CN109767462A - VideoSAR interframe method for registering based on quantum particle swarm - Google Patents
VideoSAR interframe method for registering based on quantum particle swarm Download PDFInfo
- Publication number
- CN109767462A CN109767462A CN201811652772.8A CN201811652772A CN109767462A CN 109767462 A CN109767462 A CN 109767462A CN 201811652772 A CN201811652772 A CN 201811652772A CN 109767462 A CN109767462 A CN 109767462A
- Authority
- CN
- China
- Prior art keywords
- image
- calculating
- registration
- parameter
- transformation
- 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
- 239000002245 particle Substances 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 31
- 230000009466 transformation Effects 0.000 claims abstract description 17
- 230000008859 change Effects 0.000 claims abstract description 6
- 239000011159 matrix material Substances 0.000 claims description 15
- 239000013598 vector Substances 0.000 claims description 7
- 230000008602 contraction Effects 0.000 claims description 6
- 238000013519 translation Methods 0.000 claims description 6
- 230000001131 transforming effect Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 abstract description 5
- 238000012545 processing Methods 0.000 abstract description 3
- 238000002474 experimental method Methods 0.000 description 6
- 238000009826 distribution Methods 0.000 description 5
- 241000321453 Paranthias colonus Species 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000010845 search algorithm Methods 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 206010028980 Neoplasm Diseases 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000010339 dilation Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
Landscapes
- Image Processing (AREA)
Abstract
The invention belongs to technical field of image processing, the VideoSAR interframe method for registering based on quantum particle swarm is disclosed, this method comprises the following steps: obtaining SAR floating image to be registered and corresponding reference picture;Transformation parameter is sought using quanta particle swarm optimization, floating image is converted;Using the pixel of image as Multidimensional Point Set, point set includes the point of certain single pixel point and its neighborhood, constructs multi-dimensions histogram using point set, calculates the region mutual information between reference picture and changing image;When region mutual information does not change with the number of iterations and reaches maximum, then registration parameter at this time is optimal registration parameter;Image registration is carried out to floating image using optimal registration parameter, the SAR image after being registrated.The present invention can quickly get rid of local optimum to global convergence, the entire search space of covering, to effectively improve registration accuracy and with Quasi velosity.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a VideoSAR interframe registration method based on quantum particle swarm, which is used for high-precision and rapid registration of two or more images and can effectively solve the problems that the traditional registration method is easy to fall into local extremum and is difficult to cover the whole search space, thereby effectively improving the registration precision and the registration speed.
Background
Image registration is the process of matching and superimposing two or more images acquired at different times, with different sensors, or under different conditions. The image registration is widely applied in the fields of target detection, feature matching, tumor detection, lesion positioning, geological exploration and the like. Because the traditional method is easy to fall into local extremum to cause the error estimation of transformation parameters and is difficult to cover the whole search space, the research on the image registration method with high convergence rate and high registration precision has important significance.
At present, the main registration methods in the prior art are a mutual information-based registration method and a standard particle swarm algorithm. The mutual information-based registration method has the advantages that the obvious geometric features of the image data are not considered, so that extra errors caused by feature extraction are avoided, and automatic registration of the image can be realized. However, this method is prone to fall into local extrema leading to erroneous estimates of the transformation parameters. The standard particle swarm algorithm has the advantages of being popular and easy to understand, and simultaneously endows each particle with the characteristics of self learning and summarization so as to improve the algorithm efficiency. However, the flight speed of the particles is limited, and it is difficult to sufficiently cover the entire search space.
Disclosure of Invention
The invention aims to provide a VideoSAR interframe registration method based on quantum particle swarm aiming at the defects of the prior art, which can cover the whole search space and can rapidly jump out the local optimal solution to obtain the global optimal solution, thereby effectively improving the registration precision and the registration speed.
The technical idea of the invention is as follows: incorporating neighboring pixels as a mutual information statistical region (region mutual information) to enhance the robustness of the similarity measure; the randomness of the quantum is utilized to cover the whole search space, and the global optimal solution of the geometric transformation parameter is obtained with high probability through the quantum particle group intelligent search algorithm. The effectiveness of the invention is verified by processing results of actual measurement data of the VideoSAR in the Sandia laboratory.
