CN106447668B - The small target detecting method restored under IR Scene based on grab sample and sparse matrix - Google Patents
The small target detecting method restored under IR Scene based on grab sample and sparse matrix Download PDFInfo
- Publication number
- CN106447668B CN106447668B CN201610722917.1A CN201610722917A CN106447668B CN 106447668 B CN106447668 B CN 106447668B CN 201610722917 A CN201610722917 A CN 201610722917A CN 106447668 B CN106447668 B CN 106447668B
- Authority
- CN
- China
- Prior art keywords
- matrix
- formula
- infrared
- small target
- image
- 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
- 239000011159 matrix material Substances 0.000 title claims abstract description 102
- 238000000034 method Methods 0.000 title claims abstract description 74
- 238000001514 detection method Methods 0.000 claims abstract description 19
- 230000011218 segmentation Effects 0.000 claims abstract description 12
- 230000009466 transformation Effects 0.000 claims abstract description 9
- 238000003709 image segmentation Methods 0.000 claims abstract description 8
- 238000012545 processing Methods 0.000 claims abstract description 7
- 238000000513 principal component analysis Methods 0.000 claims abstract description 6
- 238000011084 recovery Methods 0.000 claims abstract description 5
- 230000008569 process Effects 0.000 claims description 21
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 5
- 238000012804 iterative process Methods 0.000 claims description 2
- 230000017105 transposition Effects 0.000 claims description 2
- 238000011426 transformation method Methods 0.000 claims 1
- 238000012360 testing method Methods 0.000 description 8
- 238000005070 sampling Methods 0.000 description 6
- 238000004458 analytical method Methods 0.000 description 4
- 238000000354 decomposition reaction Methods 0.000 description 4
- 238000003384 imaging method Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 102000002274 Matrix Metalloproteinases Human genes 0.000 description 1
- 108010000684 Matrix Metalloproteinases Proteins 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
Classifications
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
Abstract
The present invention proposes the small target detecting method restored under a kind of IR Scene based on grab sample and sparse matrix.Grab sample is carried out to location of pixels each in Single Infrared Image Frame, obtains the infrared image with stochastic behaviour；Patch transformation is carried out to the infrared image after grab sample, by the infrared Image Segmentation after grab sample is the small image of multiple not overlapping regions, and carries out one-dimensional vector processing, obtains the transformed two-dimensional matrix of patch；Principal component analysis is carried out to the transformed two-dimensional matrix of patch, obtains sparse matrix and low-rank matrix；Image recovery is carried out using the method for patch inverse transformation to step sparse matrix, obtains corresponding infrared small target image and infrared image background respectively；The segmentation threshold in infrared small target detection is determined using low-rank matrix, and image segmentation is carried out to infrared small target image according to the segmentation threshold, detects infrared small target.The method of the present invention is simple, operation time is short.
Description
Technical field
The invention belongs to image detections and processing technology field, and in particular under a kind of IR Scene based on grab sample and
The small target detecting method that sparse matrix restores.
Background technique
Small target deteection technology under IR Scene, is broadly divided into two classes, and one kind is to utilize infrared small mesh in single-frame images
Target imaging characteristic is analyzed, one kind be using one group of infrared image sequence, according to the motion conditions of infrared small target, in conjunction with
The mode of track association is analyzed.The detection technique of Single Infrared Image Frame Small Target is concentrated mainly on to target imaging at present
In the research of characteristic, and the target for meeting characteristic in infrared image is marked according to imaging characteristic, finally obtains infrared mesh
Mark.For combining the low-rank specificity analysis of infrared background less, especially in the method that sparse matrix before restores, interception
It obtains small image to need there are overlapping region, for guaranteeing that correlation is stronger in algorithm, but can increase in target detection process
The calculation amount of computation system, and cannot overcome well in the detection noise jamming bring influence, need to infrared image into
The pretreated work of row, there are biggish time loss, in addition, for carrying out infrared small target detection using track association mode
Method in, main thought is single frame detection, is screened by the method for the track association of interframe to single frames target, to obtain
The testing result for obtaining Small object, can be improved the accuracy of algorithm, but there are the consumption of algorithm calculating process to a certain extent
Time is longer, the higher problem of system complexity.
