CN111882594A - ORB feature point-based polarization image rapid registration method and device - Google Patents
ORB feature point-based polarization image rapid registration method and device Download PDFInfo
- Publication number
- CN111882594A CN111882594A CN202010730007.4A CN202010730007A CN111882594A CN 111882594 A CN111882594 A CN 111882594A CN 202010730007 A CN202010730007 A CN 202010730007A CN 111882594 A CN111882594 A CN 111882594A
- Authority
- CN
- China
- Prior art keywords
- image
- registered
- reference image
- feature
- characteristic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- 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
Abstract
The invention relates to a polarization image fast registration method and a device based on ORB characteristic points, which comprises the steps of obtaining at least two intensity images of a target in different polarization directions, and determining a reference image and an image to be registered; respectively extracting feature points in the reference image and the image to be registered by utilizing an ORB algorithm; constructing a search tree of the characteristic points of the reference image by using a rapid nearest neighbor search algorithm, and constructing characteristic point pairs matched between the image to be registered and the reference image according to the search tree; and estimating a projective transformation matrix from the image to be registered to the reference image according to the characteristic point pairs, and performing projective transformation on the image to be registered to obtain an image registration result. The method has good noise immunity, can realize the rapid registration of the polarization images, and effectively solves the problem of mismatching of the polarization images of targets, particularly dynamic targets, acquired by a time-sharing polarization imaging system at different moments.
Description
Technical Field
The invention relates to the technical field of polarization imaging, in particular to a polarization image fast registration method and device based on ORB feature points, electronic equipment and a computer readable storage medium.
Background
In recent years, the polarization imaging technology is widely applied to various fields by virtue of the advantages of highlighting targets, penetrating cloud mist, distinguishing authenticity and the like. On the basis of intensity detection, the technology can enhance the edge contour characteristics of the target and improve the target detection and identification capability in the low signal-to-noise ratio environment by utilizing the polarization characteristic difference of the target and the background radiation light wave. Meanwhile, different types of polarization imaging instruments are developed and used in the fields of material classification, target recognition, atmospheric detection, biological diagnosis and the like. The time-sharing polarization imaging system collects intensity images in different polarization directions at different moments by adopting a mode of rotating a polarizing film, calculates a plurality of images to obtain infrared polarization information of a target, has the advantages of low cost, high extinction ratio, simple and convenient structure and the like, and is well applied to the field of aerospace. In order to realize real-time polarization imaging detection in a changing scene, a time-sharing high-frame-frequency polarization imaging system needs to be developed, the speed of the output polarization image reaches 50 frames/second, and the polarization information of a target scene can be rapidly acquired, however, in the process of polarization imaging of a dynamic target, intensity images in different polarization directions at different moments have certain position difference. If the polarization information is characterized based on these images with position difference, false polarization characteristics about the target will be generated. Therefore, the obtained intensity images with different polarization directions must be quickly and accurately registered, so that real-time and accurate polarization information about the target is calculated, and the capability of detecting and identifying the target in real time is improved.
The existing polarization image registration methods mainly include an infrared polarization image partition registration method based on matrix recovery, a polarization image registration algorithm based on KAZE characteristics, an infrared polarization image registration algorithm based on phase correlation and sub-images, and the like. However, these registration methods have the disadvantage of long registration time or sensitivity to noise, and the algorithms are usually designed for specific polarization systems and scenes. For a time-sharing high frame rate polarization imaging system, a corresponding registration method is particularly required to have the characteristic of real-time and high speed. Obviously, the existing polarized image registration methods cannot meet these requirements.
Therefore, in view of the above disadvantages, it is desirable to provide a new fast registration method for polarized images with high noise immunity, especially with real-time characteristics.
Disclosure of Invention
The invention aims to solve the technical problems that the existing polarized image registration method has the defects of long registration time and sensitivity to noise, cannot meet the requirement of real-time registration of the polarized image of a time-sharing high frame frequency polarized imaging system, and provides a polarized image rapid registration method based on ORB (ordered FAST and Rotated BRIEF) characteristic points aiming at the defects in the prior art.
