CN109325446A - A kind of method for detecting infrared puniness target based on weighting truncation nuclear norm - Google Patents
A kind of method for detecting infrared puniness target based on weighting truncation nuclear norm Download PDFInfo
- Publication number
- CN109325446A CN109325446A CN201811097652.6A CN201811097652A CN109325446A CN 109325446 A CN109325446 A CN 109325446A CN 201811097652 A CN201811097652 A CN 201811097652A CN 109325446 A CN109325446 A CN 109325446A
- Authority
- CN
- China
- Prior art keywords
- image
- weighting
- matrix
- infrared
- target
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
- Aiming, Guidance, Guns With A Light Source, Armor, Camouflage, And Targets (AREA)
Abstract
The invention discloses a kind of method for detecting infrared puniness target based on weighting truncation nuclear norm, method and steps are as follows: obtain original infrared image D, and construct infrared block of imageIn infrared block of image of buildingOn the basis of, in conjunction with the l of the truncation nuclear norm and weighting of weighting1Norm constructs objective function, solves objective function and obtains optimal low-rank matrix and sparse matrix, respectively corresponds as background block imageWith object block imageAccording to obtained background block imageWith object block imageReconstructed background image B and target image T;Adaptive threshold fuzziness is carried out to obtained target image T, determines the position of target, output test result.Small IR targets detection problem is converted objective function Solve problems by the present invention, adaptively isolate target and background, the truncation nuclear norm of weighting is applied in small IR targets detection problem for the first time, the value of nuclear norm is truncated by different weight constraints, can efficiently and accurately detect Weak target.
Description
Technical field
The invention belongs to infrared image processing and target detection technique fields, and in particular to one kind is based on weighting truncation core model
Several method for detecting infrared puniness target.
Background technique
Infrared imagery technique has the characteristics that untouchable, capture details ability is strong, can also be achieved continuous probe round the clock,
Remote target is detected, and is not influenced by barriers such as cigarette, mists, therefore, can realize target using infrared imagery technique
Detection and identification.Infrared search and tracking (Infrared search and track, IRST) system is in military, civilian etc.
Field is used widely, and military value is high, wherein small IR targets detection technology is basic as one of IRST system
Function is of great significance in infrared reconnaissance, infrared early warning, distant object detection.But due in infrared band,
The texture of target, structural information lack, and along with remote, complex background, various clutters etc. influence, infrared target is often in spot
Point is dotted, or even floods in the background, and it is extremely difficult that this has resulted in small IR targets detection, designs a kind of accuracy rate
Small IR targets detection algorithm high, positioning is quasi-, robustness is high, can contribute to infrared target detection and tracking system has
Effect property.
Small IR targets detection technology can be divided into two major classes on the whole: the small and weak mesh based on single frames and based on multiframe
Detection technique is marked, but captures the motion profile of target since the detection technique based on multiframe needs joint multiframe, excludes to make an uproar
The interference of sound, it is therefore desirable to very big calculation amount and amount of storage, it is very high to hardware requirement, using seldom in Practical Project.Currently,
Commonly the detection method based on single frames can be roughly divided into following three classes:
(1) background inhibits.Background inhibit class method based in infrared image background uniformity it is assumed that using filter pair
The background of infrared image is predicted, then subtracts background from original image again, is carried out Threshold segmentation finally with this and is detected small and weak mesh
Mark.Maximum intermediate value (Max-Median) filtering, Largest Mean (Max-Mean) filtering, top cap (Top-Hat) transformation, two dimension are minimum
Square (TDLMS) filtering etc. belongs to the scope of background inhibition.Although such methods are realized simply, due to noise and it is not inconsistent
Close consistency it is assumed that the method that background inhibits easily is influenced by noise clutter, therefore for the red of most of low signal-to-noise ratio
For outer image, poor effect;
(2) vision significance.Human visual system (Human Visual System, HVS) refers at least to contrast, view
Feel and pay attention to and three kinds of mechanism of eye movement, this kind of methods are related to most being contrast mechanisms, i.e., in hypothesis infrared image, target is
Most significant object.For example, Difference of Gaussian filter (DoG) calculates Saliency maps using two different Gaussian filters, and
