CN111160181B - Small target detection method based on infrared video image - Google Patents

Small target detection method based on infrared video image Download PDF

Info

Publication number
CN111160181B
CN111160181B CN201911325125.0A CN201911325125A CN111160181B CN 111160181 B CN111160181 B CN 111160181B CN 201911325125 A CN201911325125 A CN 201911325125A CN 111160181 B CN111160181 B CN 111160181B
Authority
CN
China
Prior art keywords
target
background
solving
infrared video
infrared
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.)
Active
Application number
CN201911325125.0A
Other languages
Chinese (zh)
Other versions
CN111160181A (en
Inventor
梁军利
程志伟
刘睿恺
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.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical University
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 Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN201911325125.0A priority Critical patent/CN111160181B/en
Publication of CN111160181A publication Critical patent/CN111160181A/en
Application granted granted Critical
Publication of CN111160181B publication Critical patent/CN111160181B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • 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

Abstract

The invention discloses a small target detection method based on an infrared video image, which comprises the following steps: step 1: inputting each frame of the M frames of infrared video sequences as a column vector to form a data matrix; step 2: establishing a modeling mode by taking a robust matrix approximation theory as a background and additionally establishing an infrared video image target detection model based on a target sparse expression item and an anti-noise interference item; and step 3: solving the detection model by using the ADMM to obtain the optimal solution of the background part and the target part of the infrared video; and 4, step 4: and converting the background part and the target part into a video form and outputting. The small target detection method based on the infrared video image can accurately detect the infrared small target, and can realize accurate and effective extraction of the small target even under the condition that the target is weak to only a few pixels; secondly, the detection result is not limited by the movement condition of the target, and accurate detection can be realized when the target has a repeated track.

