CN116452647A - Dynamic image registration method, system and device based on matching pursuit - Google Patents

Dynamic image registration method, system and device based on matching pursuit Download PDF

Info

Publication number
CN116452647A
CN116452647A CN202310705533.9A CN202310705533A CN116452647A CN 116452647 A CN116452647 A CN 116452647A CN 202310705533 A CN202310705533 A CN 202310705533A CN 116452647 A CN116452647 A CN 116452647A
Authority
CN
China
Prior art keywords
matching
matrix
points
image registration
representing
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
Application number
CN202310705533.9A
Other languages
Chinese (zh)
Other versions
CN116452647B (en
Inventor
朱正东
杨祖元
李陵江
谢双龙
冯浩天
谢胜利
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ante Laser Co ltd
Original Assignee
Guangdong University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN202310705533.9A priority Critical patent/CN116452647B/en
Publication of CN116452647A publication Critical patent/CN116452647A/en
Application granted granted Critical
Publication of CN116452647B publication Critical patent/CN116452647B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G06T7/33—Determination of transform parameters for the alignment of images, i.e. image registration using feature-based methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G06V10/757—Matching configurations of points or features
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • Databases & Information Systems (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Software Systems (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a dynamic image registration method, a system and a device based on matching pursuit, wherein the method comprises the following steps: reading an image and performing preliminary feature point matching; constructing an initial homography matrix; selecting a matching point with the largest residual error based on the residual error matrix, and updating the distance weights of all the matching points in the inhibition radius of the selected matching point; updating the homography matrix; returning to the step of selecting the matching points until the iteration stopping condition is met, and obtaining a final homography matrix; image registration is completed. The system comprises: the device comprises a preliminary matching module, an initial matrix module, an updating module, a judging module and a registering module. The apparatus includes a memory and a processor for performing the matching pursuit-based dynamic image registration method described above. By using the method and the device, the characteristic points of the effective description image can be selected, the homography matrix reconstruction precision is improved, and the image registration quality is further improved. The invention can be widely applied to the field of image registration.