In order to solve the above technical problem, the present invention provides a method for fast registering a polarization image based on ORB feature points, which includes the following steps:
acquiring at least two intensity images of a target in different polarization directions, and determining a reference image and an image to be registered;
respectively extracting feature points in the reference image and the image to be registered by utilizing an ORB algorithm;
constructing a search tree of the characteristic points of the reference image by using a rapid nearest neighbor search algorithm, and constructing the matched characteristic point pairs in the image to be registered and the reference image according to the search tree;
and estimating a projective transformation matrix from the image to be registered to the reference image according to the characteristic point pairs, and performing projective transformation on the image to be registered to obtain an image registration result.
Preferably, the feature points are duplets composed of coordinates and corresponding feature descriptors.
Preferably, the constructing the feature point pairs matched between the image to be registered and the reference image according to the search tree specifically includes:
and taking the Hamming distance between the feature descriptors as a similarity measurement criterion of the feature points, searching the corresponding feature point with the minimum Hamming distance for the feature point of the image to be registered in the constructed search tree of the feature points of the reference image, and constructing the matched feature point pair in the image to be registered and the reference image.
Preferably, the constructing the feature point pairs matched between the image to be registered and the reference image further includes:
and filtering out the characteristic point pairs with the Hamming distance between the characteristic points larger than a preset threshold value from the characteristic point pairs.
Preferably, the estimating, according to the feature point pairs, a projective transformation matrix from the image to be registered to the reference image specifically includes:
extracting a coordinate pair in the feature point pairs;
estimating a projective transformation matrix of the image to be registered to the reference image according to the following formula:
wherein the content of the first and second substances,
i, j ∈ 1,2,3, which is a projective transformation matrix of the image to be registered to the reference image,
coordinates representing the kth feature point in the reference image,
and representing the coordinates of the t-th characteristic point in the m-th image to be registered.
The invention also provides a polarization image fast registration device based on ORB characteristic points, which comprises:
the image acquisition unit is used for acquiring at least two intensity images of the target in different polarization directions and determining a reference image and an image to be registered;
the characteristic extraction unit is used for respectively extracting characteristic points in the reference image and the image to be registered by utilizing an ORB algorithm;
the characteristic matching unit is used for constructing a search tree of the characteristic points of the reference image by utilizing a rapid nearest neighbor search algorithm and constructing the characteristic point pairs matched with the reference image in the image to be registered according to the search tree;
and the image registration unit is used for estimating a projection transformation matrix from the image to be registered to the reference image according to the characteristic point pairs, and performing projection transformation on the image to be registered to obtain an image registration result.
Preferably, the feature points are duplets composed of coordinates and corresponding feature descriptors.
Preferably, the feature matching unit is configured to perform the following operations:
using the Hamming distance between the feature descriptors as a similarity measurement criterion of the feature points, searching the corresponding feature point with the minimum Hamming distance for the feature point of the image to be registered in the constructed search tree of the feature points of the reference image, and constructing the matched feature point pair in the image to be registered and the reference image;
preferably, the feature matching unit is further configured to perform the following operations:
and filtering out the characteristic point pairs with the Hamming distance between the characteristic points larger than a preset threshold value from the characteristic point pairs.
Preferably, the image registration unit is configured to perform the following operations:
extracting a coordinate pair in the feature point pairs;
estimating a projective transformation matrix of the image to be registered to the reference image according to the following formula:
wherein the content of the first and second substances,
i, j e 1,2,3, for the image to be registered to the reference mapA projective transformation matrix of the image,
coordinates representing the kth feature point in the reference image,
and representing the coordinates of the t-th characteristic point in the m-th image to be registered.
The invention also provides an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the polarization image fast registration method according to any one of the preceding claims when executing the computer program.
The invention also provides a computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of the polarization image fast registration method of any one of the preceding claims.
