CN112462367B - Vehicle detection method based on polarized synthetic aperture radar - Google Patents
Vehicle detection method based on polarized synthetic aperture radar Download PDFInfo
- Publication number
- CN112462367B CN112462367B CN202011183412.5A CN202011183412A CN112462367B CN 112462367 B CN112462367 B CN 112462367B CN 202011183412 A CN202011183412 A CN 202011183412A CN 112462367 B CN112462367 B CN 112462367B
- Authority
- CN
- China
- Prior art keywords
- target
- matrix
- training
- training set
- test set
- 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.)
- Expired - Fee Related
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 56
- 238000012549 training Methods 0.000 claims abstract description 97
- 238000012360 testing method Methods 0.000 claims abstract description 78
- 239000011159 matrix material Substances 0.000 claims abstract description 70
- 238000000034 method Methods 0.000 claims abstract description 40
- 230000010287 polarization Effects 0.000 claims abstract description 22
- 230000009467 reduction Effects 0.000 claims abstract description 21
- 230000001427 coherent effect Effects 0.000 claims abstract description 18
- 238000000513 principal component analysis Methods 0.000 claims abstract description 13
- 238000005259 measurement Methods 0.000 claims abstract description 12
- 238000012545 processing Methods 0.000 claims abstract description 5
- 238000001914 filtration Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 239000013598 vector Substances 0.000 description 14
- 230000006870 function Effects 0.000 description 8
- 238000009826 distribution Methods 0.000 description 6
- 230000015654 memory Effects 0.000 description 6
- 238000010586 diagram Methods 0.000 description 5
- 238000003384 imaging method Methods 0.000 description 5
- 238000003860 storage Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 3
- 238000012216 screening Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000006835 compression Effects 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013179 statistical model Methods 0.000 description 2
- OAICVXFJPJFONN-UHFFFAOYSA-N Phosphorus Chemical compound [P] OAICVXFJPJFONN-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000012669 compression test Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000003912 environmental pollution Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
- G01S13/89—Radar or analogous systems specially adapted for specific applications for mapping or imaging
- G01S13/90—Radar or analogous systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques, e.g. synthetic aperture radar [SAR] techniques
- G01S13/9021—SAR image post-processing techniques
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/411—Identification of targets based on measurements of radar reflectivity
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
- G01S7/41—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00 using analysis of echo signal for target characterisation; Target signature; Target cross-section
- G01S7/418—Theoretical aspects
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Radar, Positioning & Navigation (AREA)
- Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Physics & Mathematics (AREA)
- Electromagnetism (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention provides a vehicle detection method based on a polarization synthetic aperture radar, and belongs to the technical field of radar image signal processing. The method comprises the following steps: acquiring a training set and a test set; compressing the coherent matrix of each sample in the obtained training set and test set channel by a two-dimensional principal component analysis algorithm of L1 norm to obtain a training set and a test set after dimensionality reduction; detecting a potential target by using the complex Weissett distance measurement according to the obtained training set and the test set after dimension reduction; carrying out target identification on potential targets detected by the rewwester distance measurement by using template matching based on a correlation coefficient; and if a plurality of candidate targets obtained after target identification belong to a rectangular frame of a vehicle, weighting and combining the candidate targets according to the importance degree of the candidate targets to obtain a vehicle detection result. By adopting the invention, the interference of strong reflectors such as buildings and the like can be eliminated, and the positioning accuracy of dense targets is improved.
Description
Technical Field
The invention relates to the technical field of radar image signal processing, in particular to a vehicle detection method based on a polarization synthetic aperture radar.
Background
With the rapid development of economy, the number of private cars in cities is increasing, traffic jam and environmental pollution are caused, and the pressure can be relieved through intelligent regulation and control of vehicles.
Target detection is a key and difficult problem of polar Synthetic Aperture Radar (polar SAR), and is also an important application direction of the polar SAR system. The echo received by the polarized SAR has four basic characteristics of amplitude, phase, frequency and polarization, and can completely describe a target physical scattering process. Different targets have different backscattering strength, dielectric property and geometric property, so that different polarization properties can be shown in a polarized SAR image after the incident electromagnetic waves are subjected to information modulation and loading by the targets, and the basis for realizing detection of different targets is realized.
Because the polarized SAR data contains time domain, frequency domain, time frequency domain and polarization information with rich targets, more resources can be provided for realizing stable target detection, and the method has wide application in many civil and military fields. Generally, conventional target detection methods based on polarized SAR images can be classified into the following three categories:
1) target detection algorithm based on polarization characteristics
The polarized SAR image contains rich polarized scattering features of the target, such as polarized entropy, polarized scattering total power (Span), characteristic value, polarization degree, cross entropy and the like, so many scholars propose many target detection algorithms based on the polarized scattering features. Touzi et al analyzed The performance of ship detection based on polarization information at different incident angles using The polarized SAR data recorded by SAR-580, and The contrast between The ship target and The sea was significantly improved when The incident angle was below 60 °, thereby improving The performance of ship detection (Touzi R. calibrated polar SAR data for shift detection [ C ]// IGARSS 2000.IEEE 2000International geographic and motion Sensing Systemum. labeling The Pulse of The ship: The roll of motion Sensing in manufacturing The environment. proceedings. IEEE 2000,1: 144-146.). Fan et al propose a cross entropy based target detection algorithm, which first uses cross entropy to segment sea and land, extracts the region containing the target of interest, and then uses similarity and Span value to eliminate false target, thereby realizing ship target detection (Fan L, Yang J, Pen Y N.A cross-based parameter for ship detection from a polar SAR image [ C ]//2005IEEE International Symposium on Microwave, Antenna, Propagation and EMC Technologies for Wireless communications.IEEE,2005,1: 6-9.).