Summary of the invention
The present invention proposes the small target detecting method restored under a kind of IR Scene based on grab sample and sparse matrix, side
Method is simple, operation time is short.
In order to solve the above technical problem, the present invention provides extensive based on grab sample and sparse matrix under a kind of IR Scene
Multiple small target detecting method, steps are as follows:
Grab sample is carried out to location of pixels each in Single Infrared Image Frame, obtains the infrared image with stochastic behaviour；
Patch transformation is carried out to the infrared image after grab sample, the infrared Image Segmentation after grab sample is not had to be multiple
There is the small image of overlapping region, and carry out one-dimensional vector processing, obtains the transformed two-dimensional matrix of patch；
Principal component analysis is carried out to the transformed two-dimensional matrix of patch, obtains sparse matrix and low-rank matrix；
Image recovery is carried out using the method for patch inverse transformation to step sparse matrix, obtains corresponding infrared small mesh respectively
Logo image and infrared image background；
The segmentation threshold in infrared small target detection is determined using low-rank matrix, according to the segmentation threshold to infrared small mesh
Logo image carries out image segmentation, detects infrared small target.
Compared with prior art, the present invention its remarkable advantage is: (1) the present invention is based on sparse matrix recoveries to come to infrared
Small object under scene in single-frame images is detected, and using sparse characteristic of the infrared small target in entire IR Scene, is mentioned
The high accuracy of detection；(2) it is made an uproar using the method for neighborhood territory pixel grab sample to present in infrared image in the present invention
Sound is inhibited, and to simplify, there are overlapping regions between common small image in traditional detection method by this method
Method reduces operation time；(3) present invention obtains Target Segmentation threshold value using background information, uses threshold adaptive
Method, ensure that the versatility of method.
Detailed description of the invention
Fig. 1 is the method for the present invention flow chart.
Fig. 2 is Single Infrared Image Frame to be processed.
Fig. 3 is the Single Infrared Image Frame carried out after grab sample.
Fig. 4 be in inventive method carry out sparse matrix analysis and low-rank matrix analysis schematic diagram, wherein (a) indicate to
Single Infrared Image Frame after machine sampling carries out patch transformation, and carries out the partial results intercepted after one-dimensional；(b) it indicates to carry out odd
The part sparse matrix result intercepted after different value resolution process；(c) the part low-rank matrix of singular value decomposition processing interception is indicated
As a result.
Fig. 5 is the background in the infrared image obtained after being restored low-rank matrix using the method for the present invention.
Fig. 6 is to be carried out using the method for the present invention and other several pairs of methods to other testing result, wherein (a) indicates this hair
The testing result of bright method (b) indicates the testing result obtained using BHP method, and (c) expression is obtained using TDLMS method
Testing result (d) indicates the testing result obtained using TopHat method.
Specific embodiment
It is readily appreciated that, technical solution according to the present invention, in the case where not changing connotation of the invention, this field
Those skilled in the art can imagine the Small object inspection restored under IR Scene of the present invention based on grab sample and sparse matrix
The numerous embodiments of survey method.Therefore, following specific embodiments and attached drawing are only the examples to technical solution of the present invention
Property explanation, and be not to be construed as whole of the invention or be considered as the limitation or restriction to technical solution of the present invention.
In conjunction with Fig. 1, the small target detecting method restored under IR Scene based on grab sample and sparse matrix is red by single frames
Outer image array is divided into background parts and target part, using the direct linear dependence of background gray levels in infrared image, obtains
The low-rank matrix of background must be corresponded to, while using the saltant type relative to background gray levels of infrared small target, utilizing singular value
The method of decomposition obtains the corresponding sparse matrix of Small object, and using the gray threshold obtained in low-rank matrix to sparse matrix into
Row screening, after restoring to the sparse matrix after screening, completes the detection of infrared small target.Detailed process is as follows:
Step 1, the original infrared image I that infrared detector is obtainedsStochastical sampling is carried out, obtaining has stochastic behaviour
Infrared image I.