Description

Dynamic image registration method, system and device based on matching pursuit
Technical Field
The present invention relates to the field of image registration, and in particular, to a dynamic image registration method, system and apparatus based on matching tracking.
Background
Image registration is a technique for achieving geometric calibration of images, and a mapping relation between a reference image and an image to be registered is obtained by calculating a homography matrix. In image registration, homography matrix plays a role of importance, and registration and reconstruction of images can be realized by finding corresponding feature points in a plurality of images and calculating homography matrixes corresponding to the points. However, when the feature point detection operator is used to detect an image, a large number of trivial and aggregated feature points often appear, and the homography matrix calculated according to the feature points may appear over-describing a local area of the image, thereby resulting in low registration quality.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a dynamic image registration method, a dynamic image registration system and a dynamic image registration device based on matching pursuit, which can select feature points for effectively describing images, improve homography matrix reconstruction accuracy and further improve image registration quality.
The first technical scheme adopted by the invention is as follows: a dynamic image registration method based on matching pursuit comprises the following steps:
reading an image and performing preliminary feature point matching to obtain matching points;
constructing an initial homography matrix according to the matching points;
selecting a matching point with the largest residual error based on the residual error matrix, and updating the distance weights of all the matching points in the inhibition radius of the selected matching point;
updating the homography matrix based on the selected matching points;
returning to the step of selecting the matching points until the iteration stopping condition is met, and obtaining a final homography matrix;
and (5) completing image registration based on the final homography matrix.
Further, the step of reading the image and performing preliminary feature point matching to obtain a matching point specifically includes:
acquiring an image to be processed;
performing feature point detection processing on an image to be processed to obtain feature points corresponding to the image to be processed;
and calculating the Hamming distance according to the characteristic points, removing the characteristic points which are erroneously matched, and completing the initial matching of the characteristic points to obtain matching points.
Through the preferred step, the conventional feature point detection and matching method is utilized to perform preliminary matching, and the result of the preliminary matching is processed.
Further, the method further comprises the following steps:
and assigning position weights to the matching points based on a target recognition method and performing primary screening.
Through the preferred step, position weights are given to the matching points through a deep learning target recognition technology, unstable matching points are screened out, and the fact that the rest matching points can describe images more accurately is ensured.
Further, the step of selecting the matching point with the largest residual error based on the residual error matrix and updating the distance weights of all the matching points within the inhibition radius of the selected matching point specifically includes:
introducing a residual error matrix;
calculating the residual size of the matching points based on the residual matrix and selecting the matching point with the largest residual;
defining a storage matrix and adding transposed vectors of a preset number of matching points to the storage matrix;
setting a suppression radius and updating the distance weights of all matching points in the selected matching point suppression radius.
Through the optimization step, the residual matrix is iteratively updated, the matching point with the largest residual is selected each time, the stability and the significance of the selected characteristic point are comprehensively considered, other insignificant trivial characteristic points are eliminated, and the reconstruction precision of the homography matrix is improved; and updating the distance weight for the matching points with similar characteristics in the inhibition radius of the selected matching points, and reducing the possibility that the matching points in the range are selected again in the next iteration so as to reduce redundant description of the image and be beneficial to improving the discreteness of the selected matching points.
Further, the residual matrix formula is expressed as follows:
in the above-mentioned method, the step of,Ethe residual matrix is represented and,representing the expanded form of the residual matrix,/->Representing the first row of the residual matrix,/->Representing the first of the residual matrixnGo (go)/(go)>Represents the expanded form of the position weights, +.>The first row representing the position weight,the first one of the position weightsnGo (go)/(go)>Represents the expanded form of the distance weights, +.>First row representing distance weights, +.>The first one of the distance weightsnGo (go)/(go)>Representing coefficient matrix->Is arranged in the row corresponding to the row,hrepresenting homography matrix,/->Representing multiplication of corresponding elements->Representing coordinate information of corresponding matching points in the images to be registered on the second image,ndetermined by the number of matching points.
Further, the update formula of the distance weight is as follows:
in the above-mentioned method, the step of,representing distance weight,/-, and>representing the distance weight before update, +.>A geometric average representing the distances of a pair of matching points to their respective centers to which the distance weights are assigned,a、b、krepresenting a preset constant，eRepresenting natural constants.
Further, the method further comprises the following steps:
and carrying out normalization processing on the feature points.
Through the preferred step, the coordinate normalization processing is performed on all the characteristic points, so that the calculation processing of the subsequent coordinates is facilitated.
Further, the step of completing image registration based on the final homography matrix specifically includes:
restoring the final homography matrix into a homography matrix of actual coordinates;
and performing projection mapping according to the homography matrix of the actual coordinates to finish image registration.
Through the preferred step, the points of the first image in the two images to be processed are finally mapped into the second image through the homography matrix, so that the registration of the images is completed.
The second technical scheme adopted by the invention is as follows: a matching pursuit-based dynamic image registration system, comprising:
the preliminary matching module is used for reading the image and performing preliminary feature point matching to obtain matching points;
the initial matrix module is used for constructing an initial homography matrix according to the matching points;
the updating module is used for selecting the matching point with the largest residual error based on the residual error matrix and updating the distance weights of all the matching points in the inhibition radius of the selected matching point; updating the homography matrix based on the selected matching points;
