CN106019255A - Radar target type recognition method based on one-dimensional image data layer fusion under multiple viewing angles - Google Patents
Radar target type recognition method based on one-dimensional image data layer fusion under multiple viewing angles Download PDFInfo
- Publication number
- CN106019255A CN106019255A CN201610584279.1A CN201610584279A CN106019255A CN 106019255 A CN106019255 A CN 106019255A CN 201610584279 A CN201610584279 A CN 201610584279A CN 106019255 A CN106019255 A CN 106019255A
- Authority
- CN
- China
- Prior art keywords
- radar target
- dimensional
- radar
- picture
- under
- 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
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
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses an implementation method for radar target type recognition based on one-dimensional image data layer fusion under multiple viewing angles. The radar target type recognition method is mainly applied to radar target type recognition of a conventional broadband coherent warning radar in a collaboration system. The radar target type recognition method comprises the mains steps of: carrying out data preprocessing on one-dimensional images under the viewing angles; resolving radar target attitude information under the viewing angles; conducting data fusion registration on the one-dimensional images under the viewing angles; setting a feature extraction threshold; extracting leading-edge and trailing-edge positions of a target one-dimensional image; estimating length of a radar target; and finally carrying out target type recognition. The radar target type recognition method provided by the invention has the advantages that project implementation is easy, the data fusion effect is good, the theoretical foundation of the adopted method is full and the like, and the accuracy rate of radar target type recognition under multiple viewing angles is improved by 5% or more when compared with that of radar target type recognition under a single viewing angle.
Description
Technical field
The present invention is a kind of for based on conventional broadband coherent surveillance radar element under collaboration system, it is achieved under the most one-dimensional picture
Radar target kind identification method, it is achieved the rough sort of radar target.
Background technology
Under each visual angle on the basis of one-dimensional picture, realize radar target kind identification method by data fusion etc., utilize radar mesh
Target type identification information can realize the rough sort of radar target, improves recognition correct rate.Coordinate the transformation of existing detecting devices,
Collaborative target recognition ability can be promoted energetically on the basis of existing detecting devices.
Radar target recognition is the important extension to radar detection function, is also a study hotspot of current radar signal processing,
And technical research of based on one-dimensional picture is one of the study hotspot in radar target recognition field, the most a lot of researchs are based on single-view
Under one-dimensional as carrying out, as in April, 2013 Xian Electronics Science and Technology University's academic dissertation " Radar High Range Resolution target
Identify technical research " in the Radar High Range Resolution target identification method based on TSB-HMM and based on Modifying model that proposes
HRRP noise robust identification method.
Different from the method proposed in other documents, the present invention is directed on the basis of based on picture one-dimensional under various visual angles, by number
The radar target kind identification method merged according to layer, it is achieved the rough sort of radar target.
Summary of the invention
It is an object of the invention to provide the radar target that under a kind of various visual angles solved under collaboration system, one-dimensional picture data Layer merges
Kind identification method, is effectively realized the rough sort of radar target.By means of the invention it is possible to fleet system at sea is fought
System realizes effectively classifying the radar target under various kinestates in sighting distance, and the thunder under the one-dimensional picture of various visual angles
5% is improved on the basis of reaching target recognition accuracy radar target recognition accuracy under the one-dimensional picture of single-view.
The technical solution realizing the present invention is:
First one-dimensional under each visual angle is rejected as data the pretreatment such as bad picture, no-coherence cumulating;Utilize radar target
Course and azimuth information calculate the attitude angle under each visual angle of radar target;Homomorphism is utilized to stretch affine transformation method, in conjunction with minimum entropy pair
Together picture one-dimensional under each visual angle is carried out data fusion registration;One-dimensional picture after merging registration is arranged feature extraction thresholding;Utilize sliding
Along position before and after dynamic average stretching method extraction radar target picture;Before and after utilizing radar target picture, edge calculates radar mesh
Target radical length, the attitude information in conjunction with radar target estimates radar target length;Finally carry out radar target type identification.
Compared with prior art, its remarkable advantage is the present invention:
Use homomorphism stretching affine transformation method that picture one-dimensional under each visual angle is carried out data fusion method, it is possible to disappear accurately and efficiently
One-dimensional picture targe-aspect sensitivity, translation sensitivity and the strength sensitive caused except detection ranges different under different visual angles, different attitude
Impact, and its implementation is simple.Before moving average stretching method can extract radar target picture quickly and efficiently
Back edge position, it is good that the method has adaptivity, and amount of calculation is little, the feature that operational efficiency is high.The proposition of the present invention and engineering are real
Radar target recognition field has highly application value now.
Below in conjunction with the accompanying drawings the present invention is described in further detail.
Accompanying drawing explanation
Fig. 1 is the data flowchart of the present invention.