According to the above thought, the technical scheme of the invention comprises the following steps:
step 1, acquiring an SAR floating image to be registered and a corresponding reference image;
step 2, calculating a transformation parameter by using a quantum particle group algorithm, and transforming the floating image;
step 3, calculating the regional mutual information between the reference image and the transformed image obtained in the step 2;
step 4, when the mutual region information does not change along with the iteration times and reaches the maximum, the registration parameter at the moment is the optimal registration parameter
And 5, carrying out image registration by using the optimal registration parameters to obtain a registered SAR image.
Compared with the prior art, the invention has the following advantages:
① the invention adopts quantum particle swarm intelligent search algorithm, can get rid of local optimum to global convergence rapidly;
② can cover the whole search space by using the randomness of quanta.
Drawings
To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described as follows:
FIG. 1 is a flow diagram of a VideoSAR interframe registration method based on quantum particle swarm;
FIG. 2(a) is a reference image of a registration experiment;
FIG. 2(b) is a floating image of a registration experiment;
fig. 3 is a graph of registration results;
FIG. 4 is a superimposed display of the registration result image and the reference image;
FIG. 5 is a graph of a comparison of optimization parameter errors based on regional mutual information;
fig. 6 is a graph of the change of the mutual information of the regions of the two methods with the iteration number.
Detailed Description
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 only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Fig. 1 is a schematic flow chart of a video sar interframe registration method based on quantum particle swarm provided in an embodiment of the present invention.
As shown in fig. 1, a video sar interframe registration method based on quantum particle swarm provided in the embodiment of the present invention includes the following steps:
step 1, acquiring an SAR floating image and a reference image to be registered.
The reference image and the floating image are shown in fig. 2, and their sizes are m × n.
And 2, solving a transformation parameter by using a quantum particle group algorithm, and transforming the floating image.
Each particle i contains the following information: setting the problem to be optimized in an N-dimensional solution space, wherein the group X is { X ═ X1,X2,…,XmContains m particles, and at time t, the coordinate of the ith particle in the solution space is marked as Xi(t)=[Xi,1(t),Xi,2(t),…,Xi,N(t)]The particle has only position information and no velocity information, where i is 1,2, …, m. By Pi(t)=[Pi,1(t),Pi,2(t),…,Pi,N(t)]To indicate the optimum fitness value of the individual particles, i.e.A local optimum value; with G (t) ═ G1(t),G2(t),…,GN(t)]To represent the best fitness value, i.e. the global optimum, of the population. And has G (t) ═ Pg(t), where g represents the sequence number corresponding to the particle with the global best position in the population, g ∈ {1,2, …, m }.
(2a) Initializing the rotation angle gamma, X direction translation T of the particle in the search spacexY direction translation TyConstituting a solution space X. The positions of the n particles in the solution space are randomly assigned. The number of particles is n, and the maximum number of iteration steps is itermax。
(2b) Calculating a cost function value from the objective function C (R, T);
wherein x isiThe number of the control points in the floating image X is N; y isiThe number of the control points is M; r is the translation transformation in 3 axes.
(2c) And calculating the average best position C (t) of the particles, and updating the local optimal value and the global optimal value.
(2d) Calculating a random point pi,j(t) updating the position X of the particlei,j(t+1);
Xi,j(t+1)=pi,j(t)±α·|Cj(t)-Xi,j(t)|·ln(1/ui,j(t))
Wherein,and ui,j(t) is a random number uniformly distributed between 0 and 1, ± according to ui,jThe difference of (t) is dynamically adjusted correspondingly, if the random number is larger than 0.5, the negative sign is taken, otherwise, the positive sign is taken, α is called as expansion and contraction factor, and plays a role in determining the contraction speed of the particles.
A control strategy is employed that decreases linearly with the progress of the iteration to determine α, as shown by:
α=(α0-α1)×(itermax-k)/itermax+α1
in the formula, α0>α1,α0To control the initial values of the parameters, α1For final values of the control parameters, itermaxAnd k is the current iteration number.
(2e) And judging whether the iteration step number is met. Not satisfying return (2 b); and if the condition is met, the operation is ended. With derived transformation parameters (gamma, T)x,Ty) And carrying out space transformation on the floating image to obtain a transformed image. And calculating the gray value of each pixel after image space transformation by utilizing bilinear interpolation.