the judging module returns to the step of selecting the matching points until the iteration stop condition is met, and a final homography matrix is obtained;
and the registration module is used for completing image registration based on the final homography matrix.
The third technical scheme adopted by the invention is as follows: a dynamic image registration apparatus based on matching pursuits, comprising:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement a matching pursuit based dynamic image registration method as described above.
The method, the system and the device have the beneficial effects that: according to the invention, different position weights are given to each type of matching points through a deep learning target recognition technology, so that the matching points capable of accurately describing images are obtained; through iterative updating of the residual matrix and selection of the matching point with the largest residual, the insignificant trivial matching point can be eliminated, and the reconstruction precision of the homography matrix is improved; by updating the distance weights for all feature-similar matching points within the selected matching point suppression radius, redundant description of the image can be reduced, which is beneficial to improving the discreteness of the selected matching points. The reconstructed homography matrix can describe the mapping relation between the images more accurately so as to realize dynamic image registration better finally.
Drawings
FIG. 1 is a flow chart of steps of a dynamic image registration method based on matching pursuits of the present invention;
FIG. 2 is a graph showing distance weights in intervals according to an embodiment of the present inventionA mapped image thereon;
FIG. 3 is a schematic diagram illustrating mapping of pixel points according to an embodiment of the present invention;
fig. 4 is a block diagram of a dynamic image registration system based on matching pursuits according to the present invention.
Reference numerals: 1. a first image; 2. and a second image.
Detailed Description
The invention will now be described in further detail with reference to the drawings and to specific examples. The step numbers in the following embodiments are set for convenience of illustration only, and the order between the steps is not limited in any way, and the execution order of the steps in the embodiments may be adaptively adjusted according to the understanding of those skilled in the art.
As shown in fig. 1, the present invention provides a dynamic image registration method based on matching pursuit, which includes the following steps:
s1, reading an image and performing preliminary feature point matching to obtain matching points;
s1.1, acquiring a plurality of images to be processed;
s1.2, carrying out feature point detection processing on an image to be processed to obtain feature points corresponding to the image to be processed;
specifically, detection of image feature points is accomplished by feature point detection operators, such as SIFT operator, SURF operator, ORB operator, etc., and these feature points are described using 128-dimensional vectors as descriptors.
S1.3, calculating the Hamming distance according to the characteristic points and eliminating the wrong matching points, and completing the initial matching of the characteristic points to obtain the matching points.
Specifically, the hamming distance of descriptors between every two image feature points is calculated to complete the initial matching of the feature points, and then the RANSAC algorithm is used to exclude mismatching points, so that the initial matching of every two image feature points is finally completed.
S2, assigning position weights to the matching points based on a target recognition method and performing primary screening.
Using deep learning target recognition technology to recognize dynamic targets in images and obtain corresponding target frames to realize position weight assignment to different matching points, wherein any matching point is given：
(1)
Screening out the matching points with the position weight value of 0, and finishing the primary screening.
S3, constructing an initial homography matrix according to the matching points;
s3.1, translating the image coordinates so that the origin of the image coordinates is the center of gravity of the feature point. The average offset of the coordinates is as follows:
(2)
in the above-mentioned method, the step of,for the +.>For matching points, the image to be registered comprises a first image and a second image, +.>Indicate->For the first image of the matching pointxCoordinates of->Indicate->For the first image of the matching pointyThe coordinates of the two points of the coordinate system,indicate->For the second image of the matching pointxCoordinates of->Indicate->For the y-coordinate of the second image of the matching points, n is determined by the number of matching points, +.>X-coordinate average offset representing the center of gravity of all matching points of the first image, +.>Representing the center of gravity of all matching points of the first imageyAverage offset of coordinates>Representing the centre of gravity of all matching points of the second imagexThe average amount of offset of the coordinates,representing the centre of gravity of all matching points of the second imageyAverage offset of coordinates.
And S3.2, scaling the image coordinates to make the average distance from the feature point to the original point be 1. The coordinate scaling scale is as follows:
(3)
in the above-mentioned method, the step of,representing all matching points of the first imagexScale of coordinates>Representing all matching points of the first imageyScale of coordinates>Representing all matching points of the second imagexScale of coordinates>Representing all matching points of the second imageyScaling of the coordinates.
S3.3, constructing a corresponding coordinate transformation matrix, and realizing normalization processing of all feature point coordinates:
(4)
in the above-mentioned method, the step of,the coordinates of the normalized pair of feature points are the normalized matching points.
S3.4, solving a standard linear equation set of the homography matrix as follows:
(5)
in the above, homography matrix，/>For the +.>For the matching point of the matching point,。
s3.5, due to the scale equivalence, letSingle sheetThe stress matrix is fixed in scale. The equivalent form of the above equation set is as follows:
(6)
in the above, a group of matching points contributes two rows, so the coefficient matrixIs->In the form of (a) and (b),Mrepresenting a target matrix for storing a coefficient matrix +.>Target values corresponding to each row of coefficient vectors; because the first image is mapped to the second image, the target matrixMThe actual meaning of (2) is: for storing the coordinate values of the pair of matching points on the second image. Coefficient matrixEach row of coefficient vectors in (a) is multiplied byhEqual to the target value, each constraint equation obtained is used to construct a least squares problem solutionhIs a parameter of 8 unknown parameters.
S4, selecting a matching point with the largest residual error based on the residual error matrix, and updating the distance weights of all matching points in the inhibition radius of the selected matching point;
s4.1, introducing a residual error matrix for searching a homography matrixAnd a relatively stable matching point.