Target is detected and is identified.Method (LCM) based on local contrast is high using the small neighbourhood local contrast comprising target,
And the low feature of the background area local contrast for the target not included can protrude mesh by calculating local contrast figure
Mark inhibits background, achievees the purpose that detection.Later, many new methods are derived by DoG and LCM again, it is more such as square
It is also introduced into wherein to features such as, entropys.When infrared image, which meets vision significance, to be assumed, the available excellent effect of such methods
Fruit, still, under practical application scene, this hypothesis is difficult to meet, and causes accuracy rate low, the sources for false alarms with conspicuousness
In the presence of, erroneous detection problem is difficult to overcome;
(3) target background separates.What this kind of methods utilized is the non local autocorrelation and mesh of infrared image background
Target detection problems are converted to optimization problem by target sparsity.And under this major class, and can be subdivided into based on super complete word
Allusion quotation, the method for low-rank representation and the method restored based on low-rank background and sparse target.Based on super complete dictionary, low-rank representation
Method (including being based on low-rank representation LRR, the detection method etc. based on low-rank and rarefaction representation LRSR) needs strong by Gauss in advance
The super complete dictionary of Construction of A Model different target size and shape is spent, the process for constructing target dictionary is cumbersome, and testing result is by word
Allusion quotation influences big, and if Gaussian intensity model will be no longer applicable in when target size and larger change in shape.Based on low-rank background with
The method that sparse target is restored is by block iconic model (Patch-Image Model, PIM), the original block of available low-rank
Image, then by the characteristic of target sparse, by optimization object function, while recovering background and target image, finally obtain
Testing result.Many emerging in large numbers for method demonstrate the superiority of such methods.But due to strong edge, partial noise, false-alarm
Source also has the characteristics that sparse, they can reduce the accuracy rate of detection.It is restored the present invention is based on low-rank background and sparse target
Method proposes a kind of based on the infrared small and weak of weighting truncation nuclear norm in the research direction for improving Detection accuracy and high efficiency
Object detection method.
Summary of the invention
It is an object of the invention to: the prior art is solved in complicated infrared background, noise jamming, low signal-to-noise ratio and low right
In the case where degree, infrared small object is difficult to accurately be detected, and the problem that Detection accuracy is low and efficiency is not high enough proposes
A kind of method for detecting infrared puniness target based on weighting truncation nuclear norm.
The technical solution adopted by the invention is as follows:
A kind of method for detecting infrared puniness target based on weighting truncation nuclear norm, method include the following steps:
Step 1 obtains original infrared image D, and constructs infrared block of image
Step 2, the infrared block of image in buildingOn the basis of, in conjunction with the l of the truncation nuclear norm and weighting of weighting1Norm,
Construct objective function:
Wherein λ and β indicates that coefficient of balance, L represent low-rank ingredient, and S represents sparse ingredient, and N represents noise contribution, | | |
|W, rThe truncation nuclear norm of weighting is represented, i.e.,σi(L) i-th of singular value of L is indicated, r represents matrix
Order, w=[w1, w2..., wMin (m, n)]TRepresent non-negative weighting coefficient, the element w in w used herein above1, w2..., wr
It is 1, | | | |W, 1Represent the l of weighting1Norm, i.e.,W∈Rm×nNon-negative weighting coefficient is represented, | |
||lAny one norm is represented, can be changed according to actual needs;
It solves objective function and obtains optimal low-rank matrix and sparse matrix, respectively correspond as background block imageAnd target
Block image
Step 3, the background block image obtained according to step 2With object block imageReconstructed background image B and target figure
As T;
Step 4 carries out adaptive threshold fuzziness to the target image T that step 3 obtains, and determines the position of target, output inspection
Survey result.
Further, infrared block of image is constructed in the step 1Specifically: use size for the sliding window of k × k, from
It is left-to-right, from top to bottom, original image D is traversed with the step-length of s, and the matrix of k × k size acquired in each sliding window
Vector turns to k2× 1 column vector, after the completion of traversal, all obtained Column vector groups are at a new matrix to get to infrared
Block imageWherein t is the number of window sliding.