Random sampling procedure is to infrared image IsIn each location of pixels (x, y) carry out stochastical sampling, to including neighborhood BN
(x, y) and pixel value IsItself the region (x, y) carries out equiprobability random selection, is represented using the result of stochastical sampling current
The gray value I (x, y) of location of pixels, the gray value I (x, y) after stochastical sampling is such as shown in formula (1):
I (x, y)=Rand (Is(x,y),BN(x,y)) (1)
Wherein, BN(x, y) indicates N neighborhood, such as takes 8, in order to guarantee the consistency in treatment process, to original infrared figure
As IsMarginal position deleted there is no the location of pixels of 8 neighborhoods.
Step 2 carries out patch transformation to the infrared image I with stochastic behaviour, infrared image I is divided into and multiple is not had
There is the small image of overlapping region, and carry out one-dimensional vector processing, obtains the transformed two-dimensional matrix I of patchP。
Infrared image I after grab sample is mainly transformed to small image of the M width having a size of W × H by patch transformation, is gone forward side by side
Row one-dimensional handles to obtain the two-dimensional matrix of (W × H) × M, uses IPTo indicate the two-dimensional matrix.For convenience of description, take m=W ×
H, n=M respectively indicate infrared image two-dimensional matrix IPWidth and height, then haveIndicate the real number of m row n column
Matrix.
Step 3, two-dimensional matrix I transformed to patchPThe analysis of sparse matrix and low-rank matrix is carried out, is obtained sparse
Matrix and low-rank matrix.
Present invention assumes that infrared image is made of the strong Small object of the higher background parts of correlation and sparsity, in this way
Small target deteection can be converted to by infrared image two-dimensional matrix IPIt carries out principal component analysis and obtains low-rank matrix,
The objective function for then carrying out principal component analysis can be expressed with formula (2):
In formula (2), λ indicates weight parameter, for the complexity of reduction method, | | | |0Indicate 0 norm sign, S table
Show sparse matrix, noise information has been further comprised in S, L indicates low-rank matrix, due to being not 0 in 0 norm representing matrix element
Number, then | | S | |0=# { i, j:aij≠ 0 }, wherein # indicates a numerical symbol, i.e. formula (2) is expressed as obtaining infrared image I
The smallest sparse matrix.Due to rank () and | | | |0There are nonconvex properties and Non-smooth surface characteristic in optimization, so target letter
Number can be with corresponding conversion for a loose convex optimization problem, as shown in formula (3):
In above formula | | | |*Indicate nuclear norm, the sum of all singular values in representing matrix, can for low-rank matrix
With the size for representing matrix order, it is contemplated that low-rank matrix L is made of the low-rank space of multiple linear correlations, then can be used
Matrix A indicates linear space, indicates low-rank matrix L using the form that a parameter matrix Z is multiplied:
L=AZ (4)
Using the conclusion of formula (4), in the detection process of Small object, in infrared image number shared by object pixel compared with
It is few, two-dimensional matrix I can be usedPAs linear space, then L=I can be obtainedpZ, and the nuclear norm of parameter matrix Z can be with
Indicate low-rank as a result, then objective function can further indicate that are as follows:
In order to introduce a transition variable J using lagrange's method of multipliers solution formula (5) described objective function,
And J=Z, then objective function can be converted further and is expressed as:
Using lagrange's method of multipliers, final goal function can be indicated with formula (7) are as follows:
In formula (7), Y1, Y2Indicate Lagrange's multiplier, tr [Y1 T(Ip-IpZ-S)] andIt respectively indicates
Matrix Y1 T(Ip-Ip) and matrix Z-SMark, μ > 0 indicates penalty parameter, and objective function F can pass through the side of iteration
Method is solved, and is first determined to the initial value of parameters, and the initial value selected in the present invention is Z=0, S=0, Y1=Y2
=0, μ=10-6, μmax=108, ρ=1.5, iteration result is sparse matrix S and low-rank matrix L,1 norm of representing matrix S.Indicate Frobenius norm
Square, T expression takes transposition to operate in matrix.