2) Target detection algorithm based on constant false alarm rate
Constant False Alarm Rate (CFAR) has been widely applied to target detection of SAR images due to its simple and efficient characteristics, and many scholars also propose many target detection algorithms based on CFAR in combination with the polarization characteristics of polarized SAR. Wu et al designs a CFAR-based ship algorithm by using a polarized feature vector such as Span, feature value and anisotropy, and verifies the performance of the algorithm on RADARSAR-2 measured data (Bingjie W, Chao W, Bo Z, et al. Shield detection based on Radarsat-2full-polar images [ C ]// Proceedings of 2011IEEE CIE International Conference on radio, IEEE,2011,1: 634-. However, the traditional statistical model of the background clutter cannot accurately represent the statistical characteristics of the background clutter under many conditions, and thus the performance of the target detection algorithm is reduced. Therefore, many scholars propose improved algorithms such as two-parameter CFAR, order statistic CFAR, mean CFAR and CFAR based on gray scale correlation, but the detection effect is improved to a limited extent. For polarized SAR image data in a plurality of complex scenes, such as urban complex ground target detection tasks, the CFAR algorithm based on the statistical model still has the problems of high detection result omission ratio and false alarm rate caused by inaccurate clutter representation model in target detection.
3) Target detection algorithm based on polarization statistical distribution
Many researchers have conducted research work on statistical distribution modeling and model parameter estimation for polarized SAR image data. Song et al proposed a variational Bayesian-based target detection algorithm using the complex covariance matrix of polarized SAR and obeying to multivariate Gaussian distribution, and achieved a good effect on the data set of RADASARSET-2 (Song S, Xu B, Yang J.Ship detection in polar SAR images using targets' sparse property [ C ]//2016IEEE International geographic science and motion Sensing Symposium (IGARSS). IEEE,2016: 5706-. We et al assume that the covariance matrix of the results of polarized SAR imaging obeys complex Wishart distribution, propose an objective Detection algorithm Using Wishart distance and polarized Span information, and verify the validity of the algorithm on the AIRSAR dataset (Wei J, Li P, Yang J, et al. A New Automatic Shield Detection Method Using L-Band polar SAR imaging [ J ]. IEEE Journal of Selected Topics in Applied Earth observation and removal Sensing,2013,7(4): 1383-. Ding et al propose a CFAR ship target detection algorithm based on truncated Gamma distribution in order to suppress the influence of strong sea clutter and other artificial targets on ship target detection, model the statistical characteristics of the sea clutter background by a classification method and estimate model parameters, then detect the ship by using the CFAR detection algorithm, and verify the validity of the algorithm on the RADARSET-2 data set (Tao D, Doulgeris A P, Brekke C.A segmentation-based CFAR detection algorithm used on the specified statistics [ J ]. IEEE Transactions on diagnostics and removal Sensing,2016,54(5): 2887-.
Many of the methods described above require accurate acquisition of the polarization scatter information, characterization models, and statistical properties of the clutter. Under a simple uniform clutter background, such as sea surface ship detection, the ground object type is single, and information estimated by using the background clutter conforms to the actual condition; for artificial target detection in complex polarization SAR scenes including a large number of ground object types, such as cities, because clutter edges and interference caused by multiple targets, information estimated by utilizing clutter often does not conform to reality, and the practicability of many methods is greatly reduced. Therefore, most of the existing algorithms are directed at sea surface target detection tasks, and few detection researches on ground targets are carried out, especially the problem of intensive target detection in complex environments.
Disclosure of Invention
The embodiment of the invention provides a vehicle detection method based on a polarized synthetic aperture radar, which can eliminate the interference of strong reflectors such as buildings and the like and improve the positioning accuracy of dense targets. The technical scheme is as follows:
in one aspect, a method for detecting a vehicle based on a polarimetric synthetic aperture radar is provided, and the method is applied to an electronic device, and includes:
acquiring a training set and a test set;
compressing the coherent matrix of each sample in the obtained training set and test set channel by a two-dimensional principal component analysis algorithm of L1 norm to obtain a training set and a test set after dimensionality reduction;
detecting a potential target by using the complex Weissett distance measurement according to the obtained training set and the test set after dimension reduction;
carrying out target identification on potential targets detected by the rewwester distance measurement by using template matching based on a correlation coefficient;
and if a plurality of candidate targets obtained after target identification belong to a rectangular frame of a vehicle, weighting and combining the candidate targets according to the importance degree of the candidate targets to obtain a vehicle detection result, wherein the importance degree is represented by the complex Weissett distance and the correlation coefficient.