The polarized image rapid registration method and device based on the ORB characteristic points, the electronic equipment and the computer readable storage medium have the following beneficial effects that: the method comprises the steps of obtaining a plurality of target polarization images through a polarization detection system, determining a reference image and an image to be registered, rapidly extracting feature points in the reference image and the image to be registered by utilizing an ORB algorithm, then constructing a search tree of the feature points of the reference image by utilizing a rapid nearest neighbor search algorithm, matching the feature points of the image to be registered and the reference image, estimating a projection transformation matrix from the image to be registered to the reference image based on the matched feature points, performing projection transformation on the image to be registered, and registering the image to be registered and the reference image. The method has good noise immunity, can realize the rapid registration of the polarization image, and effectively solves the problem of mismatching of the polarization image of the target, especially the dynamic target, acquired by the time-sharing polarization imaging system at different moments, thereby laying a technical foundation for the time-sharing polarization imaging system to acquire real polarization information of the target in real time and simultaneously providing a new solution for the registration of the polarization image in the polarization imaging field.
Drawings
Fig. 1 is a flowchart of a method for fast registration of polarization images based on ORB feature points according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a polarized image fast registration apparatus based on ORB feature points according to a second embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present invention.
Example one
As shown in fig. 1, the method for rapidly registering polarization images based on ORB feature points according to an embodiment of the present invention includes the following steps:
first, in step S1, at least two intensity images of the target in different polarization directions are obtained, and a reference image and an image to be registered are determined.
Specifically, a plurality of intensity images of different polarization directions in a target scene are continuously acquired by using a polarization detection system, wherein M +1 intensity images { I }are acquired in total(0),I(1),I(2),…,I(M)H × W for each intensity image, wherein H, W are the height and width of the image, respectively. In the multiple intensity images, one of the intensity images can be arbitrarily selected as a reference image, and the rest of the intensity images are used as images to be registered, optionally, the intensity image I (c) is selected0) As reference image, the remaining M intensity images { I }(m)M1, 2.. M, as the image to be registered. For example, 3 intensity images { I) with polarization directions of 0 degrees, 60 degrees and 120 degrees respectively in a target scene are continuously acquired by using a time-sharing high frame frequency polarization detection system(0),I(1),I(2)The size of the 3 intensity images is 640 multiplied by 512; selecting an intensity image I (0) That is, the image with the polarization angle of 0 ° is used as the reference image, and the remaining 2 intensity images are used as the images to be registered.
Subsequently, in step S2, feature points in the reference image and the image to be registered are extracted by the ORB algorithm, respectively.
The ORB is a very good method capable of extracting and describing image feature points in real time, and the matching accuracy based on the ORB feature points is comparable to the traditional widely-used SIFT (scale imperial feature transform) method, but the registration speed is nearly 10 times of SIFT. Meanwhile, the ORB features have invariance to target rotation, scale scaling, brightness change and the like, so that the ORB features have strong robustness or anti-noise characteristics.
In the invention, the ORB algorithm is adopted to respectively extract the reference images I(0)Characteristic point of
k＝1,2,…,N0，N0Is the number of feature points of the reference image, wherein
Representing the kth feature point of the reference image. And extracting the image to be registered { I(1),I(2),…,I(M)Feature points of
i＝1,2,…,Nm，NmAnd the number of the characteristic points of the mth image to be registered is obtained. Wherein the content of the first and second substances,
and the ith characteristic point of the mth image to be registered is shown. In some preferred embodiments, the feature points are duplets of coordinates and corresponding feature descriptors, e.g. to
Is shown in which
Coordinates of the ith characteristic point of the mth image to be registered are represented by
It is shown that,
is a corresponding feature descriptor, which is a binary vector
Wherein
N is 1, … N, N is the dimension of the vector. By adopting the ORB method, the effective characteristics of the polarization image can be extracted rapidly and robustly, so that the noise immunity of the method can be improved, and the high-precision registration of the polarization image can be realized.