The residual matrix is defined as follows:(7)
in the above-mentioned method, the step of,Ethe residual matrix is represented and,representing residual matricesECorresponding row of->Corresponding row representing position weights +.>Corresponding row representing distance weights +.>Representing coefficient matrix->Corresponding row of->Representing multiplication of corresponding elements->Representing coordinate information of corresponding matching points in the images to be registered on the second image,nas determined by the number of matching points,ndetermined by the number of matching points.
S4.2, defining a memory matrixFThe initial value is a null set for storing the selected matching points in the coefficient matrixATranspose of the corresponding two coefficient vectors. A pair of matching points in coefficient matrixATwo rows are contributed to, so that a pair of matching points stores the matrixFTo contribute two transposed coefficient vectors, such as: (8)
in the above-mentioned method, the step of, transpose of the 1 st coefficient vector corresponding to the matching point,/-> Transpose of the 2 nd coefficient vector corresponding to the matching point, wherein +.>The representation is rounded up and down to the top,krepresenting a preset constant, typically 10.
S4.3, randomly selecting four groups of matching points with position weights of 1 to be substituted into the (6) to complete the homography matrixIn (a)And adds 8 transpose vectors of 4 sets of matching points to the memory matrixFAmong them.
S4.4, setting the inhibition radius for each selected matching pointAnd update the inhibition radius +.>Distance weights for all matching points within. All matching point initial distance weights are set to 1, which is measured by the following expression:
(9)
in the above-mentioned method, the step of,representing the distance weight before update, i.e. the current distance weight, +.>A geometric average representing the distances of a pair of matching points to their respective centers to which the distance weights are assigned,a、b、kthe predetermined constant is indicated to be a predetermined constant,erepresenting natural constants.
In general, let's sayRealize the distance weight in the interval +.>The mapping on, the functional image is referred to in fig. 2.
And S4.4, calculating the residual error size of the matching point.
Specifically, the maximum correlation is obtained, and the homography matrix is obtainedSubstituting the residual calculated in the formula (7) to complete updating of the residual matrix.
S4.5, the residual error matrix is the firstTwo transposed vectors of the matching point corresponding to the largest residual error +.>Added to a memory matrixFAnd updating the distance weight according to the formula (9).
S5, updating the homography matrix based on the selected matching points;
transpose of the corresponding two coefficient vectors of the matching point selected by the iteration into a storage matrixFThen according to the updated memory matrixFUpdating homography matrix. The process corresponds to solving a least squares problem:
(10)
in the above-mentioned method, the step of,to transpose the symbol, in this case->Comprises onlyFTarget value corresponding to existing vector in +.>Representing the L2 norm of the vector.
S6, returning to the step S4 until the iteration stop condition is met, and jumping out of the loop to obtain a final homography matrix;
specifically, steps S4.4-S5 are looped until the norm of the residual matrix is less than a preset threshold or the memory matrixFIf the number of middle column vectors exceeds a preset threshold, the iterative process stops, indicating that a sufficiently sparse representation has been found for reconstructing the homography matrix.
Memory matrixFThe number of middle column vectors is equal to the number of matching points.
And S7, completing image registration based on the final homography matrix.
S7.1, reducing the reconstructed final homography matrix into a final homography matrix in a general form:
(11)
s7.2, the final homography matrix of the general form is also needed as the characteristic point coordinates are normalized beforeRestoring to final homography matrix under actual coordinates>：/>(12)
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing a first recovery matrix->Representing a second recovery matrix->And->The expression of (2) is as follows:
(13)
and S7.3, performing projection mapping according to the homography matrix of the actual coordinates to complete image registration.
When the camera rotates around a point to obtain two different images (as shown in fig. 3), the two images are projection mapping, and the relationship between the two images can be obtained by using the final homography matrix under the actual coordinatesTo describe. In the figure->Is one point in the first image, then mapped to another point in the second image +.>. The points in the first image finally pass the final homography matrix in real coordinates +.>Mapping into the second image, thereby completing registration of the images.
As shown in fig. 4, a dynamic image registration system based on matching pursuits includes:
the preliminary matching module is used for reading the image and performing preliminary feature point matching to obtain matching points;
the initial matrix module is used for constructing an initial homography matrix according to the matching points;
the updating module is used for selecting the matching point with the largest residual error based on the residual error matrix and updating the distance weights of all the matching points in the inhibition radius of the selected matching point; updating the homography matrix based on the selected matching points;
the judging module returns to the step of selecting the matching points until the iteration stop condition is met, and a final homography matrix is obtained;
and the registration module is used for completing image registration based on the final homography matrix.
The content in the method embodiment is applicable to the system embodiment, the functions specifically realized by the system embodiment are the same as those of the method embodiment, and the achieved beneficial effects are the same as those of the method embodiment.
A dynamic image registration device based on matching pursuit:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement a matching pursuit based dynamic image registration method as described above.
The content in the method embodiment is applicable to the embodiment of the device, and the functions specifically realized by the embodiment of the device are the same as those of the method embodiment, and the obtained beneficial effects are the same as those of the method embodiment.
A storage medium having stored therein processor-executable instructions which, when executed by a processor, are for implementing a matching pursuit based dynamic image registration method as described above.
The content in the method embodiment is applicable to the storage medium embodiment, and functions specifically implemented by the storage medium embodiment are the same as those of the method embodiment, and the achieved beneficial effects are the same as those of the method embodiment.
While the preferred embodiment of the present invention has been described in detail, the invention is not limited to the embodiment, and various equivalent modifications and substitutions can be made by those skilled in the art without departing from the spirit of the invention, and these modifications and substitutions are intended to be included in the scope of the present invention as defined in the appended claims.