Further, prior to acquiring the training set and the test set, the method further comprises:
acquiring a polarized synthetic aperture radar image corresponding to a scene to be detected on land;
sequentially extracting an S matrix corresponding to each pixel from the polarized synthetic aperture radar image;
conjugate multiplication is carried out through Pauli bases, and the S matrix is converted into a coherent matrix T;
and filtering the coherent matrix T by adopting fine Lee filtering to obtain a filtered T matrix.
Further, the S matrix is:
wherein the S matrix satisfies a reciprocal symmetry law: shv＝Svh，ShhRepresenting horizontally transmitted, horizontally received polarization channel data, ShvRepresenting horizontally transmitted, vertically received polarization channel data, SvhRepresenting polarisation channel data transmitted in a vertical manner, received in a horizontal manner, SvvThe data of the polarization channels transmitted in a vertical mode and received in a vertical mode are shown.
Further, the coherence matrix T is:
wherein k represents a Pauli group,
the superscript H denotes the conjugate transpose, denotes the complex conjugate,<·>expressing an average value;
therein, phaseAll information of the stem matrix T is: t is11、T22、T33、real(T12)、real(T13)、real(T23)、imag(T12)、imag(T13) And imag (T)23) The data of the 9 channels are replaced, real represents the real part of the complex number, and imag represents the imaginary part of the complex number.
Further, the acquiring the training set and the test set includes:
setting the size of a sliding window according to the size of a target, wherein the target is a vehicle, and the size of the sliding window comprises the following steps: the width and height of the sliding window;
setting the sliding step length of a sliding window;
sliding the sliding window on the whole polarimetric synthetic aperture radar image according to the set size and sliding step length of the sliding window, and taking out an image block covered by the sliding window each time as a test sample in a test set, wherein all the test samples form the test set;
and forming a training set by the sections of the partial target and various backgrounds selected on the whole polarimetric synthetic aperture radar image, wherein each section in the training set is used as a training sample, the size of each section is consistent with that of the sliding window, the labels of the sections of the partial target are positive, and the labels of the sections of various backgrounds are negative.
Further, the compressing the coherent matrix of each sample in the obtained training set and test set channel by using a two-dimensional principal component analysis algorithm of L1 norm to obtain the training set and test set after dimensionality reduction includes:
selecting the number of the principal components of the training sample according to the use purpose;
under the condition of selecting the number of principal components, processing the training set through a two-dimensional principal component analysis algorithm of an L1 norm to obtain a projection matrix;
and projecting the training set and the test set on the projection matrix to obtain the training set and the test set after dimension reduction.
Further, the selecting the number of the training sample principal components according to the purpose of use includes:
if the distance of the rewwexate is calculated, the number of the main components of the training sample is selected to be 1;
and if calculating the correlation coefficient for measuring the target space structure, selecting the number of the principal components of the training samples according to the accumulated contribution rate of the eigenvalues of the scatter matrix of the training samples.
Further, the detecting the potential target by using the rewwestert distance metric according to the obtained training set and the test set after the dimensionality reduction includes:
calculating the complex Weissett distance between each test sample and all training samples;
classifying the category of the test sample into a label to which the training sample with the minimum distance belongs, wherein the label comprises: a positive class and a negative class;
and if the label of the test sample belongs to the positive class, keeping the reserved test sample as a potential target, otherwise, discarding the reserved test sample.
Further, the target identification using correlation coefficient-based template matching for potential targets detected by the rewwexate distance metric comprises:
calculating the correlation coefficient of each potential target and all training samples;
classifying the category of the potential target into the label to which the training sample with the maximum correlation coefficient belongs
If the label of the potential target belongs to the positive class, keeping the label, otherwise, abandoning the label;
and if the correlation coefficient of the potential target to be reserved is greater than or equal to a preset threshold value, continuing to reserve the potential target as a candidate target, otherwise, abandoning the potential target, wherein the preset threshold value is the product of a preset constant and the maximum value of the correlation coefficient.
Further, the expression of weighted combination is:
wherein M, N is the average value and correlation coefficient of the complex Weissett distance converted into positive valueThe average value of (a) of (b),
n is the number of candidate objects belonging to a vehicle, WiIs the complex Weissett distance, D, after the conversion to a positive value corresponding to the ith candidate targetiIs the correlation coefficient, X, corresponding to the ith candidate targetiIs the coordinate value of the top left corner of the ith candidate target, and Z is the coordinate of the candidate target after weighted combination.
In one aspect, an electronic device is provided and includes a processor and a memory, where at least one instruction is stored in the memory and loaded into and executed by the processor to implement the method for detecting a vehicle based on a polar synthetic aperture radar.
In one aspect, a computer-readable storage medium having at least one instruction stored thereon is provided, the at least one instruction being loaded and executed by a processor to implement the method for detecting a vehicle based on a polar synthetic aperture radar as described above.