Description

Small target detection method based on infrared video image
Technical Field
The invention belongs to the technical field of video image processing methods, and particularly relates to a small target detection method based on an infrared video image.
Background
With the continuous development of Infrared imaging technology, the Infrared small target detection technology is widely applied in the military field and the civil field, on one hand, in the military application field, an all-weather Infrared Search and tracking system (IRST) is significant, and compared with the active combat of radar detection, the Infrared small target detection system not only can be passively detected and has better concealment performance and is not easy to be found by enemies, but also has all-weather detection capability and anti-electromagnetic interference capability, so that the Infrared imaging technology plays an extremely important role in the military field, and for example, one of the core technologies of space-based Infrared early warning is an Infrared detection system. On the other hand, in the civil field, the effect of infrared imaging is not a little different, for example, with the development of scientific technology, the unmanned aerial vehicle raises a new trend of the era, however, the illegal use of the unmanned aerial vehicle also brings threats to national security and safety of people's life, and a powerful and reliable system capable of realizing the detection function is needed; in addition, infrared imaging also makes outstanding contributions in the aspects of safety monitoring, biomedicine and the like of residential quarters.
However, infrared imaging has some disadvantages, and the imaging result is greatly influenced by the radiation intensity of the target and the detection distance, and firstly, as the detection distance increases, the smaller the infrared imaging target is, the smaller the target radiation is, and the weaker the imaging target intensity is. However, in general, the distance between the target and the infrared detector is relatively long, the target information captured by the detector is usually a weak target point with a size of only a few pixels to tens of pixels, and the characteristic information is not obvious (such as texture, shape, etc.), and in addition, the infrared imaging result is relatively easy to be influenced by environmental factors (atmospheric radiation, etc.), the signal-to-noise ratio of the imaging result is often low, which causes the gray value of the target point to be not greatly distinguished from the background and noise parts, which brings great difficulty to the detection of the target. Therefore, the research on the infrared video image-based weak and small target detection technology has great theoretical innovation significance and practical use value.
The target detection is generally performed based on background extraction and foreground separation technologies, that is, a video image is divided into a background part and a foreground part through a certain background modeling strategy, and currently, common target detection methods include a gaussian background modeling method, a frame difference method, a low-rank decomposition method and the like, but the algorithms are still limited by many factors. Firstly, the gaussian background modeling method has a high false alarm rate under a complex background and is easy to generate a ghost phenomenon; the detection result obtained by adopting the frame difference method is greatly influenced by the motion condition of the target, the target which moves slowly or has a repeated motion track is difficult to detect, and a 'void phenomenon' exists; the problems bring great limitations to practical application, for example, when the unmanned aerial vehicle needs to hover at a certain position to fly when executing a reconnaissance task, the target cannot be completely and continuously detected by adopting the method; in addition, when the target is small, the interference noise caused by the environmental factors still has a great influence on the detection result by using the low-rank matrix decomposition method.
Disclosure of Invention
The invention aims to provide a small target detection method based on an infrared video image, which can accurately detect an infrared small target and is not limited by the motion condition of the target.
The technical scheme adopted by the invention is as follows: a small target detection method based on infrared video images comprises the following steps:
step 1: inputting each frame of the M frames of infrared video sequences as a column vector to form a data matrix;
step 2: assuming that the data matrix input in the step 1 consists of three parts, namely a background, a target and interference noise, and establishing an infrared video image target detection model which takes a robust matrix approximation theory as a background modeling mode and is additionally based on a target sparse expression item and an anti-noise interference item based on an array signal processing theory;
and step 3: solving the infrared video image target detection model obtained in the step 2 by using an Alternating Direction Method of Multipilers (ADMM) to obtain the optimal solution of the background part and the target part of the infrared video;
and 4, step 4: and (4) converting the background part and the target part obtained by the solution in the step (3) into a video form and outputting.
The invention has the beneficial effects that: the small target detection method based on the infrared video image can accurately detect the infrared small target, can realize accurate and effective extraction of the small target even under the condition that the target is weak to only a few pixels, and has no ghost phenomenon and void phenomenon in a detection result; secondly, the influence of the target motion condition on the detection result is reduced, and accurate detection can be realized when the target has a repeated track; in addition, the target detection method has better anti-interference capability, does not need to carry out additional denoising treatment, and has stronger applicability.
Drawings
FIG. 1 is a flow chart of a method for detecting a small target based on an infrared video image according to the present invention;
FIG. 2 is a schematic diagram of a detection model of a small target detection method based on an infrared video image according to the present invention;
fig. 3(a) is an input image for detecting a pedestrian according to the present invention, and fig. 3(b) is a detection result for detecting a pedestrian according to the present invention;
FIG. 4(a) is an input image of a first embodiment of the present invention for detecting batons, and FIG. 4(b) is a detection result of the first embodiment of the present invention for detecting batons;
FIG. 5(a) is an input image of a second embodiment of the present invention for detecting batswarms, and FIG. 5(b) is a detection result of the second embodiment of the present invention for detecting batswarms;
fig. 6(a) is an input image of the present invention detecting a drone, and fig. 6(b) is a detection result of the present invention detecting a drone;
7-1-1 is the result of pedestrian detection by low rank matrix decomposition, 7-1-2 is the result of pedestrian detection by the present invention, 7-1-3 is the result of pedestrian detection by frame difference method, and 7-1-4 is the result of pedestrian detection by Gaussian background modeling;
FIG. 7-2-1 is the detection result of the first embodiment of detecting bat groups by the low rank matrix decomposition method, FIG. 7-2-2 is the detection result of the first embodiment of detecting bat groups by the present invention, FIG. 7-2-3 is the detection result of the first embodiment of detecting bat groups by the frame difference method, FIG. 7-2-4 is the detection result of the first embodiment of detecting bat groups by the Gaussian background decomposition method;
FIG. 7-3-1 is the detection result of the second embodiment of bat group detection by low rank matrix factorization, FIG. 7-3-2 is the detection result of the second embodiment of bat group detection by the present invention, FIG. 7-3-3 is the detection result of the second embodiment of bat group detection by frame difference method, FIG. 7-3-4 is the detection result of the second embodiment of bat group detection by Gaussian background modeling;
fig. 7-4-1 is a detection result of detecting the unmanned aerial vehicle by the low rank matrix decomposition method, fig. 7-4-2 is a detection result of detecting the bat group according to the present invention, fig. 7-4-3 is a detection result of detecting the unmanned aerial vehicle by the frame difference method, and fig. 7-4-4 is a detection result of detecting the unmanned aerial vehicle by the gaussian background modeling method.
Detailed Description
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
The invention provides a small target detection method based on an infrared video image, which comprises the following steps as shown in figures 1 and 2:
step 1, inputting each frame of an M-frame infrared video sequence as a column vector to form a data matrix r;
step 2, assuming that the data matrix r is ═ r1(t),r2(t)…rn(t)]The method comprises the following steps of constructing a background model based on a robust matrix approximation theory, conducting weighted summation on the output of each array element, guiding an input infrared video signal to the background part, namely receiving an expected background signal and inhibiting other signals; modeling and solving the foreground object by adopting a sparse expression item; the modeling of the interference suppression item is carried out through the residual error of the noise which is similar to the original data formed by the background and the foreground, namely, the effect of removing the interference is achieved through the suppression factor on the assumption that the change of the scene and the influence brought by the environment are considered as the interference item, and a small target detection model based on the infrared video image is established:
