CN102663437A - Spacecraft classifying and identifying method based on generalized Hough transformation - Google Patents

Spacecraft classifying and identifying method based on generalized Hough transformation Download PDF

Info

Publication number
CN102663437A
CN102663437A CN2012101347867A CN201210134786A CN102663437A CN 102663437 A CN102663437 A CN 102663437A CN 2012101347867 A CN2012101347867 A CN 2012101347867A CN 201210134786 A CN201210134786 A CN 201210134786A CN 102663437 A CN102663437 A CN 102663437A
Authority
CN
China
Prior art keywords
image
rcs
spacecraft
target
template
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
Application number
CN2012101347867A
Other languages
Chinese (zh)
Inventor
牛威
苏威
寇鹏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
China Xian Satellite Control Center
Original Assignee
China Xian Satellite Control Center
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by China Xian Satellite Control Center filed Critical China Xian Satellite Control Center
Priority to CN2012101347867A priority Critical patent/CN102663437A/en
Publication of CN102663437A publication Critical patent/CN102663437A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Image Analysis (AREA)

Abstract

The invention discloses a spacecraft classifying and identifying method based on generalized Hough transformation, which comprises the following steps of: drawing a radar cross-section (RCS) sequence image aiming at the RCS sequence of a certain radar observation arc section, calculating the compactness of the pixels of the RCS sequence image, carrying out initial classification, filtering the RCS sequence image, also generating a profile diagram and determining the image of a coarse positioning area by using generalized Hough transformation; extracting the skeleton characteristics of the image of the coarse positioning area and carrying out fine matching with the skeleton characteristic of the image of a template; and finally, carrying out matching verification based on the compactness degree of the pixels. According to the spacecraft classifying and identifying method based on the generalized Hough transformation, the aim of automatic target identification can be realized, and the identification rate is also enhanced.