The technical scheme provided by the embodiment of the invention has the beneficial effects that at least:
in the embodiment of the invention, a training set and a test set are obtained; compressing the coherent matrix of each sample in the obtained training set and test set channel by a two-dimensional principal component analysis algorithm of L1 norm to obtain the training set and test set after dimension reduction, improving the robustness of the algorithm to abnormal data and solving the problem of overlarge dynamic range of image data of the polarimetric synthetic aperture radar; according to the obtained training set and test set after dimensionality reduction, a potential target is detected by using the complex Weissett distance measurement, and statistical information of image data is fully utilized; carrying out target identification on potential targets detected by the rewwester distance measurement by using template matching based on a correlation coefficient, and considering the spatial structure information of the targets; and if a plurality of candidate targets obtained after target identification belong to a rectangular frame of a vehicle, weighting and combining the candidate targets according to the importance degree of the candidate targets to obtain a vehicle detection result, wherein the importance degree is represented by the complex Weissett distance and the correlation coefficient. Therefore, through double screening of effective information reflected by the complex Weissett distance and the correlation coefficient, the interference of strong reflectors such as buildings and the like can be eliminated, the effective identification of vehicle targets under the complex background environment is realized, the candidate targets are combined according to the importance degree of the candidate targets in the detection process, the accuracy of positioning the dense targets is improved, and the robustness of the method is improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic flow chart of a method for detecting a vehicle based on a polarized synthetic aperture radar according to an embodiment of the present invention;
FIG. 2 is a detailed flowchart of a method for detecting a vehicle based on a polarized synthetic aperture radar according to an embodiment of the present invention;
fig. 3 is a schematic diagram of a target detection scenario of a polarized SAR vehicle according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of a polarized SAR vehicle target identification provided by an embodiment of the present invention;
FIG. 5 is a schematic diagram of a merged repeating block of a polarized SAR vehicle provided by an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings.
As shown in fig. 1, an embodiment of the present invention provides a method for detecting a vehicle based on a polar synthetic aperture radar, where the method may be implemented by an electronic device, and the electronic device may be a terminal or a server, and the method includes:
s101, acquiring a training set and a test set;
s102, compressing a coherent matrix of each sample in an acquired training set and a test set channel by channel through a two-dimensional Principal Component Analysis (L1 norm-2 dimensional Principal Component Analysis, L1-2DPCA) algorithm of an L1 norm to obtain a training set and a test set after dimensionality reduction;
s103, detecting a potential target by utilizing a Wishart distance measurement according to the obtained training set and the test set after dimension reduction;
s104, identifying the potential target detected by the fuweishate distance measurement by using template matching based on a correlation coefficient;
and S105, if a plurality of candidate targets obtained after target identification belong to a rectangular frame of a vehicle, weighting and combining the candidate targets according to the importance degree of the candidate targets to obtain a vehicle detection result, wherein the importance degree is represented by the complex Weissett distance and the correlation coefficient.
The vehicle detection method based on the polarized synthetic aperture radar obtains a training set and a test set; compressing the coherent matrix of each sample in the obtained training set and test set channel by a two-dimensional principal component analysis algorithm of L1 norm to obtain the training set and test set after dimension reduction, improving the robustness of the algorithm to abnormal data and solving the problem of overlarge dynamic range of image data of the polarimetric synthetic aperture radar; according to the obtained training set and test set after dimensionality reduction, a potential target is detected by using the complex Weissett distance measurement, and statistical information of image data is fully utilized; carrying out target identification on potential targets detected by the rewwester distance measurement by using template matching based on a correlation coefficient, and considering the spatial structure information of the targets; and if a plurality of candidate targets obtained after target identification belong to a rectangular frame of a vehicle, weighting and combining the candidate targets according to the importance degree of the candidate targets to obtain a vehicle detection result, wherein the importance degree is represented by the complex Weissett distance and the correlation coefficient. Therefore, through double screening of effective information reflected by the complex Weissett distance and the correlation coefficient, the interference of strong reflectors such as buildings and the like can be eliminated, the effective identification of vehicle targets under the complex background environment is realized, the candidate targets are combined according to the importance degree of the candidate targets in the detection process, the accuracy of positioning the dense targets is improved, and the robustness of the method is improved.
In the foregoing specific embodiment of the method for detecting a vehicle based on a polar synthetic aperture radar, further, as shown in fig. 2 and 3, before the training set and the test set are acquired, the method further includes:
a1, acquiring a polarimetric synthetic aperture radar image, namely PolSAR data, corresponding to a scene to be detected on land, wherein the data comprises: hh. hv, vh and vv polarization channel data;
a2, sequentially extracting an S matrix corresponding to each pixel from the polarized synthetic aperture radar image, wherein the S matrix is as follows:
wherein the S matrix satisfies a reciprocal symmetry law: shv＝Svh，ShhRepresenting horizontally transmitted, horizontally received polarization channel data, ShvRepresenting horizontally transmitted, vertically received polarization channel data, SvhRepresenting polarisation channel data transmitted in a vertical manner, received in a horizontal manner, SvvThe data of the polarization channels transmitted in a vertical mode and received in a vertical mode are shown.
A3, converting the S matrix into a coherent matrix T by conjugate multiplication of Pauli bases:
wherein k represents a Pauli group,
In this embodiment, the diagonal elements of the coherence matrix T are real numbers, other elements are complex numbers, and other elements are conjugate equal with respect to the diagonal, so the coherence matrix T is conjugate symmetric, and thus all information of the coherence matrix T can be replaced by data of the following 9 channels: t is11、T22、T33、real(T12)、real(T13)、real(T23)、imag(T12)、imag(T13) And imag (T)23) Real denotes taking the real part of the complex number and imag denotes taking the imaginary part of the complex number.