In step S3, a fast nearest neighbor search algorithm is used to construct a search tree of the feature points of the reference image, and feature point pairs matching the reference image and the image to be registered are constructed according to the search tree.
In the invention, a rapid Nearest neighbor searching algorithm of high-dimensional binary data based on FLANN (fast Library for Approximate neighbors) is adopted to construct a reference image I(0)Middle characteristic point
k＝1,2,…,N0To realize fast search of feature points.
In some preferred embodiments, the hamming distance between feature descriptors is used as the similarity measurement criterion of the feature points, and the feature points of the image to be registered are used in the constructed search tree of the feature points of the reference image
Searching the most similar characteristic point, namely the characteristic point with the minimum Hamming distance, thereby constructing an image to be registered and the imageMatching pairs of feature points in a reference image
j＝1,2,…,NmWherein
Representing images to be registered I(m)J th characteristic point and reference image I(0)The kth feature point is matched, k ∈ {1,2, …, N0}. The Hamming distance between the two feature descriptors is calculated as follows, namely, the corresponding positions of the two feature descriptors are subjected to exclusive OR operation one by one and then summed.
Compared with a Local Sensitive Hash (LSH) algorithm with fixed parameters widely adopted in the prior art, the FLANN algorithm adopted by the invention is a self-adaptive or data-driven fast search algorithm for high-dimensional binary data, and can dynamically adjust a construction method and a fast search strategy of a search tree according to the number of characteristic points in an image and the characteristic dimensions thereof, thereby realizing faster search. In addition, the data driving characteristics also enable the method to be dynamically adaptive to complex and changeable polarized image matching scenes (for example, the number of characteristic points in the polarized images of different scenes is greatly different), and the robustness and the adaptability of the polarized image rapid matching method are enhanced.
It should be noted that, when there are at least two images to be registered, the most similar feature points can be searched for the feature points of all the images to be registered in parallel according to the constructed search tree of the feature points of the reference image, so that the matching between the feature point pairs of all the images to be registered and the reference image can be realized quickly.
In some more preferred embodiments, constructing pairs of feature points that match in the image to be registered and the reference image further includes: and filtering the characteristic point pairs with the Hamming distance between the characteristic points larger than a preset threshold value in the constructed characteristic point pairs. In the present invention, the preset threshold may be twice the minimum hamming distance, where the minimum hamming distance refers to the minimum hamming distance between the feature point pairs constructed as described above. Order to
For filtering and treating I(m)The rest characteristic point re-labeled characteristic point pairs, wherein T is 1,2, … and Tm,Tm≤Nm. By filtering the mismatched characteristic point pairs, the registration accuracy of the polarization image can be improved.
And finally, in step S4, estimating a projective transformation matrix from the image to be registered to the reference image according to the feature point pairs obtained in step S3, performing corresponding projective transformation on the image to be registered, and registering the image to be registered with the reference image to obtain an image registration result.
In some preferred embodiments, a projective transformation matrix of the image to be registered to the reference image is estimated according to the feature point pairs, specifically as follows:
firstly extracting the coordinate pair of the corresponding characteristic point pair
Re-estimating the image I to be registered(m)To reference picture I(0)Projection transformation matrix H(m)I.e. solve the following optimization problem:
wherein the content of the first and second substances,
i, j ∈ 1,2,3, which is a projective transformation matrix of the image to be registered to the reference image,
coordinates representing the kth feature point in the reference image,
and representing the coordinates of the t-th characteristic point in the m-th image to be registered, namely matching the t-th characteristic point in the m-th image to be registered with the k-th characteristic point of the reference image.
And after solving to obtain a projective transformation matrix from the image to be registered to the reference image, performing corresponding projective transformation on all the images to be registered, and registering the images to be registered with the reference image to obtain an image registration result.