Description

A kind of spacecraft classifying identification method based on Generalized Hough Transform
Technical field
The invention belongs to aerospace measurement and control field, relate to a kind of spacecraft classifying identification method of the RCS of utilization sequence.
Background technology
China's active service radar adopts the Narrow-band Radar system mostly at present, and the extraterrestrial target identification problem of therefore exploring and studying under the Narrow-band Radar system has important practical significance and wide practical value.Target RCS has comprised abundant target information, is the general data of the target radar characteristic that can obtain of Narrow-band Radar, and therefore utilizing the target classification identification of rcs measurement data is importance of extraterrestrial target radar identification.Because radar is different to the observation angle of spacecraft in difference observation segmental arc, and RCS is relatively more responsive to the variation of attitude angle, makes that the RCS sequence statistical nature and the transform characteristics thereof of the different observation of each spacecraft segmental arc are unstable.The statistic of RCS sequence commonly used comprises distribution characteristic parameters such as the extreme difference, standard deviation, standard inequality, the coefficient of variation of position characteristic parameter such as the average, maximum value minimal value, median of the mean place of describing the target RCS sequence and ad-hoc location and expression RCS sequence degree of scatter on whole real number axis.People such as Jin Sheng, Gao Meiguo is according to the radargrammetry data of 10 observation of 7 targets segmental arc; Adopt nearest neighbor method, utilize above-mentioned statistical nature to discern, the result shows that this method can only carry out preliminary classification identification by the bigger spacecraft of enantiomorph difference; And discrimination is low, is roughly 70%.In transform characteristics, traditional Fourier transform is a kind of global change, the time-frequency local character of the RCS sequence that is beyond expression; Though Short Time Fourier Transform improves to some extent, say the signal analysis method that is still a kind of single resolution in essence, on to the application that presents non-stationary signal characteristic RCS sequence; Still there is limitation; Time-Frequency Analysis Method has bigger advantage with respect to said method, but in practical application, to identification of targets performance and bad; Basically be in the theory study stage, be mainly used in judgement in the reality targeted attitude stability.
Summary of the invention
In order to overcome the deficiency of prior art, the present invention provides a kind of spacecraft classifying identification method based on Generalized Hough Transform, can realize automatic target identification, and improves discrimination.
The technical solution adopted for the present invention to solve the technical problems may further comprise the steps:
(1) to the RCS sequence of certain radar observation segmental arc, be transverse axis with time, the RCS amplitude is that the longitudinal axis is drawn bianry image, obtains the RCS sequence image.
(2) compactness of the pixel of calculating RCS sequence image is carried out classification just, if the absolute value of the compactness difference of certain width of cloth template image is less than 0.05 in this compactness and the ATL, then this template image can be participated in follow-up matching operation; If there is not the compactness of a width of cloth template image to satisfy above-mentioned requirements in the ATL, think that then this target can't discern, be unknown object.
(3) use the mean filter of neighborhood averaging that the RCS sequence image is carried out filtering, the profile that in filtering, extracts the RCS sequence image generates profile diagram.Profile diagram is carried out hole fill, and carry out the outline feature extraction, externally contour feature carries out extracted at equal intervals then, reduces the redundancy of information, obtains discrete outline figure.The principle that extracts is the resolution identification that does not influence characteristics of image; Just to satisfy the accuracy requirement of images match computing; Be controlled in the scope that error allows; Because the template image size of using is 200 * 200 pixels, the outline allowable error is 10 pixels, and it is feasible therefore carrying out 8 times of extracted at equal intervals.Use Generalized Hough Transform to voting with the zone of template image most probable coupling in the corresponding RCS sequence image of discrete outline figure at last, the zone that poll is the highest is the coarse positioning area image.
(4) utilize Mathematical Morphology Method to extract the framework characteristic of coarse positioning area image; Utilization Hausdorff distance is carefully mated the framework characteristic of coarse positioning area image and the framework characteristic of template image; If the Hausdorff that obtains distance thinks then that less than 20 pixels both are complementary, the coarse positioning area image is the matching area image; Otherwise think and mate failure, target to be identified is a unknown object.
(5) draw the respectively compact degree distribution figure of pixel of matching area image and template image.Calculate the Hausdorff distance between the compact degree distribution figure of two pixels; If the Hausdorff distance is less than 20 pixels; Think that then the compact degree distribution of pixel of matching area image and template image is roughly the same; The recognition result that obtains through above-mentioned steps is accurate, and target to be identified is the target of template image representative, otherwise thinks that target to be identified is a unknown object.