Claims (10)

1. The dynamic image registration method based on the matching pursuit is characterized by comprising the following steps:
reading an image and performing preliminary feature point matching to obtain matching points;
constructing an initial homography matrix according to the matching points;
selecting a matching point with the largest residual error based on the residual error matrix, and updating the distance weights of all the matching points in the inhibition radius of the selected matching point;
updating the homography matrix based on the selected matching points;
returning to the step of selecting the matching points until the iteration stopping condition is met, and obtaining a final homography matrix;
and (5) completing image registration based on the final homography matrix.
2. The method for matching pursuit-based dynamic image registration according to claim 1, wherein the step of reading the image and performing preliminary feature point matching to obtain matching points specifically comprises:
acquiring an image to be processed;
performing feature point detection processing on an image to be processed to obtain feature points corresponding to the image to be processed;
and calculating the Hamming distance according to the characteristic points, removing the characteristic points which are erroneously matched, and completing the initial matching of the characteristic points to obtain matching points.
3. The matching pursuit-based dynamic image registration method of claim 2, further comprising:
and assigning position weights to the matching points based on a target recognition method and performing primary screening.
4. A matching pursuit-based dynamic image registration method according to claim 3, wherein the step of selecting a matching point with the largest residual error based on a residual error matrix and updating the distance weights of all matching points within the selected matching point inhibition radius specifically comprises:
introducing a residual error matrix;
calculating the residual size of the matching points based on the residual matrix and selecting the matching point with the largest residual;
defining a storage matrix and adding transposed vectors of a preset number of matching points to the storage matrix;
setting a suppression radius and updating the distance weights of all matching points in the selected matching point suppression radius.
5. The matching pursuit-based dynamic image registration method of claim 4, wherein the residual matrix formula is expressed as follows:in the above-mentioned method, the step of,Erepresenting residual matrix->Representing the expanded form of the residual matrix,/->Representing the first row of the residual matrix,/->Representing the first of the residual matrixnThe number of rows of the device is,represents the expanded form of the position weights, +.>First row representing position weights, +.>The first one of the position weightsnThe number of rows of the device is,represents the expanded form of the distance weights, +.>First row representing distance weights, +.>The first one of the distance weightsnThe number of rows of the device is,representing coefficient matrix->Is arranged in the row corresponding to the row,hrepresenting homography matrix,/->Representing the multiplication of the corresponding elements,representing coordinate information of corresponding matching points in the images to be registered on the second image,ndetermined by the number of matching points.
6. The matching pursuit-based dynamic image registration method according to claim 4, wherein the distance weight update formula is as follows:in the above, the->The distance weight is represented by a value of the distance,representing the distance weight before update, +.>A geometric average representing the distances of a pair of matching points to their respective centers to which the distance weights are assigned,a、b、krepresenting a preset constant，eRepresenting natural constants.
7. The matching pursuit-based dynamic image registration method of claim 2, further comprising:
and carrying out normalization processing on the feature points.
8. The matching pursuit-based dynamic image registration method as defined in claim 7, wherein the step of performing image registration based on a final homography matrix comprises:
restoring the final homography matrix into a homography matrix of actual coordinates;
and performing projection mapping according to the homography matrix of the actual coordinates to finish image registration.
9. A matching pursuit-based dynamic image registration system, comprising:
the preliminary matching module is used for reading the image and performing preliminary feature point matching to obtain matching points;
the initial matrix module is used for constructing an initial homography matrix according to the matching points;
the updating module is used for selecting the matching point with the largest residual error based on the residual error matrix and updating the distance weights of all the matching points in the inhibition radius of the selected matching point; updating the homography matrix based on the selected matching points;
the judging module returns to the step of selecting the matching points until the iteration stop condition is met, and a final homography matrix is obtained;
and the registration module is used for completing image registration based on the final homography matrix.
10. A dynamic image registration apparatus based on matching pursuits, comprising:
at least one processor;
at least one memory for storing at least one program;
the at least one program, when executed by the at least one processor, causes the at least one processor to implement a matching pursuit-based dynamic image registration method as recited in any one of claims 1-8.
CN202310705533.9A 2023-06-15 2023-06-15 Dynamic image registration method, system and device based on matching pursuit Active CN116452647B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310705533.9A CN116452647B (en) 2023-06-15 2023-06-15 Dynamic image registration method, system and device based on matching pursuit

