 2009-07-28 | 2011-02-03 | 주식회사 유진로봇 | Control method for localization and navigation of mobile robot and mobile robot using same |
| CN103823887A (en) * | 2014-03-10 | 2014-05-28 | 北京大学 | Based on low-order overall situation geometry consistency check error match detection method |
Non-Patent Citations (3)
| Title |
|---|
| 曾峦等: "基于SIFT的自动匹配策略", 《光电工程》 * |
| 胡春海等: "视差生长与张量相结合的多基线稠密匹配", 《光电工程》 * |
| 顾大龙等: "基于SURF的图像匹配算法改进", 《现代电子技术》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107980140A (en) * | 2017-10-16 | 2018-05-01 | 厦门中控智慧信息技术有限公司 | A kind of recognition methods of vena metacarpea and device |
| CN107980140B (en) * | 2017-10-16 | 2021-09-14 | 厦门熵基科技有限公司 | Palm vein identification method and device |
| CN108171738A (en) * | 2018-01-25 | 2018-06-15 | 北京雅森科技发展有限公司 | Multimodal medical image registration method based on brain function template |
| CN108171738B (en) * | 2018-01-25 | 2022-02-01 | 北京雅森科技发展有限公司 | Multi-modal medical image registration method based on brain function template |
| CN112116622A (en) * | 2020-09-25 | 2020-12-22 | 贝壳技术有限公司 | Contour extraction method, search library establishment method, contour matching method and system |
| CN112990231A (en) * | 2021-03-12 | 2021-06-18 | 河北经贸大学 | Method, device and terminal for removing image mismatching pairs |
| CN112990231B (en) * | 2021-03-12 | 2022-08-26 | 河北经贸大学 | Method, device and terminal for removing image mismatching pairs |
| CN113409372A (en) * | 2021-06-25 | 2021-09-17 | 浙江商汤科技开发有限公司 | Image registration method, related device, equipment and storage medium |
| CN113727095A (en) * | 2021-08-27 | 2021-11-30 | 杭州萤石软件有限公司 | Method, device, equipment and system for detecting movement of camera and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105551012B (en) | 2019-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tang et al. | Quadtree attention for vision transformers | |
| CN105551012A (en) | Method and system for reducing wrong matching pair in computer image registration | |
| Kong et al. | Generalizing Laplacian of Gaussian filters for vanishing-point detection | |
| CN102521838B (en) | Image searching/matching method and system for same | |
| EP2724295B1 (en) | System and method for identifying scale invariant features of object outlines on images | |
| CN109060836A (en) | High-pressure oil pipe joint external screw thread detection method based on machine vision | |
| CN102722731A (en) | Efficient image matching method based on improved scale invariant feature transform (SIFT) algorithm | |
| CN109410255A (en) | A kind of method for registering images and device based on improved SIFT and hash algorithm | |
| CN106503704B (en) | Circular traffic sign localization method in a kind of natural scene | |
| CN113177959B (en) | QR code real-time extraction method in rapid movement process | |
| CN104978709A (en) | Descriptor generation method and apparatus | |
| CN103955690A (en) | Method for constructing compact image local feature descriptor | |
| TWI521448B (en) | Vehicle identification system and method | |
| CN105389774A (en) | Method and device for aligning images | |
| CN103994786A (en) | Image detecting method for arc ruler lines of pointer instrument scale | |
| CN103425988A (en) | Real-time positioning and matching method with arc geometric primitives | |
| Yuan et al. | Road segmentation in aerial images by exploiting road vector data | |
| CN104992400A (en) | Method and device for multispectral image registration | |
| CN112085708B (en) | Method and equipment for detecting defects of straight line edges in outer contour of product | |
| CN104978582A (en) | Contour chord angle feature based identification method for blocked target | |
| CN105493105A (en) | Keypoint identification | |
| CN103578105A (en) | Method for multisource different-type image registration based on region features | |
| CN104992433A (en) | Multispectral image registration method and device based on line segment matching | |
| CN107392948B (en) | Image registration method of amplitude-division real-time polarization imaging system | |
| CN105913065A (en) | Method and system for extracting image target area |
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 |