The invention has the beneficial effects as follows: because the RCS sequence that radar records is relatively more responsive to the variation of attitude angle; Make spacecraft unstable, cause the effect of the statistical nature that utilizes the RCS sequence and transform characteristics identification space target undesirable in the RCS sequence statistical nature and the transform characteristics thereof of difference observation segmental arc.And the spacecraft of some particular job platform; Because the influence of satellite feature (like spaceborne parabola antenna) or satellite working method (like weather satellite); The RCS sequence of different segmental arcs has similar features to occur; And obviously different with the RCS of other workbench spacecrafts, therefore utilize these characteristics to accomplish identification of targets.The present invention proposes the RCS sequence of each observation segmental arc is regarded as bianry image; Extract wherein edge, skeleton and compactness characteristic; The utilization Generalized Hough Transform is carried out coarse positioning with template image and RCS sequence image to be identified; Use a kind of image matching method to carry out the thin coupling of RCS sequence image then, realize the method for automatic target identification based on the Hausdorff distance.
Description of drawings
Fig. 1 is a method flow diagram of the present invention;
Fig. 2 is the extraterrestrial target RCS sequence of an observation segmental arc;
Fig. 3 is 8 times of extraction figure of contour feature of target to be matched;
Fig. 4 is 8 times of extraction figure of template contours characteristic, and wherein Fig. 4 (a), Fig. 4 (b), Fig. 4 (c) are respectively category-A spacecraft template, category-B spacecraft template, C class spacecraft template;
Fig. 5 treats the voting results figure that matching image carries out Generalized Hough Transform;
Fig. 6 is the framework characteristic figure of image to be matched;
Fig. 7 is the skeleton diagram of spacecraft template image, and wherein Fig. 7 (a), Fig. 7 (b), Fig. 7 (c) are respectively category-A spacecraft template, category-B spacecraft template, C class spacecraft template;
Fig. 8 is the compact degree distribution figure of spacecraft template, and wherein Fig. 8 (a), Fig. 8 (b), Fig. 8 (c) are respectively category-A spacecraft template, category-B spacecraft template, C class spacecraft template;
Fig. 9 be target RCS sequence image to be identified and template image be complementary the part compact degree distribution figure.
The practical implementation method
With the instantiation explanation, the RCS sequence that receives with certain radar detection circle time is an example, and spacecraft under it is carried out Classification and Identification, may further comprise the steps below:
1. convert the RCS sequence to the RCS sequence image
Extracting the RCS sequence of certain circle time, is transverse axis with time, and the RCS amplitude is that the longitudinal axis is drawn bianry image, and the bianry image that obtains is as shown in Figure 2.
2. based on the first classification of pixel compactness
Calculate the compactness of image to be matched, the computing formula of compactness is:
J＝A area/A MER (1)
In the formula (1), A AreaBe the area of this target object, A MERThe area of the minimum boundary rectangle of expression target.
The compactness that obtains image to be matched according to formula (1) is 0.2119; And the compactness of A, B, C, D class targets RCS sequence image is as shown in table 1 in the ATL; Because the absolute value of the compactness difference of image to be matched and A, B, C tertiary target RCS sequence image is less than 0.05; Therefore selected category-A, category-B, C class three class template images are participated in follow-up coupling, and the D class template differs far away, can get rid of.
The compact degree result of table 1 four space-like target RCS sequences
Category-A Category-B The C class The D class
Circle time 1 0.1993 0.2197 0.1535 0.0803
Circle time 2 0.2090 0.2284 0.1545 0.1138
Circle time 3 0.2160 0.2475 0.1526 0.0810
Circle time 4 0.2097 0.2384 0.1536 0.1108
3. based on the matching area coarse positioning of contour feature
Use the mean filter of neighborhood averaging that the RCS sequence image is carried out filtering, in filtering, extract the profile information of RCS sequence image.
Neighborhood averaging is a spatial domain smooth noise technology.For given image f (i, each pixel in j) (m, n).Get its neighborhood S.If S contains M pixel, get its mean value as handling back gained image picture point (m, the gray scale of n) locating.Replacing the original gray scale of this pixel with each pixel grey scale mean value in the neighborhood of pixels, promptly is the neighborhood averaging technology.Suppose that noise n is an additive noise, each point is uncorrelated mutually in the space, and expects to be 0, and variance is σ 2, g is unpolluted image, the image f of noisy through behind the neighborhood averaging is:
f ‾ ( m , n ) = 1 M Σf ( i , j ) = 1 M Σg ( i , j ) + 1 M n ( i , j ) , σ a 2 = 1 M σ 2
After the filtering, the size of noise becomes original 1/M, and the useful information of image is effectively remained owing to progressive mean.What the filtering neighborhood was big or small chooses then and will decide according to the characteristics of concrete extraterrestrial target; Especially for the more sparse point set that distributes; Choosing of neighborhood size will guarantee that effective RCS pixel that participation is calculated in the zone is abundant, and contained effective RCS pixel is enough few around the isolated point noise.Like this,, just can realize to extract the profile of nowhere-dense set, can reject the isolated point The noise again effectively the cutting apart of image as long as suitable thresholding is set when filtering.For further reducing calculated amount, can the hole in the gained contour images be filled, and then carry out outline and extract, main contour feature is extracted.Because follow-up matching operation is based on point set, and continuous point set contains too many redundant information in the contour feature, therefore under the situation that does not influence matching result, can carry out extracted at equal intervals to it, only use representational unique point to mate.The principle that extracts is the resolution identification that does not influence characteristics of image; Just to satisfy the accuracy requirement of matching operation; Be controlled in the scope of allowable error; Because the template image size of using is 200 * 200 pixels, the outline allowable error is 10 pixels, and it is feasible therefore carrying out 8 times of extracted at equal intervals.