In this embodiment, different channels of the coherence matrix T reflect different scattering characteristics.
A4, filtering by fine Lee (refined Lee), and filtering the coherent matrix T to obtain a filtered T matrix.
In an embodiment of the foregoing method for detecting a vehicle based on a polar synthetic aperture radar, further, the acquiring a training set and a test set includes:
b1, setting the size of the sliding window according to the size of a target, wherein the target is a vehicle, and the size of the sliding window comprises: the width and height of the sliding window;
b2, setting the sliding step of the sliding window;
b3, sliding the sliding window on the whole polarimetric synthetic aperture radar image according to the set size and sliding step length of the sliding window, and taking the image block covered by the sliding window each time as a test sample in a test set, wherein all the test samples form the test set;
in this embodiment, the sliding step size and the sliding window size both need to be adaptively adjusted according to the imaging size of the real target (i.e., the vehicle) on the image, the sliding step size is proportional to the imaging size, and the sliding window size is consistent with the imaging size.
And B4, forming a training set by the slices of the partial target and various backgrounds selected on the whole polarimetric synthetic aperture radar image, wherein each slice in the training set is used as a training sample, the size of each slice is consistent with that of the sliding window, the label of the slice of the partial target is a positive type, and the label of the slice of the various backgrounds is a negative type.
In this embodiment, the partial target refers to selecting partial vehicles on the polarimetric synthetic aperture radar image, for example, if there are 50 vehicles in the scene to be detected, 8 vehicle groups can be selected to form a training set positive type; the various types of backgrounds refer to: and other background interferences except the target in the scene to be detected comprise the ground, the building and the like.
In an embodiment of the foregoing method for detecting a vehicle based on a polar synthetic aperture radar, further, the compressing, channel by channel, a coherent matrix of each sample in the obtained training set and test set by a two-dimensional principal component analysis algorithm of a norm L1, and obtaining the training set and test set after the dimension reduction includes:
selecting the number of the principal components of the training sample according to the use purpose;
under the condition of selecting the number of principal components, processing the training set through a two-dimensional principal component analysis algorithm of an L1 norm to obtain a projection matrix W (namely a projection subspace);
using the obtained projection matrix W, channel-by-channel (relating to T) is performed on the coherence matrix of each sample in the acquired training set and test set11,T22,T33,real(T12),real(T13),real(T23),imag(T12),imag(T13),imag(T23) These 9 channels) to obtain a training set X 'and a test set Y' after dimensionality reduction.
In this embodiment, the nature of the compression: projecting the training set X and the test set Y on a projection matrix W, wherein the projection method comprises the following steps:
X′i＝Xi*W,Y′j＝Yj*W,i＝1...M,j＝1...N
where M represents the number of samples in the training set, N represents the number of samples in the test set, XiRepresents the ith sample of the training set before compression, YjBefore representing compressionTest set j sample, X'iRepresenting the ith sample, Y 'in the training set after dimensionality reduction'jRepresenting the j sample in the test set after dimensionality reduction.
In this embodiment, the training set and the test set share the same projection matrix W, and the projection matrix W can be obtained only through the training set.
In this embodiment, the obtaining of the projection matrix W by using the training set may specifically include the following steps:
firstly, determining an objective function for solving an optimal projection matrix W:
wherein, WoptRepresents the optimal projection matrix W, | · | | non-woven phosphor1Representing the 1 norm, X of the vectori(j,: represents a training set sample XiThe j-th row of (a), W represents the projection matrix,
l1 norm representing the matrix, defined as
DijRepresenting the elements of matrix D at row i and column j.
For the solution of the objective function, an iterative greedy algorithm may be adopted to solve, and the objective function formula of the algorithm for each projection vector W in the projection matrix W is:
then, a sign function u is definedij(t)，
t represents the number of iterations, and this sign function is substituted into the equation:
the method can be obtained by the following steps:
to make the expression:
the value of the objective function of (d) is maximum, and the update of w (t +1) is obtained by:
and solving the projection vector of the next iteration by the above formula, and stopping the iteration until the objective function value is converged.
In this embodiment, since the projection matrix is composed of a plurality of projection vectors, but only one projection vector can be obtained each time, after the previous projection vector is determined, in order to ensure that each projection vector of the projection matrix is orthogonal to other projection vectors, it is necessary to obtain a new projection vector each time the objective function value converges, and then update the training sample, where the update formula is:
wherein, wk-1(k 1, 2.. d) is the (k-1) th projection vector of the projection matrix W,
and
training samples corresponding to the kth projection vector and the kth-1 projection vector respectively;
the next projection vector may then be found using the updated training samples.
In the foregoing specific embodiment of the method for detecting a vehicle based on a polar synthetic aperture radar, further, the selecting the number of the training sample principal components according to the purpose of use includes:
if the complex Weissett distance representing the target scattering mechanism is calculated, selecting a first principal component from the principal components, namely selecting the number of the principal components of the training sample as 1;
if the correlation coefficient for measuring the target space structure is calculated, selecting the number of the principal components of the training samples according to the accumulated contribution rate of the eigenvalues of the scatter matrix G of the training samples, wherein the scatter matrix G
In the present embodiment, the cumulative contribution rate is generally set to 90% to 95%.