And 3, calculating the regional mutual information between the reference image and the transformed image obtained in the step 2. And taking the pixel points of the image as a multi-dimensional point set, wherein the point set comprises a single pixel point and the points of the neighborhood of the single pixel point, and constructing a multi-dimensional histogram by using the point set.
(3a) And establishing a gray value matrix P representing 8 adjacent pixel points of the transformed image and the floating image.
Each group of pixel point pairs of the reference image and the transformed image may be represented as:
Q(x,y)=[R(x,y),F(x,y)]
where Q (x, y) denotes a pixel point pair, R (x, y) denotes a reference image, and F (x, y) denotes a transform image.
By halfThe pixel point pairs Q (x, y) of the square window with the diameter of r to the two images are respectively taken according to the rows in sequence to obtain two (2r +1)2The column vectors of dimension are connected end to generate d ═ 2(2r +1)2Dimension vector piN ═ m-2r (N-2r) d-dimensional vectors can be generated, forming a d × N matrix.
P=(p1,p2,...,pN)
(3b) A covariance matrix C is calculated.
Subtracting the average value from each element in P to obtain a sequence normalization vector P0The expression is:
normalizing a vector P with a sequence0And (3) calculating a covariance matrix C, wherein the expression is as follows:
(3c) computing joint entropy Hg(C) And edge entropy Hg(CR)、Hg(CF)
Assuming that the high-dimensional distribution is similar to the normal distribution, d independent one-dimensional distributions can be separated from the d-dimensional distribution because each dimension is independent, and then the normal distribution and the covariance sigma of the entropy of the d-dimensional point setdThe relationship of (a) to (b) is as follows:
then the joint entropy Hg(C) And edge entropy Hg(CR)、Hg(CF) Respectively as follows:
wherein Hg(CR) Is obtained by calculating the upper left corner of the covariance matrix CEdge entropy, H, of the reference image R obtained by the matrixg(CF) Is obtained by calculating the lower right corner of the covariance matrix CAnd obtaining the edge entropy of the reference image F by the matrix.
(3d) Calculating the region mutual information RMI:
RMI=Hg(CR)+Hg(CF)-Hg(C)
step 4, when the mutual region information does not change along with the iteration times and reaches the maximum, the current registration parameter is the optimal registration parameter;
and 5, carrying out image registration by using the optimal registration parameters obtained in the step 4 to obtain a registered SAR image.
Further, to verify the correctness of the method of the present invention, the video SAR experimental data is from sandia national Laboratories, and the SAR video is recorded in a gate traffic scene in the air force base of kortland usa. Dividing the video SAR into frames, selecting two frames as a reference image and a floating image, wherein the reference image is shown in figure 2(a), the floating image is shown in figure 2(b), and registering the reference image and the floating image by adopting the method of the invention, and comparing the method with the traditional method, the specific process is as follows:
(1) taking the mutual region information as similarity measurement, affine transformation as a geometric transformation model, and a quantum particle swarm algorithm as a search strategy, and setting the following main parameters: number of population n is 40, iteration number itermaxThe dilation/contraction factor α is initialized to 1.0 and reduced to 0.5 with iterations the registration result is shown in fig. 3 and the overlay of the registered image and the reference image is shown in fig. 4.
(2) In order to verify the superiority of the above-described method of the present invention over the conventional method, the following comparative experiment was performed.
The experiment will be divided into two groups: the first group is a registration method (RMI + PSO) combining region mutual information with a particle swarm algorithm; the second group is a registration method (RMI + QPSO) combining region mutual information with quantum particle swarm optimization.
In the experiment, the number of particles is set to 20, the number of iterations is 100, the initial values of the inertia weight w of the PSO algorithm and the contraction expansion factor α of the QPSO algorithm are both 1.0, and are reduced to 0.5 along with the iteration, and the real parameter txIs set to 4, tyIs set to 3 and θ is set to 6.
Under the above conditions, 50 registration experiments are repeated on the experimental data, and a comparison graph of the optimization parameter errors of the two optimization methods is obtained as shown in fig. 5, and a graph of the change of the region mutual information along with the iteration number is shown in fig. 6. In fig. 5, xRMSE, yRMSE, and θ RMSE represent root mean square error values for the horizontal direction, the vertical direction, and the rotation angle, respectively.