The invention adopts an ORB method, can rapidly and robustly extract effective characteristics of the polarization image, can realize registration of the polarization image by combining a rapid characteristic matching strategy, has good noise immunity, can effectively improve the registration speed of the polarization image, particularly meets the requirement of real-time registration of the polarization image of a time-sharing high frame frequency polarization imaging system, solves the problem of false polarization information caused by mismatching of the acquired image of the time-sharing high frame frequency polarization imaging system, and realizes high-precision registration of the polarization image.
Example two
As shown in fig. 2, the polarized image fast registration apparatus based on ORB feature points according to the second embodiment includes an image obtaining unit 100, a feature extracting unit 200, a feature matching unit 300, and an image registration unit 400.
The image acquiring unit 100 is configured to acquire at least two intensity images of the target in different polarization directions, and determine a reference image and an image to be registered.
In specific implementation, a polarization detection system can be used for continuously acquiring a plurality of intensity images in different polarization directions in a target scene, wherein M +1 intensity images { I }are acquired(0),I(1),I(2),…,I(M)H × W for each intensity image, wherein H, W are the height and width of the image, respectively. Selecting one of the obtained intensity images as a reference image, and the rest of the intensity images as images to be registered, optionally selecting an intensity image I(0)As reference image, the remaining M intensity images { I }(m)M is 1,2(1),I(2)…, I (M) } as the image to be registered. For example, 3 intensity images { I) with polarization directions of 0 degrees, 60 degrees and 120 degrees respectively in a target scene are continuously acquired by using a time-sharing high frame frequency polarization detection system(0),I(1),I(2)The size of the 3 intensity images is 640 multiplied by 512; selecting an intensity image I (0) I.e. with a polarisation angle of 0 DEGThe image is used as a reference image, and the remaining 2 intensity images are used as images to be registered.
A feature extraction unit 200, configured to extract feature points in the reference image and the image to be registered respectively by using an ORB algorithm.
Due to the ORB characteristics, the method has invariance to target rotation, scale scaling, brightness change and the like, and therefore has strong robustness or anti-noise characteristics. In the invention, the ORB algorithm is adopted to respectively extract the reference images I(0)Characteristic point of
k＝1,2,…,N0，N0Is the number of feature points of the reference image, wherein
Representing the kth feature point of the reference image. And extracting the image to be registered { I(1),I(2),…,I(M)Feature points of
i＝1,2,…,Nm，NmAnd the number of the characteristic points of the mth image to be registered is obtained. Wherein the content of the first and second substances,
and the ith characteristic point of the mth image to be registered is shown. In some preferred embodiments, the feature points are duplets of coordinates and corresponding feature descriptors, e.g. to
Is shown in which
Coordinates of the ith characteristic point of the mth image to be registered are represented by
It is shown that,
is a corresponding feature descriptor, which is a binary vector
Wherein
N is 1, … N, N is the dimension of the vector.
The feature matching unit 300 is configured to construct a search tree of feature points of the reference image by using a fast nearest neighbor search algorithm, and construct a feature point pair matching the image to be registered and the reference image according to the search tree.
In the invention, a fast nearest neighbor searching algorithm of high-dimensional binary data based on FLANN is adopted to construct a reference image I(0)Characteristic point
k＝1,2,…,N0To realize fast search of feature points.
In some preferred embodiments, the hamming distance between feature descriptors is used as the similarity measurement criterion of feature points, and in the constructed search tree of feature points of the reference image, all images to be registered { I } are used(1),I(2),…,I(M)Feature points of
Searching the most similar characteristic points, namely the characteristic points with the minimum Hamming distance, thereby constructing the matched characteristic point pairs in the image to be registered and the reference image
j＝1,2,…,NmWherein
Representing images to be registered I(m)J th characteristic point and reference image I(0)The kth feature point is matched, k ∈ {1,2, …, N0}. It should be noted that, when there are at least two images to be registered, the most similar feature points can be searched for the feature points of all the images to be registered in parallel according to the constructed search tree of the feature points of the reference image, so that the matching between the feature point pairs of all the images to be registered and the reference image can be realized quickly.