Further, further include in the step 2 by objective function carry out Lagrange's equation conversion, and according to optimization after
Equation solution obtains optimal low-rank matrix and sparse matrix, respectively corresponds to obtain background block imageWith object block imageTool
Body step are as follows:
The augmentation Lagrange's equation of step 2.1, the objective function constructed are as follows:
Wherein, Y indicates Lagrange multiplier, and μ indicates non-negative penalty factor, and<>indicates inner product operation, | | | |FTable
Show Frobenius norm, i.e.,
Step 2.2, in the formula of step 2.1,A∈Rr×m, B ∈ Rn ×r, I is a unit matrix, | | L | |W, *The nuclear norm of the weighting of representing matrix, i.e.,σi(L) it indicates
I-th of singular value of L, r represent rank of matrix, w=[w1, w2..., wMin (m, n)]TRepresent non-negative weight, w used herein above
In element w1, w2..., wrIt is the mark of 1, tr representing matrix, the augmentation Lagrange's equation that step 2.1 is obtained is rewritten such as
Under:
Step 2.3, infrared block of imageThe Lagrange's equation that input step 2.2 obtains solves Lagrange's equation,
And export background block imageWith object block image
Further, solution Lagrange's equation uses and is folded direction multiplier method, specific solution procedure in the step 2.3
Are as follows:
Step 2.3.1, L, S, N, Y 0, the number of iterations k=0 are initialized, maximum number of iterations maxk initializes μ >
0, ρ > 1,β > 0, w=1, W=1 × 1T, εB>=1, εT> 0;
Step 2.3.2, fixed L, N, Y, update Sk+1It is as follows:
Wherein, SτIt (x) is soft-threshold contraction operator, Sτ(x)=sgn (x) max (| x |-τ, 0);
Step 2.3.3, to LkCarry out singular value decomposition: [Uk, ∑k, Vk]=svd (Lk), it obtains Thus it obtains
Step 2.3.4, w is enabled1=w2=...=wr=1, enable tem=X-Sk+1-Nk+(Yk-(Ak)TBk/ μ), tem is carried out
Singular value decomposition, i.e. [Ut, ∑t, Vt]=svd (tem);
Step 2.3.5, fixed S, N, Y, update Lk+1It is as follows:
Wherein, diag indicates that diagonal element is max (∑ii-wi/ μ) diagonal matrix;
Step 2.3.6, fixed L, S, Y, update Nk+1It is as follows:
Step 2.3.7, fixed L, S, N, update Yk+1It is as follows:
Yk+1=Yk+μ(X-Lk+1-Sk+1-Nk+1);
Step 2.3.8, weight w is updatedk+1, Wk+1:
Wherein, εB, εTIndicate positive constant;
Step 2.3.9, the number of iterations k=k+1 is updated;
Step 2.3.10, judge whether k is greater than maxk, if it is, stopping iteration, go to step (2.3.11);If
It is not judgementIt is whether true, if so, stopping iteration, step (2.3.11) is gone to,
If it is not, going to step (2.3.2);
Step 2.3.11, the final optimal low-rank matrix L of output*With sparse matrix S*, respectively correspond as background block imageWith object block image
Further, according to background block image in the step 3KnowTake that construct infrared block of image with step 1 opposite
Mode,WithSmall images are reconstructed into, further according to sequentially successively reconstructed background image B, for the position being overlapped in each fritter
It sets, takes the mode of median filtering, determine the gray value of the position.
Further, in the step 4, when carrying out adaptive threshold fuzziness to target image T, threshold value Th=m+c* σ, wherein
M indicates the mean value of all gray scales in T, and σ indicates the standard deviation of all gray scales in T, and c indicates the constant between 1-15.
In conclusion by adopting the above-described technical solution, the beneficial effects of the present invention are:
1, it in the present invention, converts small IR targets detection problem to the Solve problems of objective function, adaptively divides
Target and background is separated out, the truncation nuclear norm of weighting is applied in the test problems of infrared small object for the first time, by not
The value of same weight constraints truncation nuclear norm, can efficiently and accurately detect Weak target;
2, in the present invention, using the optimal value for being folded direction multiplier method and solving objective function, more efficiently;
3, in the present invention, objective function introduces noise contribution, has fully considered background clutter, strong edge etc. to small and weak mesh
The influence of mark detection also enhances the robustness of algorithm while improving accuracy rate.
Detailed description of the invention
In order to illustrate the technical solution of the embodiments of the present invention more clearly, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only certain embodiments of the present invention, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 is the method for the present invention flow chart;
Fig. 2 is the infrared image that the width that the present invention enumerates contains Weak target;
Fig. 3 is the block image that the present invention is constructed by Fig. 2;
Fig. 4 is the present invention by Fig. 3 background block image isolated and object block image;
Fig. 5 is the target image and background image that the present invention is restored by Fig. 4;
Fig. 6 obtains testing result through adaptive threshold fuzziness by the target image in Fig. 5 for the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, not
For limiting the present invention, i.e., described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is logical
The component for the embodiment of the present invention being often described and illustrated herein in the accompanying drawings can be arranged and be designed with a variety of different configurations.