Use above-mentioned contour feature method for distilling, obtain the discrete profile diagram of target to be matched as shown in Figure 3 and the discrete profile diagram of template contours characteristic shown in Figure 4.
The utilization Generalized Hough Transform is carried out coarse positioning to discrete profile diagram.Concrete steps are following:
In the discrete profile diagram of target to be matched, get two nearest discrete points arbitrarily, two discrete point coordinates are respectively (x 1, y 1) and (x 2, y 2), calculate the slope α of two discrete point lines, wherein α=(y 2-y 1)/(x 2-x 1), be that index is searched in the Generalized Hough Transform concordance list of template image with α, obtain with reference to angle beta and reference distance r, will bring formula (2) into reference to angle beta and reference distance r
x 0 = ( r / 1 + tg 2 β ) + ( x 1 + x 2 ) / 2 y 0 = ( r · tgβ / 1 + tg 2 β ) + ( y 1 + y 2 ) / 2 - - - ( 2 )
Obtain the position (x of RP 0, y 0), and the pixel value of this position added 1, promptly on this position, carried out single ballot.Then, take off a pair of two nearest discrete points in the clockwise direction, continue according to the method described above to calculate and these two discrete point corresponding reference point positions, and the pixel value of this position is added 1, promptly on this position, carried out single ballot.Repeat said process, until complete discrete profile diagram of traversal.Find the highest position of pixel value in the discrete profile diagram, the highest position of the poll of promptly voting, this position are the central point in coarse positioning zone, are the coarse positioning RP.In the RCS sequence image, be the center with the coarse positioning RP, mark and template image size same area image, this area image is the coarse positioning area image.
The coarse positioning result is as shown in Figure 5, and the highest point of brightness is the center of coarse positioning area image among Fig. 5.4 thin couplings based on the image framework characteristic
Ossified feature extraction algorithm is the basis with corrosion and dilation operation, and specific algorithm is following:
If S (A) represents the skeleton of A, the skeletal definition of mathematical morphology is:
Figure BDA0000160071870000052
In the formula (3), "-" is the set reducing, and difference set is asked in expression; B is a structural element; " Θ " representes erosion operation; " ° " the expression opening operation; (A Θ kB) expression is carried out k corrosion with B to A, and (A Θ kB)=((... (A Θ B) Θ B) Θ ...) Θ B; K representative is corroded into the last iterations before the empty set, i.e. k=max{k| (A Θ kB) ≠ φ with A }.
As stated above, the framework characteristic figure that obtains image to be matched is as shown in Figure 6, and the framework characteristic figure of template image is as shown in Figure 7.
The framework characteristic figure that uses skeleton template as shown in Figure 7 to treat coarse positioning area image in the matching image carefully matees, and matching result is as shown in table 2.
The bone matching result of table 2 image to be matched and three types of spacecraft templates
Category-A Category-B The C class
18.3576 49.0408 37.5366
We are according to the checking of a large amount of measured datas in the past, and matching result can be judged to be this class targets less than 20 pixels, so we can judge that target to be matched possibly be the category-A target.
5. verify based on the coupling of the compact degree of pixel
From image to be matched, be partitioned into the zone of mating most with template image, relatively should the zone and the compact degree distribution situation of pixel of category-A To Template image.
Be illustrated in figure 8 as the compact degree distribution figure of pixel of category-A, category-B, C class spacecraft template image.
Be illustrated in figure 9 as the compact degree distribution figure of pixel of matching area in the image to be matched.
Calculate the Hausdorff distance of above-mentioned image, obtain result shown in the table 3.
The checking result of three types of spacecraft platforms of table 3
Category-A Category-B The C class
7.4773 49.6820 70.3920
We are according to the checking of a large amount of measured datas in the past, and the Hausdorff distance thinks that less than 20 pixels recognition result is accurately credible, so our final judgements target to be matched is the category-A target.
6 measured datas compare the test of recognition performance
Through a large amount of actual measurement RCS sequential test, use the method that proposes among conventional statistical clustering methods, small wave converting method and this paper under identical hardware condition, certain three types of spacecraft platform to be carried out Classification and Identification respectively, test result is as shown in table 4.
Three kinds of methods of table 4 are to the recognition success rate of certain three types of spacecraft platform
Method Category-A/% Category-B/% C class/%
Statistical method 21.42 26.34 24.76
Small wave converting method 59.47 62.68 60.39
This paper method 86.03 88.62 86.21
Visible by table 4, statistical method is minimum to the discrimination of these three types of spacecraft platforms, can't make Classification and Identification basically, though the discrimination of small wave converting method increases, the result is still undesirable.And the method discrimination that this paper adopts is higher than 85%, satisfies the demand of practical application.
Table 5 uses the success ratio of before and after the Generalized Hough Transform certain extraterrestrial target being discerned to reach on average consuming time
Figure BDA0000160071870000061
Figure BDA0000160071870000071
Table 5 has been listed the success ratio of using Generalized Hough Transform to carry out coarse positioning and not using Generalized Hough Transform to carry out the identification of coarse positioning extraterrestrial target and has been reached on average consuming time.Can find out that from table 5 recognition success rate before and after the use characteristic changes little, and recognition speed has improved 30 times nearly, satisfied in the practical application requirement the time.