Figure BDA0002328186180000051
wherein a is a guide matrix, a sparse matrix S is a target, b is a background, and lambda1And λ2To compromise the coefficient, | O | | livepIs 1pNorm, | · | luminanceFIs Frobenius norm.
Different from background modeling modes of other target detection theories, the method skillfully utilizes related contents of array signal processing, and the model can be regarded as an intelligent antenna system and can receive and process target data in real time. In the target detection theory, data consists of three parts of a background, a target and interference noise, the invention establishes a system, can realize an effective background modeling mode by utilizing a robust matrix approximation theory, can inhibit the noise, achieves a certain anti-interference effect and finally accurately detects the infrared small target;
step 3, solving the non-convex optimization model by using an Alternating Direction Multiplier Method (ADMM);
in order to simplify the calculation, the intermediate variable Z ═ r-S is introduced into the model proposed in step 2, and the model is expressed as:
Figure BDA0002328186180000052
the above formula is a non-convex optimization model, and is solved by using an alternating direction multiplier (ADMM) iterative algorithm, so as to obtain a separated background b and a separated target S. The method comprises the following specific steps:
(1) the augmented Lagrangian function is established according to the model as follows:
Figure BDA0002328186180000053
wherein Y is a Lagrange multiplier, and rho controls the iteration step length of the ADMM algorithm. For this model, the choice of p is important, with p equal to 1, then l evolving into S1Norm, p is 0, then l evolves to S0Norm,/, of0The norm represents the number of non-zero elements of the vector, in fact l0The norm is not solvable, so to more sparsely constrain the signal, 0 < p < 1 is chosen, with p closer to 0 and S more sparse.
(2) The optimization step by step is as follows:
fixing { Z (t), S (t), Y (t) }, and respectively updating { b (t +1) and S (t +1) };
solving the background b, neglecting the irrelevant terms, and simplifying the formula as follows:
Figure BDA0002328186180000061
namely solving:
Figure BDA0002328186180000062
solving the available b (t +1)
Solving the sparse foreground S, neglecting the irrelevant terms, and simplifying the formula as follows:
Figure BDA0002328186180000063
the above equation can be converted into:
Figure BDA0002328186180000064
namely solving:
Figure BDA0002328186180000065
wherein, the first and the second end of the pipe are connected with each other,
Figure BDA0002328186180000066
further, the above equation can be divided into n subproblem solutions:
Figure BDA0002328186180000071
wherein Sn，
Figure BDA0002328186180000072
Are respectively S,
Figure BDA0002328186180000073
Becomes:
Figure BDA0002328186180000074
where ρ iss＝ρ/λ1(ii) a From the above formula, it can be easily seen，
a)|Sn|pIs an even function and follows | SnIncreasing with increasing | the center of symmetry being Sn＝0；
b)
Figure BDA0002328186180000075
In that
Figure BDA0002328186180000076
Symmetrical, monotonically decreasing at (- ∞,0), at
Figure BDA0002328186180000077
Monotonically increasing;
c)f(Sn) In that
Figure BDA0002328186180000078
All are monotonous, therefore
Figure BDA0002328186180000079
f (0) is the local minimum over the two intervals, respectively.
Thus, f (S)n) Must occur at
Figure BDA00023281861800000710
Above, f (S) can thus be further discussed by discussing its first, second and third derivativesn) In that
Figure BDA00023281861800000711
Unevenness of the surface; it is clear that, given any value of p, we can obtain f (S) when 0 < p < 1n) Each element of S (t +1), which is the global minimum of (c), is available.
Fixing { b (t +1), S (t +1), Y (t) }, and solving { Z (t +1) };
ignoring the extraneous terms, the formula reduces to:
Figure BDA00023281861800000712
namely solving:
Figure BDA00023281861800000713
namely solving:
Figure BDA0002328186180000081
solving the above formula by using a fixed point iteration method to obtain: z (t + 1);
(iii) fixing { b (t +1), S (t +1), Z (t +1) }, updating { Y (t +1) }
Y(t+1)＝Y(t)+ρ(Z(t+1)-r+S(t+1)) (14)
Repeating the first step and the third step until the maximum iteration times or the convergence threshold value is reached, and obtaining a background b and a target S by adopting the method;
and 4, converting the background part and the target part obtained by solving into a video form and outputting.
Results of the experiment
The method of the invention is adopted to identify the small targets in the four infrared videos respectively. Fig. 3(a), fig. 4(a), fig. 5(a), and fig. 6(a) are original video images, and fig. 3(b), fig. 4(b), fig. 5(b), and fig. 6(b) are small object detection results according to the present invention; fig. 7 is a graph showing the effectiveness of the present invention by comparing the results of the low rank matrix decomposition method, the frame difference method, and the gaussian mixture background modeling method mentioned in the background art.
Fig. 3 shows a sequence of input images for pedestrian detection, and the images in fig. 3(a) are subjected to object detection using the method of the present invention as a preferred embodiment, and the results are shown in fig. 3 (b). 7-1-4 and 7-1-1, the Gaussian mixture background modeling and the low-rank sparse matrix decomposition can obtain complete detection results, but the effect of the low-rank sparse matrix decomposition is influenced by noise interference; the detection results obtained by the frame difference method in fig. 7-1-3 have the phenomena of "ghost" and "void". Aiming at the sequence, the method (shown in figure 7-1-2) provided by the invention has strong anti-interference capability, does not have the phenomena of ghost and void, and can achieve better detection effect.
The input images of the sequence shown in fig. 4 and 5, bat group detection, are subject to target detection on the images in fig. 4(a) and 5(a) with the method provided by the present invention as a preferred embodiment, and the results are shown in fig. 4(b) and 5 (b). As can be seen from fig. 4 and 5, the detection effect is greatly affected due to the rapidity and disorder of bat movement. The detection result obtained by adopting the frame difference method (figure 7-2-3) is greatly influenced by noise, the detection result obtained by adopting the mixed Gaussian background modeling method (figure 7-2-4) has a fuzzy phenomenon, and the detection result obtained by adopting the low-rank sparse matrix decomposition (figure 7-2-1) and threshold filtering and the method (figure 7-2-2) provided by the invention can achieve good effects. When the bat number is large, as shown in fig. 5(a), although a complete detection result can be obtained by adopting a frame difference method (fig. 7-3-3), a low-rank sparse matrix decomposition diagram (fig. 7-3-1) and a mixed gaussian background modeling method (fig. 7-3-4), the bat number is affected by noise caused by swaying branches and the like, and interference points of a lower right corner area in the picture corresponding to a tree shadow contour part of an original image are mistakenly detected as a target part; aiming at the sequence, the method (figure 7-3-2) provided by the invention has obvious denoising effect and does not have interference points.
Fig. 6 shows the sequence of input images for unmanned aerial vehicle detection, and the images in fig. 6(a) are subjected to target detection using the method provided by the present invention as a preferred embodiment, and the results are shown in fig. 6 (b). As can be seen from fig. 6, the scene of the sequence is most complex, the target is the smallest, the motion trail has repeated segments, and bright human buildings exist in the field of view, which makes the detection of the infrared micro target in the scene extremely difficult. As shown in fig. 7-4-4 and fig. 7-4-3, although the target can be extracted by the gaussian mixture background modeling and the frame difference method, the influence caused by the isolated noise point cannot be ignored, which brings great difficulty to practical application, and when the target motion trajectory is repeated, the two methods both generate a "void" phenomenon and a frame loss phenomenon, and cannot continuously capture target information; as shown in fig. 7-4-1, the target part can be extracted more completely by using a low-rank sparse matrix decomposition method, but is also influenced by isolated noise points; the method (figure 7-4-2) provided by the invention further improves the robustness of the frame on the basis of completely detecting the target, and reduces the influence of noise interference on the detection result to the greatest extent.
It can be seen that the method of the present invention well realizes an effective background modeling manner, can accurately detect the infrared small target, such as fig. 3-5, and can also realize accurate and effective extraction of the target even under the condition that the target is weak (such as fig. 6). Secondly, the method is less influenced by the motion condition of the target, and the detection result has no ghost phenomenon and void phenomenon; in addition, the invention can inhibit noise and achieve ideal anti-interference effect.