Therefore, the detailed description of the embodiment of the present invention provided in the accompanying drawings is not intended to limit below claimed
The scope of the present invention, but be merely representative of selected embodiment of the invention.Based on the embodiment of the present invention, those skilled in the art
Member's every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
It should be noted that the relational terms of term " first " and " second " or the like be used merely to an entity or
Operation is distinguished with another entity or operation, and without necessarily requiring or implying between these entities or operation, there are any
This actual relationship or sequence.Moreover, the terms "include", "comprise" or its any other variant be intended to it is non-exclusive
Property include so that include a series of elements process, method, article or equipment not only include those elements, but also
Further include other elements that are not explicitly listed, or further include for this process, method, article or equipment it is intrinsic
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including described
There is also other identical elements in the process, method, article or equipment of element.
In this text, matrix is indicated with capitalization overstriking letter such as X, element indexing such as Xij, No. * expression of subscript is optimal, such as
L*。
A kind of method for detecting infrared puniness target based on weighting truncation nuclear norm, as shown in Figure 1, comprising the following steps:
Step 1: obtaining an original infrared image D ∈ R to be processedm×n, as shown in Figure 2.
Step 2: use size for the sliding window of k × k, it is from left to right, from top to bottom, original red with the step-length traversal of s
Outer image D, and the matrix-vector of k × k size acquired in each sliding window is turned to k2× 1 column vector, traversal are completed
Afterwards, all obtained Column vector groups arrive infrared block of image at a new matrixWherein t is window sliding
Number, infrared block of image are as shown in Figure 3.
Step 3: in the infrared block of image that step 2 constructsOn the basis of, in conjunction with the truncation nuclear norm of weighting and weighting
L1Norm constructs objective function, optimal low-rank matrix and sparse matrix is solved according to objective function, respectively corresponds as back
Scape block imageWith object block imageAs shown in Figure 4, the specific steps are as follows:
Infrared block of step 3.1, input image
Step 3.2, the l that nuclear norm and weighting is truncated in conjunction with weighting1Norm constructs objective function, the specific steps are as follows:
Step 3.2.1, assume image X ∈ Rm×nIt is made of low-rank ingredient L, sparse ingredient S and noise contribution N, in order to restore
Low-rank ingredient L and sparse ingredient S, then it is as follows can to construct objective function:
Wherein λ and β indicates that coefficient of balance, L represent low-rank ingredient, and S represents sparse ingredient, and N represents noise contribution, | | |
|W, rThe truncation nuclear norm of weighting is represented, i.e.,σi(L) i-th of singular value of L is indicated, r represents matrix
Order, w=[w1, w2..., wMin (m, n)]TRepresent non-negative weighting coefficient, the element w in w used herein above1, w2..., wr
It is 1, | | | |W, 1Represent the l of weighting1Norm, i.e.,W∈Rm×nNon-negative weighting coefficient is represented,Generation
Any one norm of table, can change according to actual needs, due in most cases, noise, especially strong edge ingredient, for
It is sparse for entire image, therefore uses l here1Norm constrains N, i.e. l=1;
Step 3.2.2, the augmentation Lagrange's equation of the objective function constructed are as follows:
Wherein, Y indicates Lagrange multiplier, and μ indicates non-negative penalty factor, and<>indicates inner product operation, | | | |FTable
Show Frobenius norm, i.e.,
Step 3.2.3, in the equation of step 3.2.2,A∈Rr×m, B
∈Rn×r, I is a unit matrix, | | L | |W, *The nuclear norm of the weighting of representing matrix, i.e.,σi(L)
Indicate i-th of singular value of L, r represents rank of matrix, w=[w1, w2..., wMin (m, n)]TNon-negative weight is represented, it is used here
W in element w1, w2..., wrIt is the mark of 1, tr representing matrix, the augmentation Lagrange's equation that step 2.1 obtains is changed
It writes as follows:
Step 3.3, infrared block of imageThe Lagrange's equation that input step 3.2.3 is obtained is multiplied using direction is folded
Sub- method solves Lagrange's equation, and exports background block imageWith object block imageSpecific solution procedure
Are as follows:
Step 3.3.1, L, S, N, Y 0, the number of iterations k=0 are initialized, maximum number of iterations maxk initializes μ >
0, ρ > 1,β > 0, w=1, W=1 × 1T, εB>=1, εT> 0, ρ, εB, εTAlgorithm is used when being all equation solution
Variable;
Step 3.3.2, fixed L, N, Y, update Sk+1It is as follows:
Wherein, SτIt (x) is soft-threshold contraction operator, Sτ(x)=sgn (x) max (| x |-τ, 0);
Step 3.3.3, to LkCarry out singular value decomposition: [Uk, ∑k, Vk]=svd (Lk), it obtains Thus it obtains
Step 3.3.4, w is enabled1=w2=...=wr=1, enable tem=X-Sk+1-Nk+(Yk-(Ak)TBk/ μ), tem is carried out
Singular value decomposition, i.e. [Ut, ∑t, Vt]=svd (tem);
Step 3.3.5, fixed S, N, Y, update Lk+1It is as follows:
Wherein, diag indicates that diagonal element is max (∑ii-wi/ μ) diagonal matrix;
Step 3.3.6, fixed L, S, Y, update Nk+1It is as follows:
Step 3.3.7, fixed L, S, N, update Yk+1It is as follows:
Yk+1=Yk+μ(X-Lk+1-Sk+1-Nk+1);
Step 3.3.8, weight w is updatedk+1, Wk+1:
Wherein, εB, εTIndicate positive constant;
Step 3.3.9, the number of iterations k=k+1 is updated;
Step 3.3.10, judge whether k is greater than maxk, if it is, stopping iteration, go to step (2.3.11);If
It is not judgementIt is whether true, if so, stopping iteration, step (2.3.11) is gone to,
If it is not, going to step (2.3.2);
Step 3.3.11, the final optimal low-rank matrix L of output*With sparse matrix S*, respectively correspond as background block imageWith object block image
Step 4: the background block image obtained according to step 3With object block imageReconstructed background figure
As B ∈ Rm×nWith target image T ∈ Rm×n, as shown in figure 5, specific steps are as follows: for the background block image of inputIt adopts
The mode opposite with step 2 is taken,In t column vector be reconstructed into the small images of t k × k size, further according to sequence according to
Secondary reconstructed background image B takes the mode of median filtering for the position being overlapped in each fritter, determines the gray scale of the position
Value, target image T in the same way byReconstruct.
Step 5: adaptive threshold fuzziness being carried out to the target image T that step 4 obtains, determines the position of target, and export
Testing result, as shown in fig. 6, threshold value Th=m+c* σ, wherein m is indicated in T when carrying out adaptive threshold fuzziness to target image T
The mean value of all gray scales, σ indicate the standard deviation of all gray scales in T, and c indicates the constant between 1-15.
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all in essence of the invention
Made any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention within mind and principle.
Claims (6)
1. a kind of method for detecting infrared puniness target based on weighting truncation nuclear norm, it is characterised in that: method includes following step
It is rapid:
Step 1 obtains original infrared image D, and constructs infrared block of image
Step 2, the infrared block of image in buildingOn the basis of, in conjunction with the l of the truncation nuclear norm and weighting of weighting1Norm, building
Objective function:
Wherein λ and β indicates that coefficient of balance, L represent low-rank ingredient, and S represents sparse ingredient, and N represents noise contribution, | | | |W, rGeneration
The truncation nuclear norm of table weighting, i.e.,σi(L) i-th of singular value of L is indicated, r represents rank of matrix, w
=[w1, w2..., wMin (m, n)]TRepresent non-negative weighting coefficient, the element w in w used herein above1, w2..., wrIt is 1,
||·||W, 1Represent the l of weighting1Norm, i.e.,W∈Rm×nNon-negative weighting coefficient is represented, | | | |lGeneration
Any one norm of table, can change according to actual needs;
It solves objective function and obtains optimal low-rank matrix L*With sparse matrix S*, respectively correspond as background block imageAnd object block
Image
Step 3, the background block image obtained according to step 2With object block imageReconstructed background image B and target image T:
Step 4 carries out adaptive threshold fuzziness to the target image T that step 3 obtains, and determines the position of target, output detection knot
Fruit.
2. a kind of method for detecting infrared puniness target based on weighting truncation nuclear norm according to claim 1, feature
It is: constructs infrared block of image in the step 1Specifically: use size for the sliding window of k × k, from left to right, from upper
It arrives down, original image D is traversed with the step-length of s, and the matrix-vector of k × k size acquired in each sliding window is turned to k2
× 1 column vector, after the completion of traversal, all obtained Column vector groups arrive infrared block of image at a new matrixWherein t is the number of window sliding.