Claims (2)

1. the spacecraft classifying identification method based on Generalized Hough Transform is characterized in that comprising the steps:
(1) to the RCS sequence of certain radar observation segmental arc, be transverse axis with time, the RCS amplitude is that the longitudinal axis is drawn bianry image, obtains the RCS sequence image;
(2) compactness of the pixel of calculating RCS sequence image is carried out classification just, if the absolute value of the compactness difference of certain width of cloth template image is less than 0.05 in this compactness and the ATL, then this template image can be participated in follow-up matching operation; If there is not the compactness of a width of cloth template image to satisfy above-mentioned requirements in the ATL, think that then this target can't discern, be unknown object;
(3) use the mean filter of neighborhood averaging that the RCS sequence image is carried out filtering, the profile that in filtering, extracts the RCS sequence image generates profile diagram; Profile diagram is carried out hole fills; And carry out the outline feature extraction; Externally contour feature carries out extracted at equal intervals then; Use Generalized Hough Transform to voting with the zone of template image most probable coupling in the corresponding RCS sequence image of discrete outline figure at last, the zone that poll is the highest is the coarse positioning area image;
(4) utilize Mathematical Morphology Method to extract the framework characteristic of coarse positioning area image; Utilization Hausdorff distance is carefully mated the framework characteristic of coarse positioning area image and the framework characteristic of template image; If the Hausdorff that obtains distance thinks then that less than 20 pixels both are complementary, the coarse positioning area image is the matching area image; Otherwise think and mate failure, target to be identified is a unknown object;
(5) draw the respectively compact degree distribution figure of pixel of matching area image and template image; Calculate the Hausdorff distance between the compact degree distribution figure of two pixels; If the Hausdorff distance, thinks then that the compact degree distribution of pixel of matching area image and template image is roughly the same less than 20 pixels, the recognition result that obtains through above-mentioned steps is accurate; Target to be identified is the target of template image representative, otherwise thinks that target to be identified is a unknown object.
2. the spacecraft classifying identification method based on Generalized Hough Transform according to claim 1 is characterized in that: described extracted at equal intervals is 8 times of extracted at equal intervals.
CN2012101347867A 2012-05-03 2012-05-03 Spacecraft classifying and identifying method based on generalized Hough transformation Pending CN102663437A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN2012101347867A CN102663437A (en) 2012-05-03 2012-05-03 Spacecraft classifying and identifying method based on generalized Hough transformation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN2012101347867A CN102663437A (en) 2012-05-03 2012-05-03 Spacecraft classifying and identifying method based on generalized Hough transformation

Publications (1)

Publication Number Publication Date
CN102663437A true CN102663437A (en) 2012-09-12

Family

ID=46772921

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2012101347867A Pending CN102663437A (en) 2012-05-03 2012-05-03 Spacecraft classifying and identifying method based on generalized Hough transformation

Country Status (1)

Country Link
CN (1) CN102663437A (en)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102944868A (en) * 2012-11-23 2013-02-27 北京航空航天大学 Low-radar cross section metal rack and design method thereof
WO2014134860A1 (en) * 2013-03-04 2014-09-12 苏州瑞派宁科技有限公司 Method and apparatus for automatically identifying and segmenting position spectrum
CN106022337A (en) * 2016-05-22 2016-10-12 复旦大学 Planar object detection method based on continuous edge characteristic
RU2620727C2 (en) * 2015-06-08 2017-05-29 Федеральное государственное образовательное бюджетное учреждение высшего профессионального образования Московский технический университет связи и информатики (ФГОБУ ВПО МТУСИ) Method of recognition of space appliances on reduced radar location images
CN108333561A (en) * 2018-03-12 2018-07-27 电子科技大学 Multi-emitting source Passive Location based on position Yu waveform parameter Combined estimator
CN108520205A (en) * 2018-03-21 2018-09-11 安徽大学 A kind of human motion recognition method based on Citation-KNN
CN109664321A (en) * 2018-12-27 2019-04-23 四川文理学院 Mechanical arm, explosive trolley and method for searching
CN112163616A (en) * 2020-09-25 2021-01-01 电子科技大学 Local sparse constraint transformation RCS sequence feature extraction method
CN113692215A (en) * 2021-07-30 2021-11-23 广州佳帆计算机有限公司 System, method and device for adjusting position of patch element

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101733746A (en) * 2009-12-22 2010-06-16 哈尔滨工业大学 Autonomously identifying and capturing method of non-cooperative target of space robot

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101733746A (en) * 2009-12-22 2010-06-16 哈尔滨工业大学 Autonomously identifying and capturing method of non-cooperative target of space robot