In this embodiment, the classifier used for detecting the potential target by using the complex Wishart distance metric is a minimum distance classifier.
In this embodiment, the complex Wishart minimum distance classification may be performed on the test set by using a complex Wishart minimum distance classifier, and specifically may include the following steps:
c1, calculating the complex Weisset distance between each test sample and all training samples;
in this embodiment, the compound weixate distance is represented as:
dW＝ln(|T2|)+tr(T2 -1T1)
wherein d isWRepresenting the rewwexate distance, T1Is a coherence matrix, T, of the test sample2Is the coherent matrix of the training sample, | T2I represents to solve the coherent matrix T2The determinant of (a), tr (-) represents a trace of the matrix.
C2, classifying the category of the test sample into a label to which the training sample with the smallest distance belongs, wherein the label comprises: a positive class and a negative class;
and C3, if the label of the test sample belongs to the positive class, keeping the test sample as a potential target, otherwise, discarding the test sample.
In this embodiment, the coherent matrix obeys the complex weixate distribution, and a suspected target area can be extracted by using the complex weixate distance, that is: a potential target.
In this embodiment, the classifier used for target identification by using template matching based on a correlation coefficient is a maximum similarity classifier, where template matching refers to finding a portion most similar to an image B in a current image a, generally, the image a is referred to as an input image, and the image B is referred to as a template image, and a specific operation method is to slide the template image B on the image a, traverse all pixels to complete matching, and when the template image B moves to a certain position of the input image a, a correlation coefficient coef between the template image B and a corresponding image block in the input image a is:
wherein, I1(I, j) represents the pixel value of the template image B, I2(i, j) represents the pixel value in the input image a corresponding to the slide position of the template image B, and a larger correlation coefficient indicates a higher degree of similarity.
In this embodiment, the identifying the target by using the classifier with the maximum similarity of the correlation coefficient may specifically include the following steps:
d1, calculating the correlation coefficient of each potential target and all training samples;
d2, classifying the categories of the potential targets into the labels to which the training samples with the maximum relevant coefficients belong
D3, if the label of the potential target belongs to the positive class, keeping, otherwise, abandoning;
d4, if the correlation coefficient of the potential target is greater than or equal to the preset threshold, continuing to retain as the candidate target, as shown in fig. 4, otherwise, discarding, wherein the preset threshold is the product of the preset constant and the maximum value of the correlation coefficient.
In this embodiment, based on the correlation coefficient coef between images, context information reflected by a target neighborhood (the context information refers to information obtained from neighborhood data of the target, and the obtained information mainly includes spatial position information) is used, because the test set is obtained in a sliding window manner, different slices are obtained around the target, that is, the information of the target neighborhood, that is, the context information is referred to, and a spatial structure (the spatial structure refers to that when the correlation coefficient between two images with the same size is calculated by using a coef calculation formula, pixel values of corresponding pixel points of the two images participate in calculation) and scattering characteristics (embodied by 9 channels of a coherence matrix T) are combined, so that vehicle target detection is performed subsequently.
In this embodiment, in order to increase the speed of the method for detecting a vehicle based on the polar synthetic aperture radar, a potential target that is very similar to the training set positive class is left by setting a high threshold, where the threshold is a preset constant C multiplied by the maximum value of the correlation coefficient, then the correlation coefficient of the potential target left in step D3 is compared with the set threshold, if the threshold is smaller than the threshold, the potential target left in step D3 is discarded, and otherwise, the retention is continued.
In this embodiment, the larger the constant C is set, the more false targets are filtered, but the true targets may be mistakenly classified, so C is generally set to 0.75-0.9.
In this embodiment, the candidate targets left after being screened in step D4 may have a plurality of candidate targets (embodied as rectangular frames) corresponding to the same vehicle, so that the plurality of candidate targets in this case need to be merged, the final target of vehicle detection is to mark the vehicle with a rectangular frame, as shown in fig. 5, but after the height and width of the target (embodied by the width and height of the sliding window) have been set, the rectangular frame can be replaced by only the coordinates of the upper left corner of the rectangular frame (candidate target), the upper left corner is represented as a red dot, and one red dot represents one rectangular frame, so that a plurality of red dots belonging to one vehicle can be weighted and averaged into one red dot according to their importance degrees, thereby obtaining the rectangular frame of the vehicle.
In this embodiment, for a plurality of candidate targets belonging to a vehicle, the complex Wishart distances of the candidate targets are all converted into positive values, and then are combined after weighted averaging according to the importance degree (weighted combining, which is essentially a combining repeat box), where the expression of weighted combining is:
wherein M, N are the average value of the complex Weissett distance and the average value of the correlation coefficient after being converted into positive values,
n is the number of candidate objects (i.e., rectangular frames) belonging to a vehicle, WiIs the complex Weissett distance, D, after the conversion to a positive value corresponding to the ith candidate targetiIs the correlation coefficient, X, corresponding to the ith candidate targetiIs the coordinate value of the top left corner of the ith candidate target, and Z is the coordinate of the candidate target after weighted combination.
In this embodiment, the importance degree is represented by a complex Wishart distance and a correlation coefficient, where the smaller the complex Wishart distance is, the larger the correlation coefficient is, the more similar it is to the vehicle target.