Fig. 2 be the present invention different visual angles under radar target as normalized schematic diagram.
Fig. 3 be the present invention different visual angles under radar target picture merge registration after one-dimensional picture schematic diagram.
Detailed description of the invention
The present invention is embodied as step and sees accompanying drawing 1.
(1) one-dimensional picture data prediction under each visual angle, method is as follows:
Calculate the kurtosis matrix K of one-dimensional picture, find greatest member max (K) and the one-dimensional picture of correspondence thereof of kurtosis matrix, statistics
Remaining one-dimensional image set closes { xiAnd the kurtosis set { k that peels offi}:
Wherein N is the number of one-dimensional picture, XiIt is the one-dimensional decent notebook data of remaining i-th, μiIt is sample average, σiIt it is sample canonical
Variance, E (Xi-μi)4It is 4 rank centre-to-centre spacing of the one-dimensional decent notebook data of i-th.If kiAnon-normal, then it is assumed that the one-dimensional picture of i-th
For abnormal one-dimensional picture, sample data is set to 0.With one-dimensional picture corresponding to kurtosis matrix greatest member for base picture, use minimum entropy
Power estimation criterion carries out registration process, and the one-dimensional picture after alignment is done no-coherence cumulating.
I(xk)=-log pk
Wherein xkIt is one-dimensional decent notebook data, I (xk) it is quantity of information, X is discrete random variable, PkIt it is event X=xkThe probability occurred,It it is one-dimensional picture sample data sets.
Wherein siFor the no-coherence cumulating result of the i-th distance unit after alignment, N is the number of one-dimensional picture.
(2) radar target attitude algorithm under each visual angle, method is as follows:
Radar target attitude angle under wherein θ is each visual angle, α is radar target course under each visual angle, and β is radar target side under each visual angle
Position.
(3) homomorphism stretching affine transformation method carries out data fusion registration to picture one-dimensional under each visual angle, and its method is as follows:
A) the one-dimensional picture under each visual angle is normalized, such as Fig. 2.
B) combine the radar target attitude information under each visual angle resolved, the radar target picture under each visual angle is stretched to same
Under one attitude angle θ ', calculate the one-dimensional image distance after radar target homomorphism stretches under each visual angle respectively from unit number:
Wherein θ is the attitude angle of radar target, and f is radar target image distance from unit number, and f ' is the radar mesh after homomorphism stretching
Mark one-dimensional image distance from unit number.
C) the radar target picture under each visual angle is stretched to the thunder under attitude angle θ ', after being stretched by integrating step b)
Reach target one-dimensional picture f1、f2。
D) statistics f1、f2Main scattering point, and to f1、f2Main scattering point mates, and obtains belonging to same scattering point
Echo, according to the alignment schemes in (1) by f1、f2Carry out alignment registration, such as Fig. 3.
E) the radar target picture after alignment registration is carried out no-coherence cumulating.
(4) arranging feature extraction thresholding, method is as follows:
Calculate in (3) average and the variance of noise section in the one-dimensional picture after no-coherence cumulating, feature extraction thresholding is set to
The average of noise section and the form of variance sum.
Gate=mean (X)+k*std (X)
Wherein gate is local threshold, and X is noise section data acquisition system in one-dimensional range profile, and k is a constant value coefficient.
(5) edge before and after moving average stretching method extracts radar target picture, method is as follows:
(m, n) splits, and obtains segmentation result P to the one-dimensional picture P after no-coherence cumulating to utilize threshold value gateg(m, n):
To the P after segmentationg(m, n) region, extracted before and after extraction thresholding part along position, at front back edge position by 10
Distance length does moving average, if average is less than 0.2 times of Largest Mean, this territory element is multiplied by the weighted value of 0.05,
The most again with thresholding carry out radar target as zone boundary demarcate, behind spotting zone boundary, from the beginning of maximum, to
Both direction is carried out together with process, if exceeding a certain scope together with territory spacing, then it is assumed that be not a target, then this direction is stopped
Only together with process, the one-dimensional picture region P finally obtainedg' (m, n), one-dimensional picture region Pg' (m, n) distance unit minima is as thunder
Reach target one-dimensional picture forward position, one-dimensional picture region Pg' (m, n) distance unit maximum as radar target as tailing edge.
(6) estimation radar target length, method is as follows:
The one-dimensional picture radical length of calculating radar target:
F=(pEnd-pStart) * s
Wherein pEnd be radar target as tailing edge, pStart be radar target as forward position, s is distance by radar resolution, and f is
Radical length.
Calculating radar target length:
Wherein f is the one-dimensional picture radical length of radar target, and θ ' is the attitude angle after homomorphism stretching affine transformation, and L is that radar target is long
Degree.