In some more preferred embodiments, constructing pairs of feature points that match in the image to be registered and the reference image further includes: and filtering the characteristic point pairs with the Hamming distance between the characteristic points larger than a preset threshold value in the constructed characteristic point pairs. In the present invention, the preset threshold may be twice the minimum hamming distance, where the minimum hamming distance refers to. Order to
For filtering and treating I(m)The rest characteristic point re-labeled characteristic point pairs, wherein T is 1,2, … and Tm,Tm≤Nm. By filtering the mismatched characteristic point pairs, the registration accuracy of the polarization image can be improved.
And the image registration unit 400 is configured to estimate a projective transformation matrix from the image to be registered to the reference image according to the feature point pairs, and perform projective transformation on the image to be registered to obtain an image registration result.
In some preferred embodiments, the image registration unit 400 is configured to, when estimating a projective transformation matrix of the image to be registered to the reference image according to the feature point pairs, specifically perform the following operations:
firstly extracting the coordinate pair of the corresponding characteristic point pair
Re-estimating the image I to be registered(m)To reference picture I(0)Projection transformation matrix H(m)I.e. solve the following optimization problem:
wherein the content of the first and second substances,
i, j ∈ 1,2,3, which is a projective transformation matrix of the image to be registered to the reference image,
coordinates representing the kth feature point in the reference image,
and representing the coordinates of the t-th characteristic point in the m-th image to be registered.
And after solving to obtain a projective transformation matrix from the image to be registered to the reference image, performing corresponding projective transformation on all the images to be registered, and registering the images to be registered with the reference image to obtain an image registration result.
In addition, an embodiment of the present invention further provides an electronic device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor, when executing the computer program, implements the steps of the polarization image fast registration method according to any one of the foregoing embodiments.
The present invention further provides a computer-readable storage medium storing a computer program, which when executed by a processor implements the steps of the polarization image fast registration method according to any of the foregoing embodiments.
In summary, the present invention obtains a plurality of target polarization images through a polarization detection system, determines a reference image and an image to be registered, rapidly extracts feature points in the reference image and the image to be registered by using an ORB algorithm, then constructs a search tree of the feature points of the reference image by using a rapid nearest neighbor search algorithm, then matches the feature point pairs of the image to be registered and the reference image, and finally estimates a projection transformation matrix from the image to be registered to the reference image based on the matched feature point pairs, performs projection transformation on the image to be registered, and registers the image to be registered with the reference image. The method has good noise immunity, can realize the rapid registration of the polarization image, and effectively solves the problem of mismatching of the polarization image of the target, especially the dynamic target, acquired by the time-sharing polarization imaging system at different moments, thereby laying a technical foundation for the time-sharing polarization imaging system to acquire real polarization information of the target in real time and simultaneously providing a new solution for the registration of the polarization image in the polarization imaging field.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (10)
1. A polarization image fast registration method based on ORB feature points is characterized by comprising the following steps:
acquiring at least two intensity images of a target in different polarization directions, and determining a reference image and an image to be registered;
respectively extracting feature points in the reference image and the image to be registered by utilizing an ORB algorithm;
constructing a search tree of the characteristic points of the reference image by using a rapid nearest neighbor search algorithm, and constructing the matched characteristic point pairs in the image to be registered and the reference image according to the search tree;
and estimating a projective transformation matrix from the image to be registered to the reference image according to the characteristic point pairs, and performing projective transformation on the image to be registered to obtain an image registration result.
2. The polarized image fast registration method according to claim 1, wherein: the feature points are binary groups formed by coordinates and corresponding feature descriptors.
3. The polarization image fast registration method according to claim 2, wherein the constructing the matched feature point pairs in the image to be registered and the reference image according to the search tree specifically comprises:
and taking the Hamming distance between the feature descriptors as a similarity measurement criterion of the feature points, searching the corresponding feature point with the minimum Hamming distance for the feature point of the image to be registered in the constructed search tree of the feature points of the reference image, and constructing the matched feature point pair in the image to be registered and the reference image.