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310705533.9A CN116452647B (en) 2023-06-15 2023-06-15 Dynamic image registration method, system and device based on matching pursuit

Publications (2)

Publication Number Publication Date
CN116452647A true CN116452647A (en) 2023-07-18
CN116452647B CN116452647B (en) 2023-12-08

Family

ID=87130546

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310705533.9A Active CN116452647B (en) 2023-06-15 2023-06-15 Dynamic image registration method, system and device based on matching pursuit

Country Status (1)

Country Link
CN (1) CN116452647B (en)

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104751465A (en) * 2015-03-31 2015-07-01 中国科学技术大学 ORB (oriented brief) image feature registration method based on LK (Lucas-Kanade) optical flow constraint
CN105957007A (en) * 2016-05-05 2016-09-21 电子科技大学 Image stitching method based on characteristic point plane similarity
CN107170001A (en) * 2017-04-25 2017-09-15 北京海致网聚信息技术有限公司 Method and apparatus for carrying out registration to image
CN107220999A (en) * 2017-06-19 2017-09-29 江南大学 The research of workpiece circular arc Edge Feature Points matching process
CN112435278A (en) * 2021-01-26 2021-03-02 华东交通大学 Visual SLAM method and device based on dynamic target detection
KR102247288B1 (en) * 2019-11-29 2021-05-03 울산과학기술원 Warping residual based image stitching for large parallax
CN114677420A (en) * 2022-03-04 2022-06-28 武汉理工大学 Improved LLT-GST image registration algorithm
CN114998773A (en) * 2022-08-08 2022-09-02 四川腾盾科技有限公司 Characteristic mismatching elimination method and system suitable for aerial image of unmanned aerial vehicle system
WO2023045420A1 (en) * 2021-09-27 2023-03-30 上海哔哩哔哩科技有限公司 Image processing method and apparatus, electronic device, and storage medium