From the above simulation results, it can be seen that: the root mean square error of the two registration methods is not large and is within 1 pixel, which indicates that the registration reaches a sub-pixel level, but the convergence speed of the QPSO algorithm is still obviously higher than that of the PSO, and the local optimal global convergence can be quickly eliminated.
The present invention is not limited to the above preferred embodiments, and any modifications, equivalent replacements, improvements, etc. within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (3)
1. A VideoSAR interframe registration method based on quantum particle swarm is characterized by comprising the following steps:
step 1, acquiring an SAR floating image to be registered and a corresponding reference image;
step 2, calculating a transformation parameter by using a quantum particle group algorithm, and transforming the floating image;
step 3, calculating the regional mutual information between the reference image and the transformed image obtained in the step 2;
step 4, when the mutual region information does not change along with the iteration times and reaches the maximum, the current registration parameter is the optimal registration parameter;
and 5, carrying out image registration by using the optimal registration parameters to obtain a registered SAR image.
2. The method according to claim 1, characterized in that step 2 comprises in particular the following sub-steps:
(2a) initializing the rotation angle gamma, X direction translation T of the particle in the search spacexY direction translation TyForming a solution space X, randomly distributing the positions of n particles in the solution space, wherein the number of the particles is n, and the maximum iteration step number is itermax;
(2b) Calculating a cost function value from the objective function C (R, T);
wherein x isiThe number of the control points in the floating image X is N; y isiThe number of the control points is M; r is a translation transformation in 3 axes;
(2c) calculating the average best position C (t) of the particles, and updating the local optimal value and the global optimal value;
(2d) calculating a random point p by using the average best position C (t) of the particles obtained in (2c)i,j(t) updating the position X of the particlei,j(t+1);
Xi,j(t+1)=pi,j(t)±α·|Cj(t)-Xi,j(t)|·ln(1/ui,j(t))
Wherein,and ui,j(t) is a random number uniformly distributed between 0 and 1, ± according to ui,j(t) making corresponding dynamic adjustment, if the random number is greater than 0.5, taking negative sign, otherwise taking positive sign, α being called expansion and contraction factor, playing the role of determining particle contraction speed;
(2e) judging whether the iteration step number is met or not, if not, returning to the step (2b), and if so, ending; with derived transformation parameters (gamma, T)x,Ty) And carrying out space transformation on the floating image to obtain a transformed image, and calculating the gray value of each pixel after image space transformation by utilizing bilinear interpolation.
3. The method according to claim 1, characterized in that step 3 comprises in particular the following sub-steps:
(3a) establishing a gray value matrix P representing 8 adjacent pixel points around the transformed image and the floating image;
P=(p1,p2,...,pN)
(3b) calculating a covariance matrix C;
(3b1) subtracting the average value from each element in P obtained in (3a) to obtain a sequence normalization vector P0The expression is:
(3b1) using the normalized vector P obtained in (3b1)0And (3) calculating a covariance matrix C, wherein the expression is as follows:
(3c) computing joint entropy Hg(C) And edge entropy Hg(CR)、Hg(CF);
Wherein Hg(CR) Is obtained by calculating the upper left corner of the covariance matrix CEdge entropy, H, of the reference image R obtained by the matrixg(CF) Is obtained by calculating the lower right corner of the covariance matrix CThe edge entropy of a reference image F obtained by the matrix;
(3d) calculating the region mutual information RMI:
RMI=Hg(CR)+Hg(CF)-Hg(C)。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811652772.8A CN109767462A (en) | 2018-12-29 | 2018-12-29 | VideoSAR interframe method for registering based on quantum particle swarm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811652772.8A CN109767462A (en) | 2018-12-29 | 2018-12-29 | VideoSAR interframe method for registering based on quantum particle swarm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109767462A true CN109767462A (en) | 2019-05-17 |
Family
ID=66453155