4. The polarized image fast registration method according to claim 3, wherein the constructing the matched pairs of feature points in the image to be registered and the reference image further comprises:
and filtering out the characteristic point pairs with the Hamming distance between the characteristic points larger than a preset threshold value from the characteristic point pairs.
5. The polarized image fast registration method according to claim 4, wherein the estimating a projective transformation matrix from the image to be registered to the reference image according to the feature point pairs specifically comprises:
extracting a coordinate pair in the feature point pairs;
estimating a projective transformation matrix of the image to be registered to the reference image according to the following formula:
wherein the content of the first and second substances,
a projective transformation matrix for the image to be registered to the reference image,
representing the kth feature in the reference imageThe coordinates of the points are such that,
and representing the coordinates of the t-th characteristic point in the m-th image to be registered.
6. A polarized image fast registration device based on ORB feature points is characterized by comprising:
the image acquisition unit is used for acquiring at least two intensity images of the target in different polarization directions and determining a reference image and an image to be registered;
the characteristic extraction unit is used for respectively extracting characteristic points in the reference image and the image to be registered by utilizing an ORB algorithm;
the characteristic matching unit is used for constructing a search tree of the characteristic points of the reference image by utilizing a rapid nearest neighbor search algorithm and constructing the characteristic point pairs matched with the reference image in the image to be registered according to the search tree;
and the image registration unit is used for estimating a projection transformation matrix from the image to be registered to the reference image according to the characteristic point pairs, and performing projection transformation on the image to be registered to obtain an image registration result.
7. The polarized image fast registration apparatus according to claim 6, wherein the feature points are binary groups composed of coordinates and corresponding feature descriptors;
the feature matching unit is used for executing the following operations:
using the Hamming distance between the feature descriptors as a similarity measurement criterion of the feature points, searching the corresponding feature point with the minimum Hamming distance for the feature point of the image to be registered in the constructed search tree of the feature points of the reference image, and constructing the matched feature point pair in the image to be registered and the reference image;
the feature matching unit is further configured to:
and filtering out the characteristic point pairs with the Hamming distance between the characteristic points larger than a preset threshold value from the characteristic point pairs.
8. The polarized image fast registration apparatus according to claim 7, wherein the image registration unit is configured to perform the following operations:
extracting a coordinate pair in the feature point pairs;
estimating a projective transformation matrix of the image to be registered to the reference image according to the following formula:
wherein the content of the first and second substances,
a projective transformation matrix for the image to be registered to the reference image,
coordinates representing the kth feature point in the reference image,
and representing the coordinates of the t-th characteristic point in the m-th image to be registered.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: the processor, when executing the computer program, implements the steps of the method for fast registration of polarized images according to any of claims 1 to 5.