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
牛威等: "利用雷达散射截面积序列的特定航天器工作平台分类识别方法", 《西安交通大学学报》, vol. 45, no. 12, 31 December 2011 (2011-12-31), pages 53 - 58 *

Cited By (13)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102944868B (en) * 2012-11-23 2014-05-07 北京航空航天大学 Low-radar cross section metal rack and design method thereof
CN102944868A (en) * 2012-11-23 2013-02-27 北京航空航天大学 Low-radar cross section metal rack and design method thereof
WO2014134860A1 (en) * 2013-03-04 2014-09-12 苏州瑞派宁科技有限公司 Method and apparatus for automatically identifying and segmenting position spectrum
RU2620727C2 (en) * 2015-06-08 2017-05-29 Федеральное государственное образовательное бюджетное учреждение высшего профессионального образования Московский технический университет связи и информатики (ФГОБУ ВПО МТУСИ) Method of recognition of space appliances on reduced radar location images
CN106022337B (en) * 2016-05-22 2019-10-15 复旦大学 A kind of planar target detection method based on continuous boundary feature
CN106022337A (en) * 2016-05-22 2016-10-12 复旦大学 Planar object detection method based on continuous edge characteristic
CN108333561B (en) * 2018-03-12 2020-04-07 电子科技大学 Multi-emission-source passive positioning method based on position and waveform parameter joint estimation
CN108333561A (en) * 2018-03-12 2018-07-27 电子科技大学 Multi-emitting source Passive Location based on position Yu waveform parameter Combined estimator
CN108520205A (en) * 2018-03-21 2018-09-11 安徽大学 A kind of human motion recognition method based on Citation-KNN
CN108520205B (en) * 2018-03-21 2022-04-12 安徽大学 motion-KNN-based human body motion recognition method
CN109664321A (en) * 2018-12-27 2019-04-23 四川文理学院 Mechanical arm, explosive trolley and method for searching
CN112163616A (en) * 2020-09-25 2021-01-01 电子科技大学 Local sparse constraint transformation RCS sequence feature extraction method
CN113692215A (en) * 2021-07-30 2021-11-23 广州佳帆计算机有限公司 System, method and device for adjusting position of patch element

Similar Documents

Publication Publication Date Title
CN102663437A (en) Spacecraft classifying and identifying method based on generalized Hough transformation
BARBER et al. SAR sea ice discrimination using texture statistics- A multivariate approach
CN103064071B (en) Radar target attribute scattering center feature extraction method based on sparse decomposition
CN102541954B (en) Method and system for searching trademarks
CN101526995B (en) Synthetic aperture radar target identification method based on diagonal subclass judgment analysis
CN104122540A (en) Matching dictionary and compressive sensing based radar range profile object identification method
CN105224937A (en) Based on the semantic color pedestrian of the fine granularity heavily recognition methods of human part position constraint
CN102928822B (en) Radar target length calculation method based on high-resolution one-dimensional range profiles
CN103473755B (en) Based on the sparse denoising method of SAR image that change detects
CN104050681A (en) Road vanishing point detection method based on video images
CN103886337A (en) Nearest neighbor subspace SAR target identification method based on multiple sparse descriptions
CN104751166A (en) Spectral angle and Euclidean distance based remote-sensing image classification method
CN101710422B (en) Image segmentation method based on overall manifold prototype clustering algorithm and watershed algorithm
CN109635789B (en) High-resolution SAR image classification method based on intensity ratio and spatial structure feature extraction
CN101853509A (en) SAR (Synthetic Aperture Radar) image segmentation method based on Treelets and fuzzy C-means clustering
CN105069796A (en) Wavelet scatternet-based SAR image segmentation method
CN103679719A (en) Image segmentation method
CN110969212A (en) ISAR image classification method based on spatial transformation three-channel convolution
CN103854290A (en) Extended target tracking method based on combination of skeleton characteristic points and distribution field descriptors
CN101369309A (en) Human ear image normalization method based on active apparent model and outer ear long axis
CN104732552A (en) SAR image segmentation method based on nonstationary condition field
CN116051822A (en) Concave obstacle recognition method and device, processor and electronic equipment
CN109544575A (en) One kind being based on the matched method for reconstructing 3 D contour of ISAR sequence polygon
CN113902969A (en) Zero-sample SAR target identification method fusing similarity of CNN and image
Vatsavayi et al. Efficient ISAR image classification using MECSM representation

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20120912