The specific calculating process of low-rank matrix S and sparse matrix L is iteratively solved such as using objective function shown in formula (7)
Under:
The 3.1 fixed other parameters in addition to J, are updated J, shown in renewal process such as formula (8):
Formula (8) can be solved using theorem, correspondence theorem are as follows: for matrixIn the case where μ > 0,
ForSolution can be carried out by the way of the singular value decomposition as shown in formula (9):
SVTμ(Y)=Udiag [(σ-μ)+]VT (9)
In formula (9),And it can be by square
Battle array Y carries out singular value decomposition and obtains σ, i.e. Y=U Σ VT, wherein Σ=diag (σ), then available
The 3.2 fixed other parameters in addition to S, are updated S, shown in renewal process such as formula (10):
In formula (10),It is that 1 norm is taken to operate after taking 2 norms to matrix S, [S]ijIt indicates
The element of the i-th row jth column of matrix S.Formula (10) can be solved using theorem, correspondence theorem are as follows: in given matrix Q=
[q1,q2,...,qi...] in the case where, forOptimum results be W*, then W*Corresponding
I column result can be indicated with formula (11) are as follows:
The 3.3 fixed other parameters in addition to Z, are updated Z, shown in renewal process such as formula (12):
In formula (12), E indicates unit matrix.
3.4 pairs of Lagrange's multipliers are updated, shown in renewal process such as formula (13) and formula (14):
Y1=Y1+μ(Ip-IpZ-E) (13)
Y2=Y2+μ(Z-J) (14)
3.5 couples of penalty parameter μ are updated, shown in renewal process such as formula (15):
μ=min (ρ μ, μmax) (15)
In formula (15), ρ indicates the decay factor in iterative process, μmaxIndicate the max-thresholds of penalty parameter μ.
Five calculating process of 3.6 couples of above-mentioned 3.1-5.5 are iterated operation, until residual shown in formula (16) and (17)
Poor judgment formula stops interative computation when setting up, and exports sparse matrix S and low-rank matrix L that last iteration obtains,
||Ip-IpZ-S||∞< ε (16)
||Z-J||∞< ε (17)
Threshold residual value ε=the 10- selected in formula (16) and (17)4, | | | |∞Indicate Infinite Norm.
Step 4 determines the segmentation threshold R in infrared small target detection using the low-rank matrix L that step 3 obtains, specifically
Shown in method such as formula (18):
R=α max { L } (18)
α expression parameter amplifies constant in formula (18), such as α=1.2, max { } is selected to indicate to choose in matrix most
Big element.
Step 5 carries out image recovery using the method for patch inverse transformation to step sparse matrix S, obtains respectively corresponding
Infrared small target image ITWith infrared image background IB, in result figure 5 shown in (a) and (b).
Step 6, according to segmentation threshold R, to infrared small target image ITImage segmentation is carried out, infrared small target, tool are obtained
Shown in body method such as formula (19):
T (x, y) is for small target deteection under finally obtained IR Scene as a result, as shown in attached drawing 6 (a).
In order to illustrate advantage of the method for the present invention in calculating speed, testing result, most using the method for the present invention and two dimension
Small side's averaging method (TDLMS), top cap algorithm (TopHat), high-pass filtering (BHP) are compared, and each method is to identical infrared figure
As carrying out small target deteection, as shown in Figure 6, it is shown that the testing result of each method, and the time used in each method is carried out
It analyzes, the time used in TDLMS processing detection Small object is 2.969824s, and TopHat algorithm process detects the time used in Small object
For 0.348399s, BHP algorithm detects the time used in Small object as 0.362685s, and used in the small target deteection of the method for the present invention
Time is 1.2431185s.As can be seen that the calculating speed of the method for the present invention is doubled relative to TDLMS detection method
It is more, it increased relative to TopHat, BHP detection method time-consuming, but be significantly improved in detection accuracy.