10. A computer-readable storage medium storing a computer program, characterized in that: the computer program when executed by a processor implements the steps of the method for fast registration of polarized images according to any of claims 1 to 5.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010730007.4A CN111882594A (en) | 2020-07-27 | 2020-07-27 | ORB feature point-based polarization image rapid registration method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010730007.4A CN111882594A (en) | 2020-07-27 | 2020-07-27 | ORB feature point-based polarization image rapid registration method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111882594A true CN111882594A (en) | 2020-11-03 |
Family
ID=73201490
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010730007.4A Pending CN111882594A (en) | 2020-07-27 | 2020-07-27 | ORB feature point-based polarization image rapid registration method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111882594A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114581326A (en) * | 2022-03-03 | 2022-06-03 | 上海应用技术大学 | OCT imaging distortion correction method and device |
| CN114937067A (en) * | 2022-06-14 | 2022-08-23 | 陕西师范大学 | Image registration method of sub-aperture polarization camera |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104167003A (en) * | 2014-08-29 | 2014-11-26 | 福州大学 | Method for fast registering remote-sensing image |
| 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 |
| CN108765476A (en) * | 2018-06-05 | 2018-11-06 | 安徽大学 | A kind of polarization image method for registering |
| CN109829853A (en) * | 2019-01-18 | 2019-05-31 | 电子科技大学 | A kind of unmanned plane image split-joint method |
-
2020
- 2020-07-27 CN CN202010730007.4A patent/CN111882594A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104167003A (en) * | 2014-08-29 | 2014-11-26 | 福州大学 | Method for fast registering remote-sensing image |
| 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 |
| CN108765476A (en) * | 2018-06-05 | 2018-11-06 | 安徽大学 | A kind of polarization image method for registering |
| CN109829853A (en) * | 2019-01-18 | 2019-05-31 | 电子科技大学 | A kind of unmanned plane image split-joint method |
Non-Patent Citations (2)
| Title |
|---|
| 梁楠等: "基于改进尺度不变特征变换的手术室多视点图像拼接算法", 《北京生物医学工程》, vol. 37, no. 1, pages 9 - 14 * |
| 陈丽莉等: "一种有效的序列图像自动拼接方法", 《光电子·激光》, vol. 22, no. 7, pages 1112 - 1118 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114581326A (en) * | 2022-03-03 | 2022-06-03 | 上海应用技术大学 | OCT imaging distortion correction method and device |
| CN114581326B (en) * | 2022-03-03 | 2024-02-02 | 上海应用技术大学 | OCT imaging distortion correction method and device |
| CN114937067A (en) * | 2022-06-14 | 2022-08-23 | 陕西师范大学 | Image registration method of sub-aperture polarization camera |
| CN114937067B (en) * | 2022-06-14 | 2023-01-31 | 陕西师范大学 | Image registration method of split-aperture polarization camera |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| WO2021000664A1 (en) | Method, system, and device for automatic calibration of differences in cross-modal target detection | |
| Mikolajczyk et al. | An affine invariant interest point detector | |
| Tell et al. | Wide baseline point matching using affine invariants computed from intensity profiles | |
| Bak et al. | Improving person re-identification by viewpoint cues | |
| CN109118523A (en) | A kind of tracking image target method based on YOLO | |
| Liu et al. | A contrario comparison of local descriptors for change detection in very high spatial resolution satellite images of urban areas | |
| CN108197604A (en) | Fast face positioning and tracing method based on embedded device | |
| Wang et al. | Airport detection in remote sensing images based on visual attention | |
| Son et al. | A multi-vision sensor-based fast localization system with image matching for challenging outdoor environments | |
| CN111882594A (en) | ORB feature point-based polarization image rapid registration method and device | |
| CN111932582A (en) | Target tracking method and device in video image | |
| CN111932579A (en) | Method and device for adjusting equipment angle based on motion trail of tracked target | |
| Cai et al. | A target tracking method based on KCF for omnidirectional vision | |
| Li et al. | Unmanned aerial vehicle image matching based on improved RANSAC algorithm and SURF algorithm | |
| CN113971684A (en) | Real-time robust target tracking method based on KCF and SURF characteristics | |
| Qiu et al. | An adaptive kernel‐based target tracking method based on multiple features fusion | |
| Xiang | Active learning for person re-identification | |
| KR100540889B1 (en) | Tracking Method for The Moving Object using Basic Points | |
| Xu et al. | Accurate registration of multitemporal UAV images based on detection of major changes | |
| Mithun et al. | Learning long-term invariant features for vision-based localization | |
| Miao et al. | Kernel-based online object tracking via gaussian mixture model learning | |
| Deshmukh et al. | A comparative study of feature detection methods | |
| Li et al. | Scale invariant feature matching using rotation-invariant distance for remote sensing image registration | |
| Osborne et al. | Temporally stable feature clusters for maritime object tracking in visible and thermal imagery | |
| Zhang | Sparse Visual Localization in GPS-Denied Indoor Environments |
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 |