3. a kind of method for detecting infrared puniness target based on weighting truncation nuclear norm according to claim 1, feature
It is: further includes objective function being subjected to Lagrange's equation conversion, and obtain according to the equation solution after optimization in the step 2
To optimal low-rank matrix and sparse matrix, respectively correspond to obtain background block imageWith object block imageSpecific steps are as follows:
The augmentation Lagrange's equation of step 2.1, the objective function constructed are as follows:
Wherein, Y indicates Lagrange multiplier, and μ indicates non-negative penalty factor, and<>indicates inner product operation, | | | |FIt indicates
Frobenius norm, i.e.,
Step 2.2, in the formula of step 2.1,A∈Rr×m, B ∈ Rn×r, I
It is a unit matrix, | | L | |W, *The nuclear norm of the weighting of representing matrix, i.e.,σi(L) indicate L's
I-th of singular value, r represent rank of matrix, w=[w1, w2..., wMin (m, n)]TNon-negative weight is represented, in w used herein above
Element w1, w2..., wrIt is the mark of 1, tr representing matrix, the augmentation Lagrange's equation that step 2.1 is obtained is rewritten as follows:
Step 2.3, infrared block of imageThe Lagrange's equation that input step 2.2 obtains solves Lagrange's equation, and defeated
Background block image outWith object block image
4. a kind of method for detecting infrared puniness target based on weighting truncation nuclear norm according to claim 3, feature
Be: solution Lagrange's equation, which uses, in the step 2.3 is folded direction multiplier method, specific solution procedure are as follows:
Step 2.3.1, L, S, N, Y 0, the number of iterations k=0 are initialized, maximum number of iterations maxk initializes μ > 0, ρ >
1,β > 0, w=1, W=1 × 1T, εB>=1, εT> 0;
Step 2.3.2, fixed L, N, Y, update Sk+1It is as follows:
Wherein, SτIt (x) is soft-threshold contraction operator, Sτ(x)=sgn (x) max (| x |-τ, 0);
Step 2.3.3, to LkCarry out singular value decomposition: [Uk, ∑k, Vk]=svd (Lk), it obtains Thus it obtains
Step 2.3.4, w is enabled1=w2=...=wr=1, enable tem=X-Sk+1-Nk+(Yk-(Ak)TBk/ μ), tem is carried out unusual
Value is decomposed, i.e. [Ut, ∑t, Vt]=svd (tem);
Step 2.3.5, fixed S, N, Y, update Lk+1It is as follows:
Wherein, diag indicates that diagonal element is max (∑ii-wi/ μ) diagonal matrix;
Step 2.3.6, fixed L, S, Y, update Nk+1It is as follows:
Step 2.3.7, fixed L, S, N, update Yk+1It is as follows:
Yk+1=Yk+μ(X-Lk+1-Sk+1-Nk+1);
Step 2.3.8, weight w is updatedk+1, Wk+1:
Wherein, εB, εTIndicate positive constant;
Step 2.3.9, λ is updatedk+1=ρ λk, update the number of iterations k=k+1;
Step 2.3.10, judge whether k is greater than maxk, if it is, stopping iteration, go to step (2.3.11);If not
JudgementIt is whether true, if so, stopping iteration, step (2.3.11) is gone to, if
It is not to go to step (2.3.2):
Step 2.3.11, the final optimal low-rank matrix L of output*With sparse matrix S*, respectively correspond as background block imageWith
Object block image
5. a kind of method for detecting infrared puniness target based on weighting truncation nuclear norm according to claim 1, feature
It is: according to background block image in the step 3WithIt takes and constructs the infrared piece of opposite mode of image with step 1,
KnowSmall images are reconstructed into, further according to sequentially successively reconstructed background image B, for the position being overlapped in each fritter, are taken
The mode of value filtering determines the gray value of the position.
6. a kind of method for detecting infrared puniness target based on weighting truncation nuclear norm according to claim 1, feature
Be: in the step 4, when carrying out adaptive threshold fuzziness to target image T, threshold value Th=m+c* σ, wherein m indicates institute in T
There is the mean value of gray scale, σ indicates the standard deviation of all gray scales in T, and c indicates the constant between 1-15.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811097652.6A CN109325446B (en) | 2018-09-19 | 2018-09-19 | Infrared weak and small target detection method based on weighted truncation nuclear norm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811097652.6A CN109325446B (en) | 2018-09-19 | 2018-09-19 | Infrared weak and small target detection method based on weighted truncation nuclear norm |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109325446A true CN109325446A (en) | 2019-02-12 |
CN109325446B CN109325446B (en) | 2021-06-22 |
Family
ID=65265194