Claims (4)
1. the small target detecting method restored under IR Scene based on grab sample and sparse matrix, which is characterized in that step is such as
Under:
Grab sample is carried out to location of pixels each in Single Infrared Image Frame, obtains the infrared image with stochastic behaviour；
Patch transformation is carried out to the infrared image after grab sample, is multiple without weight by the infrared Image Segmentation after grab sample
The small image in folded region, and one-dimensional vector processing is carried out, obtain the transformed two-dimensional matrix of patch；
Principal component analysis is carried out to the transformed two-dimensional matrix of patch, obtains sparse matrix and low-rank matrix；
To sparse matrix using patch inverse transformation method carry out image recovery, obtain respectively corresponding infrared small target image and
Infrared image background；
The segmentation threshold in infrared small target detection is determined using low-rank matrix, according to the segmentation threshold to infrared small target figure
As carrying out image segmentation, infrared small target is detected；
When carrying out principal component analysis acquisition sparse matrix and low-rank matrix to the transformed two-dimensional matrix of patch, use formula (1)
Shown objective function F iteratively solves out low-rank matrix and sparse matrix,
In formula (1), Y1, Y2Indicate Lagrange's multiplier, tr [Y1 T(Ip-IpZ-S)] andRespectively indicate matrix
Y1 T(Ip-Ip) and matrix Z-SMark, μ > 0 indicate penalty parameter, S is sparse matrix, and L is low-rank matrix, and J was
It crosses variable and J=Z, Z is parameter matrix, IPFor the transformed two-dimensional matrix of patch and L=IpZ, λ indicate weight parameter, | | |
|*Indicate nuclear norm, | | S | |1Indicate 1 norm of sparse matrix S,Indicate that square of Frobenius norm, T are indicated to phase
Answer matrix that transposition is taken to operate.
2. small target detecting method as described in claim 1, which is characterized in that iteratively solve low-rank matrix using objective function F
With the process of sparse matrix are as follows:
The 2.1 fixed other parameters in addition to J, are updated J, shown in renewal process such as formula (2):
The 2.2 fixed other parameters in addition to S, are updated S, shown in renewal process such as formula (3):
In formula (3), | | S | |2,1It indicates to take 1 norm to operate after taking sparse matrix S 2 norms；
The 2.3 fixed other parameters in addition to Z, are updated Z, shown in renewal process such as formula (4):
In formula (4), E indicates unit matrix；
2.4 pairs of Lagrange's multipliers are updated, shown in renewal process such as formula (5) and formula (6):
Y1=Y1+μ(Ip-IpZ-E) (5)
Y2=Y2+μ(Z-J) (6)
2.5 couples of penalty parameter μ are updated, shown in renewal process such as formula (7):
μ=min (ρ μ, μmax) (7)
In formula (7), ρ indicates the decay factor in iterative process, μmaxIndicate the max-thresholds of penalty parameter μ；
Five calculating process of 2.6 couples of above-mentioned 2.1-2.5 are iterated operation, until the judgement of residual error shown in formula (8) and (9)
Formula stops interative computation when setting up, and exports sparse matrix S and low-rank matrix L that last iteration obtains,
||Ip-IpZ-S||∞< ε (8)
||Z-J||∞< ε (9)
In formula (8) and (9), ε is threshold residual value.
3. small target detecting method as claimed in claim 2, which is characterized in that determine infrared small target detection using low-rank matrix
In segmentation threshold method such as formula (10) shown in:
R=α max { L } (10)
In formula (10), R is segmentation threshold, and α expression parameter amplifies constant, and max { } indicates to choose the greatest member in matrix.
4. small target detecting method as claimed in claim 3, which is characterized in that according to segmentation threshold R to infrared small target image IT
Image segmentation is carried out to obtain shown in the method such as formula (11) of infrared small target T (x, y):
T (x, y) is the Small object finally detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610722917.1A CN106447668B (en) | 2016-08-25 | 2016-08-25 | The small target detecting method restored under IR Scene based on grab sample and sparse matrix |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610722917.1A CN106447668B (en) | 2016-08-25 | 2016-08-25 | The small target detecting method restored under IR Scene based on grab sample and sparse matrix |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106447668A CN106447668A (en) | 2017-02-22 |
| CN106447668B true CN106447668B (en) | 2019-06-25 |
Family
ID=58182245
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610722917.1A Expired - Fee Related CN106447668B (en) | 2016-08-25 | 2016-08-25 | The small target detecting method restored under IR Scene based on grab sample and sparse matrix |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106447668B (en) |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107274412A (en) * | 2017-05-18 | 2017-10-20 | 北京环境特性研究所 | The method of small target deteection based on infrared image |