(7) radar target type identification, method is as follows:
Calculate and adjudicate the factor:
A1=k*L-180
A2=k*L-80
Wherein A1, A2 are the correction factor more than 0 less than 1 for the judgement factor, k, and L is radar target length.
According to judgement the factor carry out target type discrimination, work as A1 > 0, A2 > 0 time, radar target is judged to large-scale target, when
During A1<0, A2>0, radar target is judged to medium-sized target, and as A1<0, A2,<when 0, radar target is judged to precision target.
Claims (3)
1. the radar target kind identification method merged based on one-dimensional picture data Layer under various visual angles, it is characterised in that: one-dimensional by calculating
The kurtosis matrix K of Range Profile, finds greatest member max (K) and the one-dimensional range profile of correspondence thereof of kurtosis matrix, statistics remaining
Dimension Range Profile set { xiAnd the kurtosis set { k that peels offi, utilize the kurtosis method rejecting abnormalities one-dimensional range profile that peels off;With kurtosis matrix
The one-dimensional range profile that big element is corresponding is base picture, uses minimum entropy estimation criterion one-dimensional range profile to be carried out registration process, to alignment
After one-dimensional range profile do no-coherence cumulating;Course according to radar target and orientation, calculate the attitude angle of naval target;Utilize
Homomorphism stretching affine transformation method carries out data fusion registration to picture one-dimensional under each visual angle;One-dimensional picture after no-coherence cumulating is added up and makes an uproar
Average that part divides and variance, arrange feature extraction thresholding by gate=mean (X)+k*std (X) method, and wherein gate is office
Portion's threshold value, X is noise data set in one-dimensional range profile, and k is a constant value coefficient;Moving average stretching method is utilized to extract radar mesh
Mark edge before and after one-dimensional picture, by edge before and after radar target picture and distance by radar resolution according to f=(pEnd-pStart) * s
Method calculate radar target radical length, wherein pEnd be radar target as tailing edge, pStart is radar target picture
Forward position, s is distance by radar resolution, and f is radical length;By radar target attitude and the radical length of radar target and radar
Trigonometric function relation between target length according toMethod calculates the length of radar target, and wherein f is radar target
One-dimensional picture radical length, θ is attitude angle, and L is radar target length;Finally carry out radar target type identification;Pass through the party
It is correct that method carries out one-dimensional picture radar target type identification accuracy one-dimensional picture radar target type identification under single-view under various visual angles
5% is improved on the basis of rate.
The radar target kind identification method merged based on one-dimensional picture data Layer under various visual angles the most according to claim 1, it is special
Levy and be: homomorphism stretching affine transformation method carries out data fusion method for registering to picture one-dimensional under each visual angle, and the method is by by normalizing
Change process after various visual angles under radar target as homomorphism stretching affine under same attitude angle, in conjunction with main scattering point coupling and
Minimum entropy alignment carries out data fusion registration, that under reduction radar target different visual angles, different detection ranges, different attitude cause
Tie up as targe-aspect sensitivity, translation sensitivity and the impact of strength sensitive.
The radar target kind identification method merged based on one-dimensional picture data Layer under various visual angles the most according to claim 1, it is special
Levy and be: moving average stretching method extracts edge before and after radar target picture, before and after the method is passed through crossing thresholding one-dimensional picture region
Process spotting zone boundary along carrying out moving average extension, extract back edge position before radar target picture, reduce multiple target
Impact back edge position before radar target picture accurately extracted with border outlier.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610584279.1A CN106019255A (en) | 2016-07-22 | 2016-07-22 | Radar target type recognition method based on one-dimensional image data layer fusion under multiple viewing angles |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610584279.1A CN106019255A (en) | 2016-07-22 | 2016-07-22 | Radar target type recognition method based on one-dimensional image data layer fusion under multiple viewing angles |
Publications (1)
Publication Number | Publication Date |
---|---|
CN106019255A true CN106019255A (en) | 2016-10-12 |
Family
ID=57117222
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610584279.1A Pending CN106019255A (en) | 2016-07-22 | 2016-07-22 | Radar target type recognition method based on one-dimensional image data layer fusion under multiple viewing angles |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN106019255A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106405521A (en) * | 2016-10-14 | 2017-02-15 | 中国人民解放军海军七〇工厂 | One-dimensional range profile based target length calculating method and device |