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811097652.6A Active CN109325446B (en) | 2018-09-19 | 2018-09-19 | Infrared weak and small target detection method based on weighted truncation nuclear norm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109325446B (en) |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109934815A (en) * | 2019-03-18 | 2019-06-25 | 电子科技大学 | A kind of tensor recovery method for detecting infrared puniness target of combination ATV constraint |
CN110414598A (en) * | 2019-07-26 | 2019-11-05 | 国家消防工程技术研究中心 | Smog detection method, device, computer and storage medium |
CN110443209A (en) * | 2019-08-08 | 2019-11-12 | 南京邮电大学 | The infrared small target detection method restored based on matrix |
CN110674782A (en) * | 2019-09-30 | 2020-01-10 | 电子科技大学 | Infrared dim target detection method based on fractional entropy |
CN110751667A (en) * | 2019-09-17 | 2020-02-04 | 中国科学院上海技术物理研究所 | Method for detecting infrared dim small target under complex background based on human visual system |
CN110827262A (en) * | 2019-11-06 | 2020-02-21 | 西北工业大学 | Weak and small target detection method based on continuous limited frame infrared image |
CN111739055A (en) * | 2020-06-10 | 2020-10-02 | 新疆大学 | Infrared point-like target tracking method |
CN111967466A (en) * | 2020-07-08 | 2020-11-20 | 北京航空航天大学 | Infrared weak and small target detection method utilizing multidirectional sparse low-rank decomposition |
CN112329764A (en) * | 2020-09-28 | 2021-02-05 | 中国空气动力研究与发展中心计算空气动力研究所 | Infrared dim target detection method based on TV-L1 model |
CN113076659A (en) * | 2021-04-23 | 2021-07-06 | 南京理工大学 | Marine ship detection method and system based on adaptive weight kernel norm approximation and low-rank representation and sparse constraint |
CN117011196A (en) * | 2023-08-10 | 2023-11-07 | 哈尔滨工业大学 | Infrared small target detection method and system based on combined filtering optimization |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102324021A (en) * | 2011-09-05 | 2012-01-18 | 电子科技大学 | Infrared dim-small target detection method based on shear wave conversion |
CN103345729A (en) * | 2013-06-30 | 2013-10-09 | 浙江贝尔技术有限公司 | Image restoration method based on truncation nuclear norm regularization |
CN106056607A (en) * | 2016-05-30 | 2016-10-26 | 天津城建大学 | Monitoring image background modeling method based on robustness principal component analysis |
KR101740647B1 (en) * | 2016-11-04 | 2017-06-08 | 부경대학교 산학협력단 | Apparatus for Processing High Dynamic Range Imaging |
CN107133930A (en) * | 2017-04-30 | 2017-09-05 | 天津大学 | Ranks missing image fill method with rarefaction representation is rebuild based on low-rank matrix |
US20170367590A1 (en) * | 2016-06-24 | 2017-12-28 | Universita' degli Studi di Trento (University of Trento) | Self-adaptive matrix completion for heart rate estimation from face videos under realistic conditions |
-
2018
- 2018-09-19 CN CN201811097652.6A patent/CN109325446B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102324021A (en) * | 2011-09-05 | 2012-01-18 | 电子科技大学 | Infrared dim-small target detection method based on shear wave conversion |
CN103345729A (en) * | 2013-06-30 | 2013-10-09 | 浙江贝尔技术有限公司 | Image restoration method based on truncation nuclear norm regularization |
CN106056607A (en) * | 2016-05-30 | 2016-10-26 | 天津城建大学 | Monitoring image background modeling method based on robustness principal component analysis |
US20170367590A1 (en) * | 2016-06-24 | 2017-12-28 | Universita' degli Studi di Trento (University of Trento) | Self-adaptive matrix completion for heart rate estimation from face videos under realistic conditions |
KR101740647B1 (en) * | 2016-11-04 | 2017-06-08 | 부경대학교 산학협력단 | Apparatus for Processing High Dynamic Range Imaging |
CN107133930A (en) * | 2017-04-30 | 2017-09-05 | 天津大学 | Ranks missing image fill method with rarefaction representation is rebuild based on low-rank matrix |
Non-Patent Citations (1)
Title |
---|
宣晓等: "加权截断p范数在运动目标检测中的应用", 《计算机工程 》 * |
Cited By (18)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109934815A (en) * | 2019-03-18 | 2019-06-25 | 电子科技大学 | A kind of tensor recovery method for detecting infrared puniness target of combination ATV constraint |
CN109934815B (en) * | 2019-03-18 | 2023-04-14 | 电子科技大学 | Tensor recovery infrared small target detection method combined with ATV constraint |
CN110414598A (en) * | 2019-07-26 | 2019-11-05 | 国家消防工程技术研究中心 | Smog detection method, device, computer and storage medium |
CN110443209A (en) * | 2019-08-08 | 2019-11-12 | 南京邮电大学 | The infrared small target detection method restored based on matrix |
CN110443209B (en) * | 2019-08-08 | 2022-07-29 | 南京邮电大学 | Infrared small target detection method based on matrix recovery |
CN110751667A (en) * | 2019-09-17 | 2020-02-04 | 中国科学院上海技术物理研究所 | Method for detecting infrared dim small target under complex background based on human visual system |