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811652772.8A Pending CN109767462A (en) | 2018-12-29 | 2018-12-29 | VideoSAR interframe method for registering based on quantum particle swarm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109767462A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111583311A (en) * | 2020-05-14 | 2020-08-25 | 重庆理工大学 | PCBA rapid image matching method |
CN112750152A (en) * | 2021-01-18 | 2021-05-04 | 上海联影医疗科技股份有限公司 | Image registration method and device, computer equipment and storage medium |
CN116152316A (en) * | 2023-04-17 | 2023-05-23 | 山东省工业技术研究院 | Image registration method based on self-adaptive parameter particle swarm algorithm |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101216939A (en) * | 2008-01-04 | 2008-07-09 | 江南大学 | A multi-resolution medical image registration method based on quantum behaviors particle swarm algorithm |
-
2018
- 2018-12-29 CN CN201811652772.8A patent/CN109767462A/en active Pending
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101216939A (en) * | 2008-01-04 | 2008-07-09 | 江南大学 | A multi-resolution medical image registration method based on quantum behaviors particle swarm algorithm |
Non-Patent Citations (5)
Title |
---|
潘婷婷等: "QPSO算法和Powell法结合的多分辨率医学图像配准", 《计算机应用与软件》 * |
王小根等: "基于最大互信息和量子粒子群优化算法的医学图像配准研究", 《计算机应用研究》 * |
谢振平等: "一种基于混合量子行为粒子群优化算法的图像配准方法", 《计算机应用研究》 * |
赵辽英等: "基于尺度不变特征变换和区域互信息优化的多源遥感图像配准", 《物理学报》 * |
陈庆芳等: "基于分块互信息和量子粒子群算法的图像配准", 《数据采集与处理》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111583311A (en) * | 2020-05-14 | 2020-08-25 | 重庆理工大学 | PCBA rapid image matching method |
CN111583311B (en) * | 2020-05-14 | 2023-04-18 | 重庆理工大学 | PCBA rapid image matching method |
CN112750152A (en) * | 2021-01-18 | 2021-05-04 | 上海联影医疗科技股份有限公司 | Image registration method and device, computer equipment and storage medium |
CN116152316A (en) * | 2023-04-17 | 2023-05-23 | 山东省工业技术研究院 | Image registration method based on self-adaptive parameter particle swarm algorithm |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Bai et al. | Adaptive dilated network with self-correction supervision for counting | |
CN109784333B (en) | Three-dimensional target detection method and system based on point cloud weighted channel characteristics | |
WO2021196294A1 (en) | Cross-video person location tracking method and system, and device | |
CN111145228B (en) | Heterologous image registration method based on fusion of local contour points and shape features | |
CN105139420B (en) | A kind of video target tracking method based on particle filter and perception Hash | |
CN108446634B (en) | Aircraft continuous tracking method based on combination of video analysis and positioning information | |
CN103886619B (en) | A kind of method for tracking target merging multiple dimensioned super-pixel | |
CN105856230A (en) | ORB key frame closed-loop detection SLAM method capable of improving consistency of position and pose of robot | |
CN113822278B (en) | License plate recognition method for unlimited scene | |
CN107862735B (en) | RGBD three-dimensional scene reconstruction method based on structural information | |
CN109767462A (en) | VideoSAR interframe method for registering based on quantum particle swarm | |
CN110136177B (en) | Image registration method, device and storage medium | |
CN106570887A (en) | Adaptive Mean Shift target tracking method based on LBP features | |
CN115661255B (en) | Laser SLAM loop detection and correction method | |
CN106295657A (en) | A kind of method extracting human height's feature during video data structure | |
CN105427340A (en) | Video target tracking method based on improved box particle filtering | |
Yi et al. | Ucmctrack: Multi-object tracking with uniform camera motion compensation | |
CN109508674B (en) | Airborne downward-looking heterogeneous image matching method based on region division | |
CN102324043B (en) | Image matching method based on DCT (Discrete Cosine Transformation) through feature description operator and optimization space quantization | |
CN112734818B (en) | Multi-source high-resolution remote sensing image automatic registration method based on residual network and SIFT | |
CN111476167B (en) | One-stage direction remote sensing image target detection method based on student-T distribution assistance | |
Wei et al. | A SIFT-based mean shift algorithm for moving vehicle tracking | |
Han et al. | An improved RANSAC registration algorithm based on region covariance descriptor | |
CN112418250A (en) | Optimized matching method for complex 3D point cloud | |
Zhang et al. | Target tracking for mobile robot platforms via object matching and background anti-matching |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190517 |