CN106597406A (en) * | 2016-12-02 | 2017-04-26 | 中国船舶重工集团公司第七二四研究所 | Radar target recognition method based on multi-view one-dimensional image decision level fusion |
CN106778564A (en) * | 2016-12-02 | 2017-05-31 | 中国船舶重工集团公司第七二四研究所 | Based on the naval vessels of one-dimensional picture Feature-level fusion under various visual angles and freighter sorting technique |
CN108919284A (en) * | 2018-05-04 | 2018-11-30 | 中国人民解放军海军七〇工厂 | A kind of ship classification method, device and electronic equipment |
CN109682480A (en) * | 2019-01-04 | 2019-04-26 | 三峡大学 | The split type infrared temperature measurement apparatus of acquisition-processing based on wireless data transmission |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102928822A (en) * | 2012-11-02 | 2013-02-13 | 中国船舶重工集团公司第七二四研究所 | Radar target length calculation method based on high-resolution one-dimensional range profiles |
CN103839265A (en) * | 2014-02-26 | 2014-06-04 | 西安电子科技大学 | SAR image registration method based on SIFT and normalized mutual information |
-
2016
- 2016-07-22 CN CN201610584279.1A patent/CN106019255A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102928822A (en) * | 2012-11-02 | 2013-02-13 | 中国船舶重工集团公司第七二四研究所 | Radar target length calculation method based on high-resolution one-dimensional range profiles |
CN103839265A (en) * | 2014-02-26 | 2014-06-04 | 西安电子科技大学 | SAR image registration method based on SIFT and normalized mutual information |
Non-Patent Citations (1)
Title |
---|
刘桢宏: "多视角图像配准方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106405521A (en) * | 2016-10-14 | 2017-02-15 | 中国人民解放军海军七〇工厂 | One-dimensional range profile based target length calculating method and device |
CN106405521B (en) * | 2016-10-14 | 2019-06-14 | 中国人民解放军海军七〇一工厂 | A kind of calculation method and device of the target length based on one-dimensional range profile |
CN106597406A (en) * | 2016-12-02 | 2017-04-26 | 中国船舶重工集团公司第七二四研究所 | Radar target recognition method based on multi-view one-dimensional image decision level fusion |
CN106778564A (en) * | 2016-12-02 | 2017-05-31 | 中国船舶重工集团公司第七二四研究所 | Based on the naval vessels of one-dimensional picture Feature-level fusion under various visual angles and freighter sorting technique |
CN106597406B (en) * | 2016-12-02 | 2019-03-29 | 中国船舶重工集团公司第七二四研究所 | Based on the radar target identification method as Decision-level fusion one-dimensional under multi-angle of view |
CN108919284A (en) * | 2018-05-04 | 2018-11-30 | 中国人民解放军海军七〇工厂 | A kind of ship classification method, device and electronic equipment |
CN109682480A (en) * | 2019-01-04 | 2019-04-26 | 三峡大学 | The split type infrared temperature measurement apparatus of acquisition-processing based on wireless data transmission |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106019255A (en) | Radar target type recognition method based on one-dimensional image data layer fusion under multiple viewing angles | |
Ma et al. | Crossing the line: Crowd counting by integer programming with local features | |
Anishiya et al. | Number plate recognition for indian cars using morphological dilation and erosion with the aid of ocrs | |
CN102928822B (en) | Radar target length calculation method based on high-resolution one-dimensional range profiles | |
Dan et al. | Robust people counting system based on sensor fusion | |
US9928426B1 (en) | Vehicle detection, tracking and localization based on enhanced anti-perspective transformation | |
CN104463877B (en) | A kind of water front method for registering based on radar image Yu electronic chart information | |
CN103530599A (en) | Method and system for distinguishing real face and picture face | |
CN106251332B (en) | SAR image airport target detection method based on edge feature | |
CN108171193B (en) | Polarized SAR (synthetic aperture radar) ship target detection method based on super-pixel local information measurement | |
CN108108667B (en) | A kind of front vehicles fast ranging method based on narrow baseline binocular vision | |
US9928424B2 (en) | Side window detection through use of spatial probability maps | |
CN103778436B (en) | A kind of pedestrian's attitude detecting method based on image procossing | |
CN103745216B (en) | A kind of radar image clutter suppression method based on Spatial characteristic | |
CN106597406B (en) | Based on the radar target identification method as Decision-level fusion one-dimensional under multi-angle of view | |
CN105550703A (en) | Image similarity calculating method suitable for human body re-recognition | |
Pyo et al. | Front collision warning based on vehicle detection using CNN | |
Kim et al. | Autonomous vehicle detection system using visible and infrared camera | |
CN109284664A (en) | Driver assistance system and guardrail detection method | |
Jin et al. | Improved direction estimation for Di Zenzo's multichannel image gradient operator | |
CN106326851B (en) | A kind of method of number of people detection | |
Suchitra et al. | Detection & classification of arrow markings on roads using signed edge signatures | |
CN104318216A (en) | Method for recognizing and matching pedestrian targets across blind area in video surveillance | |
Cao et al. | Power line detection based on symmetric partial derivative distribution prior | |
CN104408430B (en) | License plate positioning method and device |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20161012 |
|
WD01 | Invention patent application deemed withdrawn after publication |