In this embodiment, the coordinate of Z is taken as the upper left corner of the rectangular frame, and a unique rectangular frame, that is, the rectangular frame of the vehicle, can be obtained according to the width and height of the sliding window.
In this embodiment, in order to obtain an accurate detection result, it is further necessary to remove isolated points, if the number n of rectangular frames to be merged of the suspected target is very small, the suspected target may be noise in a background with a shape close to a real target, such as a building strong reflector or a ground strong reflector, at this time, the final screening should be performed through the number of merged frames, specifically, a threshold P is set, and when n is smaller than P, the suspected target is considered as a false target and discarded, otherwise, the false target is left.
Fig. 6 is a schematic structural diagram of an electronic device 600 according to an embodiment of the present invention, where the electronic device 600 may generate relatively large differences due to different configurations or performances, and may include one or more processors (CPUs) 601 and one or more memories 602, where the memory 602 stores at least one instruction, and the at least one instruction is loaded and executed by the processor 601 to implement the above-described method for detecting a vehicle based on a polar synthetic aperture radar.
In an exemplary embodiment, a computer-readable storage medium, such as a memory, is also provided that includes instructions executable by a processor in a terminal to perform the method for polarized synthetic aperture radar-based vehicle detection described above. For example, the computer readable storage medium may be a ROM, a Random Access Memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (9)
1. A method for detecting a vehicle based on a polarized synthetic aperture radar, comprising:
acquiring a training set and a test set;
compressing the coherent matrix of each sample in the obtained training set and test set channel by a two-dimensional principal component analysis algorithm of L1 norm to obtain a training set and a test set after dimensionality reduction;
detecting a potential target by using the complex Weissett distance measurement according to the obtained training set and the test set after dimension reduction;
carrying out target identification on potential targets detected by the rewwester distance measurement by using template matching based on a correlation coefficient;
if a plurality of candidate targets obtained after target identification belong to a rectangular frame of a vehicle, weighting and combining the candidate targets according to the importance degree of the candidate targets to obtain a vehicle detection result, wherein the importance degree is represented by the complex Weisset distance and the correlation coefficient;
wherein, the expression of weighted combination is:
wherein M, N are the average value of the complex Weissett distance and the average value of the correlation coefficient after being converted into positive values,
n is the number of candidate objects belonging to a vehicle, WiIs the complex Weissett distance, D, after the conversion to a positive value corresponding to the ith candidate targetiIs the correlation coefficient, X, corresponding to the ith candidate targetiIs the coordinate value of the top left corner of the ith candidate target, and Z is the coordinate of the candidate target after weighted combination.
2. The polar synthetic aperture radar-based vehicle detection method of claim 1, wherein prior to acquiring the training set and the test set, the method further comprises:
acquiring a polarized synthetic aperture radar image corresponding to a scene to be detected on land;
sequentially extracting an S matrix corresponding to each pixel from the polarized synthetic aperture radar image;
conjugate multiplication is carried out through Pauli bases, and the S matrix is converted into a coherent matrix T;
and filtering the coherent matrix T by adopting fine Lee filtering to obtain a filtered T matrix.
3. The method for polarized synthetic aperture radar based vehicle detection according to claim 2, wherein the S matrix is:
wherein the S matrix satisfies a reciprocal symmetry law: shv＝Svh，ShhRepresenting horizontally transmitted, horizontally received polarization channel data, ShvRepresenting horizontally transmitted, vertically received polarization channel data, SvhRepresenting polarisation channel data transmitted in a vertical manner, received in a horizontal manner, SvvThe data of the polarization channels transmitted in a vertical mode and received in a vertical mode are shown.
4. The method of claim 3, wherein the coherence matrix T is:
wherein k represents a Pauli group,
the superscript H denotes the conjugate transpose, denotes the complex conjugate,<·>expressing an average value;
all information of the coherence matrix T is used: t is11、T22、T33、real(T12)、real(T13)、real(T23)、imag(T12)、imag(T13) And imag (T)23) The data of the 9 channels are replaced, real represents the real part of the complex number, and imag represents the imaginary part of the complex number.
5. The method of claim 1, wherein the acquiring a training set and a test set comprises:
setting the size of a sliding window according to the size of a target, wherein the target is a vehicle, and the size of the sliding window comprises the following steps: the width and height of the sliding window;
setting the sliding step length of a sliding window;
sliding the sliding window on the whole polarimetric synthetic aperture radar image according to the set size and sliding step length of the sliding window, and taking out an image block covered by the sliding window each time as a test sample in a test set, wherein all the test samples form the test set;
and forming a training set by the sections of the partial target and various backgrounds selected on the whole polarimetric synthetic aperture radar image, wherein each section in the training set is used as a training sample, the size of each section is consistent with that of the sliding window, the labels of the sections of the partial target are positive, and the labels of the sections of various backgrounds are negative.
6. The method according to claim 1, wherein the compressing the coherent matrix of each sample in the training set and the test set by the two-dimensional principal component analysis algorithm with L1 norm channel by channel to obtain the training set and the test set after dimension reduction comprises:
selecting the number of the principal components of the training sample according to the use purpose;
under the condition of selecting the number of principal components, processing the training set through a two-dimensional principal component analysis algorithm of an L1 norm to obtain a projection matrix;
and projecting the training set and the test set on the projection matrix to obtain the training set and the test set after dimension reduction.