| CN107301394A (en) * | 2017-06-21 | 2017-10-27 | 哈尔滨工业大学深圳研究生院 | A kind of people stream detecting method based on video data |
| CN108062523B (en) * | 2017-12-13 | 2021-10-26 | 苏州长风航空电子有限公司 | Infrared far-small target detection method |
| CN108986060B (en) * | 2018-06-25 | 2021-09-28 | 南京大学 | Multi-image reflected light suppression method based on sparse and low-rank matrix decomposition |
| CN109584303B (en) * | 2018-12-03 | 2023-04-14 | 电子科技大学 | Infrared weak and small target detection method based on Lp norm and nuclear norm |
| CN109948605B (en) * | 2019-02-15 | 2021-05-28 | 北京中电普华信息技术有限公司 | Picture enhancement method and device for small target |
| CN111862007B (en) * | 2020-07-02 | 2021-01-08 | 哈尔滨市科佳通用机电股份有限公司 | Freight car brake cylinder inclination detection method based on variance region segmentation |
| CN113034465B (en) * | 2021-03-23 | 2023-09-01 | 国网湖南省电力有限公司 | Power equipment thermal fault monitoring method, device and medium based on infrared image |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105374026A (en) * | 2015-10-09 | 2016-03-02 | 上海海事大学 | A maritime infrared small target detection method suitable for coastal defense monitoring |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9679387B2 (en) * | 2015-02-12 | 2017-06-13 | Mitsubishi Electric Research Laboratories, Inc. | Depth-weighted group-wise principal component analysis for video foreground/background separation |
-
2016
- 2016-08-25 CN CN201610722917.1A patent/CN106447668B/en not_active Expired - Fee Related
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105374026A (en) * | 2015-10-09 | 2016-03-02 | 上海海事大学 | A maritime infrared small target detection method suitable for coastal defense monitoring |
Non-Patent Citations (2)
| Title |
|---|
| Adaptive detection method of infrared small target based on target-background separation via robust principal component analysis;Chuanyun Wang等;《Infrared Physics & Technology》;20150331;第69卷;123-135 |
| 高光谱遥感影像亚像元小目标探测研究;杜博;《中国博士学位论文全文数据库 信息科技辑》;20101015;正文第17-18,110-112页 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106447668A (en) | 2017-02-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106447668B (en) | The small target detecting method restored under IR Scene based on grab sample and sparse matrix | |
| CN107680116B (en) | Method for monitoring moving target in video image | |
| CN105513026B (en) | One kind being based on the non local similar compressed sensing reconstructing method of image | |
| Bi et al. | SAR image change detection using regularized dictionary learning and fuzzy clustering | |
| CN111369487B (en) | Hyperspectral and multispectral image fusion method, system and medium | |
| Zisselman et al. | Compressed learning for image classification: A deep neural network approach | |
| CN109102477A (en) | A kind of high-spectrum remote sensing restoration methods based on the constraint of non-convex low-rank sparse | |
| CN111145131A (en) | Infrared and visible light image fusion method based on multi-scale generation type countermeasure network | |
| CN105931264B (en) | A kind of sea infrared small target detection method | |
| CN107730482B (en) | Sparse fusion method based on regional energy and variance | |
| CN103473740A (en) | Sparse representation and low-rank double restraints-based nonlocal denoising method | |
| Zhang et al. | Reweighted laplace prior based hyperspectral compressive sensing for unknown sparsity | |
| Gu et al. | SAR image super-resolution based on noise-free generative adversarial network | |
| CN107705342A (en) | A kind of infrared image reconstructing method based on self-adapting generalized orthogonal matching pursuit | |
| CN110135344A (en) | Based on the method for detecting infrared puniness target for weighting fixed order expression | |
| Zhao et al. | An adaptation of CNN for small target detection in the infrared | |
| CN104820967B (en) | In-orbit calculating imaging method | |
| CN113034371B (en) | Infrared and visible light image fusion method based on feature embedding | |
| CN106157240A (en) | Remote sensing image super resolution method based on dictionary learning | |
| Cai et al. | Anisotropic spatial–spectral total variation regularized double low-rank approximation for HSI denoising and destriping | |
| Shi et al. | Coded diffraction imaging via double sparse regularization model | |
| Lu et al. | An improved non-subsampled contourlet transform-based hybrid pan-sharpening algorithm | |
| Nirmala et al. | A novel multimodal image fusion method using shift invariant discrete wavelet transform and support vector machines | |
| Sulochana et al. | Denoising and dimensionality reduction of hyperspectral images using framelet transform with different shrinkage functions | |
| CN116051444A (en) | Effective infrared and visible light image self-adaptive fusion method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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 | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190625 Termination date: 20210825 |