CN110751667B (en) * | 2019-09-17 | 2023-09-15 | 中国科学院上海技术物理研究所 | Method for detecting infrared dim and small targets under complex background based on human visual system |
CN110674782A (en) * | 2019-09-30 | 2020-01-10 | 电子科技大学 | Infrared dim target detection method based on fractional entropy |
CN110827262A (en) * | 2019-11-06 | 2020-02-21 | 西北工业大学 | Weak and small target detection method based on continuous limited frame infrared image |
CN110827262B (en) * | 2019-11-06 | 2023-05-16 | 西北工业大学 | Weak and small target detection method based on continuous limited frame infrared image |
CN111739055A (en) * | 2020-06-10 | 2020-10-02 | 新疆大学 | Infrared point-like target tracking method |
CN111739055B (en) * | 2020-06-10 | 2022-07-05 | 新疆大学 | Infrared point-like target tracking method |
CN111967466B (en) * | 2020-07-08 | 2022-08-16 | 北京航空航天大学 | Infrared weak and small target detection method utilizing multidirectional sparse low-rank decomposition |
CN111967466A (en) * | 2020-07-08 | 2020-11-20 | 北京航空航天大学 | Infrared weak and small target detection method utilizing multidirectional sparse low-rank decomposition |
CN112329764A (en) * | 2020-09-28 | 2021-02-05 | 中国空气动力研究与发展中心计算空气动力研究所 | Infrared dim target detection method based on TV-L1 model |
CN113076659A (en) * | 2021-04-23 | 2021-07-06 | 南京理工大学 | Marine ship detection method and system based on adaptive weight kernel norm approximation and low-rank representation and sparse constraint |
CN117011196A (en) * | 2023-08-10 | 2023-11-07 | 哈尔滨工业大学 | Infrared small target detection method and system based on combined filtering optimization |
CN117011196B (en) * | 2023-08-10 | 2024-04-19 | 哈尔滨工业大学 | Infrared small target detection method and system based on combined filtering optimization |
Also Published As
Publication number | Publication date |
---|---|
CN109325446B (en) | 2021-06-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109325446A (en) | A kind of method for detecting infrared puniness target based on weighting truncation nuclear norm | |
Zhao et al. | Three-order tensor creation and tucker decomposition for infrared small-target detection | |
CN109934178A (en) | A kind of method for detecting infrared puniness target based on Kronecker base rarefaction representation | |
CN109447073A (en) | A kind of method for detecting infrared puniness target based on tensor Robust Principal Component Analysis | |
CN110007366B (en) | Life searching method and system based on multi-sensor fusion | |
CN107563433B (en) | Infrared small target detection method based on convolutional neural network | |
CN109215025A (en) | A kind of method for detecting infrared puniness target approaching minimization based on non-convex order | |
CN105488456B (en) | Method for detecting human face based on adaptive threshold adjustment rejection sub-space learning | |
CN107240122A (en) | Video target tracking method based on space and time continuous correlation filtering | |
Alonso-Gonzalez et al. | Processing multidimensional SAR and hyperspectral images with binary partition tree | |
CN109934815A (en) | A kind of tensor recovery method for detecting infrared puniness target of combination ATV constraint | |
CN106886760B (en) | A kind of EO-1 hyperion Ship Detection combined based on empty spectrum information | |
Yu et al. | Railway obstacle detection algorithm using neural network | |
CN107798345B (en) | High-spectrum disguised target detection method based on block diagonal and low-rank representation | |
Li et al. | Deep convolutional neural network based large-scale oil palm tree detection for high-resolution remote sensing images | |
CN109584303A (en) | A kind of method for detecting infrared puniness target based on Lp norm and nuclear norm | |
CN110135344A (en) | Based on the method for detecting infrared puniness target for weighting fixed order expression | |
CN112149591A (en) | SSD-AEFF automatic bridge detection method and system for SAR image | |
CN105405132A (en) | SAR image man-made target detection method based on visual contrast and information entropy | |
Qian et al. | Guided filter and convolutional network based tracking for infrared dim moving target | |
Zhang et al. | Infrared small target detection via interpatch correlation enhancement and joint local visual saliency prior | |
Sun et al. | Image target detection algorithm compression and pruning based on neural network | |
Kekre et al. | SAR image segmentation using vector quantization technique on entropy images | |
McCleary et al. | Sign language recognition using micro-Doppler and explainable deep learning | |
CN104851090A (en) | Image change detection method and image change detection device |
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 |