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104751465A (en) * 2015-03-31 2015-07-01 中国科学技术大学 ORB (oriented brief) image feature registration method based on LK (Lucas-Kanade) optical flow constraint
CN105957007A (en) * 2016-05-05 2016-09-21 电子科技大学 Image stitching method based on characteristic point plane similarity
CN107170001A (en) * 2017-04-25 2017-09-15 北京海致网聚信息技术有限公司 Method and apparatus for carrying out registration to image
CN107220999A (en) * 2017-06-19 2017-09-29 江南大学 The research of workpiece circular arc Edge Feature Points matching process
KR102247288B1 (en) * 2019-11-29 2021-05-03 울산과학기술원 Warping residual based image stitching for large parallax
CN112435278A (en) * 2021-01-26 2021-03-02 华东交通大学 Visual SLAM method and device based on dynamic target detection
WO2023045420A1 (en) * 2021-09-27 2023-03-30 上海哔哩哔哩科技有限公司 Image processing method and apparatus, electronic device, and storage medium
CN114677420A (en) * 2022-03-04 2022-06-28 武汉理工大学 Improved LLT-GST image registration algorithm
CN114998773A (en) * 2022-08-08 2022-09-02 四川腾盾科技有限公司 Characteristic mismatching elimination method and system suitable for aerial image of unmanned aerial vehicle system

Also Published As

Publication number Publication date
CN116452647B (en) 2023-12-08

Similar Documents

Publication Publication Date Title
WO2022052367A1 (en) Neural network optimization method for remote sensing image classification, and terminal and storage medium
CN111340109B (en) Image matching method, device, equipment and storage medium
CN112328715B (en) Visual positioning method, training method of related model, related device and equipment
CN111461113B (en) Large-angle license plate detection method based on deformed plane object detection network
CN109919971B (en) Image processing method, image processing device, electronic equipment and computer readable storage medium
US11822900B2 (en) Filter processing device and method of performing convolution operation at filter processing device
CN112053383B (en) Method and device for positioning robot in real time
CN112102381A (en) Hardware Trojan horse image registration method based on R-SIFT, storage medium and equipment
CN114155285B (en) Image registration method based on gray histogram
CN114549861A (en) Target matching method based on feature point and convolution optimization calculation and storage medium
CN116452647B (en) Dynamic image registration method, system and device based on matching pursuit
CN116597246A (en) Model training method, target detection method, electronic device and storage medium
CN116258873A (en) Position information determining method, training method and device of object recognition model
CN113808028B (en) Method and device for detecting countermeasure sample based on attribution algorithm
CN114998755A (en) Method and device for matching landmarks in remote sensing image
US20240153154A1 (en) Coordinate generation system, coordinate generation method, and computer readable recording medium with stored program
CN113569973B (en) Multi-view clustering method, device, electronic equipment and computer readable storage medium
US20230384235A1 (en) Method for detecting product for defects, electronic device, and storage medium
US20230401670A1 (en) Multi-scale autoencoder generation method, electronic device and readable storage medium
CN115083001B (en) Anti-patch generation method and device based on image sensitive position positioning
CN113505838B (en) Image clustering method and device, electronic equipment and storage medium
CN112633304B (en) Robust fuzzy image matching method
CN117671273A (en) Synthetic aperture radar image anti-noise identification method and system based on local constraint network
CN118052928A (en) Coordinate generation system and coordinate generation method
CN117197712A (en) ORB feature matching algorithm optimization method and device and computer equipment

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
TA01 Transfer of patent application right
TA01 Transfer of patent application right

Effective date of registration: 20231109

Address after: F201, No. 11 Caibin Road, Science City, Guangzhou Economic and Technological Development Zone, Guangdong Province, 510663

Applicant after: ANTE LASER Co.,Ltd.

Address before: 510006 Guangdong University of technology, Xiaoguwei street, Panyu District, Guangzhou City, Guangdong Province

Applicant before: GUANGDONG University OF TECHNOLOGY

GR01 Patent grant
GR01 Patent grant