7. The method of claim 6, wherein the selecting the number of training sample principal components according to the purpose of use comprises:
if the distance of the rewwexate is calculated, the number of the main components of the training sample is selected to be 1;
and if calculating the correlation coefficient for measuring the target space structure, selecting the number of the principal components of the training samples according to the accumulated contribution rate of the eigenvalues of the scatter matrix of the training samples.
8. The method of claim 1, wherein the detecting potential targets using the wiener distance metric according to the obtained reduced-dimension training set and test set comprises:
calculating the complex Weissett distance between each test sample and all training samples;
classifying the category of the test sample into a label to which the training sample with the minimum distance belongs, wherein the label comprises: a positive class and a negative class;
and if the label of the test sample belongs to the positive class, keeping the reserved test sample as a potential target, and otherwise, discarding the reserved test sample.
9. The method of claim 1, wherein the identifying the potential targets detected by the complex vessant distance metric using correlation coefficient-based template matching comprises:
calculating the correlation coefficient of each potential target and all training samples;
classifying the category of the potential target into the label to which the training sample with the maximum correlation coefficient belongs
If the label of the potential target belongs to the positive class, keeping the label, otherwise, abandoning the label;
if the correlation coefficient of the reserved potential target is larger than or equal to a preset threshold value, continuing to reserve the potential target as a candidate target, otherwise, abandoning the potential target, wherein the preset threshold value is the product of a preset constant and the maximum value of the correlation coefficient.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011183412.5A CN112462367B (en) | 2020-10-29 | 2020-10-29 | Vehicle detection method based on polarized synthetic aperture radar |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011183412.5A CN112462367B (en) | 2020-10-29 | 2020-10-29 | Vehicle detection method based on polarized synthetic aperture radar |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112462367A CN112462367A (en) | 2021-03-09 |
| CN112462367B true CN112462367B (en) | 2021-09-07 |
Family
ID=74835684
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011183412.5A Expired - Fee Related CN112462367B (en) | 2020-10-29 | 2020-10-29 | Vehicle detection method based on polarized synthetic aperture radar |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112462367B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP4105688A1 (en) | 2021-06-18 | 2022-12-21 | Aptiv Technologies Limited | Method for estimating height of object |
| CN114998606B (en) * | 2022-05-10 | 2023-04-18 | 北京科技大学 | Weak scattering target detection method based on polarization feature fusion |
-
2020
- 2020-10-29 CN CN202011183412.5A patent/CN112462367B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN112462367A (en) | 2021-03-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101221773B1 (en) | Method of classify meteorological and non-meteorological echoes using dual polarization radars | |
| CN110349160B (en) | SAR image segmentation method based on super-pixel and fuzzy C-means clustering | |
| Di Martino et al. | Scattering-based SARBM3D | |
| CN112462367B (en) | Vehicle detection method based on polarized synthetic aperture radar | |
| CN105205484B (en) | Synthetic aperture radar target detection method based on warp wavelet and Wiener filtering | |
| Yang et al. | Evaluating SAR sea ice image segmentation using edge-preserving region-based MRFs | |
| CN107742113B (en) | One kind being based on the posterior SAR image complex target detection method of destination number | |
| KR101255736B1 (en) | Method for classifying meteorological/non-meteorological echoes using single polarization radars | |
| CN107831473B (en) | Distance-instantaneous Doppler image sequence noise reduction method based on Gaussian process regression | |
| Iqbal et al. | SAR image despeckling by selective 3D filtering of multiple compressive reconstructed images | |
| CN103400383A (en) | SAR (synthetic aperture radar) image change detection method based on NSCT (non-subsampled contourlet transform) and compressed projection | |
| CN113570632A (en) | Moving small target detection method based on high-time phase spaceborne SAR sequential image | |
| CN113222898A (en) | Double-voyage SAR image trace detection method based on multivariate statistics and deep learning | |
| Reigber et al. | Detection and classification of urban structures based on high-resolution SAR imagery | |
| CN112215832A (en) | SAR trail image quality evaluation and self-adaptive detection parameter adjustment method | |
| Refice et al. | On the use of anisotropic covariance models in estimating atmospheric DInSAR contributions | |
| Li et al. | Object representation for multi-beam sonar image using local higher-order statistics | |
| CN108932520B (en) | SAR image water body probability mapping method combining prior probability estimation | |
| CN108710816B (en) | Polarized SAR image building detection method based on probability fusion | |
| Arivazhagan et al. | Optimal Gabor sub-band-based spectral kurtosis and Teager Kaiser energy for maritime target detection in SAR images | |
| CN110751652B (en) | SAR image segmentation method based on Bhattacharyya distance and texture mode measurement | |
| CN112698330A (en) | Polarized SAR image non-local mean value method based on ratio-to-digital product | |
| Kuttikkad et al. | Building wide area 2-D site models from high resolution fully polarimetric Synthetic Aperture Radar images | |
| CN114998606B (en) | Weak scattering target detection method based on polarization feature fusion | |
| Sui et al. | Automatic extraction of water in high-resolution SAR images based on multi-scale level set method and Otsu algorithm |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20210907 |