Claims (1)

1. A small target detection method based on an infrared video image is characterized by comprising the following steps:
step 1: inputting each frame of the M frames of infrared video sequences as a column vector to form a data matrix;
step 2: assuming that the data matrix input in the step 1 consists of three parts, namely a background, a target and interference noise, and establishing an infrared video image target detection model which takes a robust matrix approximation theory as a background modeling mode and is additionally based on a target sparse expression item and an anti-noise interference item based on an array signal processing theory; the method specifically comprises the following steps:
suppose the data matrix r ═ r1(t),r2(t)…rn(t)]The method comprises the following steps of constructing a background model based on a robust matrix approximation theory, conducting weighted summation on the output of each array element, guiding an input infrared video signal to the background part, namely receiving an expected background signal and inhibiting other signals; modeling and solving the foreground object by adopting a sparse expression item; modeling of interference suppression items is carried out through the residual error of the noise, which is similar to the original data formed by the background and the foreground, namely, assuming that the change of the scene and the influence caused by the environment are uniformly regarded as the interference items, the effect of removing the interference is achieved through the suppression factors, and the infrared vision based modeling is establishedAn image target detection model:
Figure FDA0003649158170000011
wherein r is the input data matrix, S is the sparse target portion, b is the background portion, a is the steering vector, λ1And λ2To compromise the coefficients, | · survivalpIs 1 ofpNorm, | \ | circumflectingFIs a Frobenius norm;
and step 3: solving the infrared video image target detection model obtained in the step 2 by using an alternating direction multiplier method to obtain the optimal solution of the infrared video background part and the target part; the method comprises the following steps:
step 3.1: according to the ADMM, introducing an auxiliary variable Z which is r-S, and constructing a Lagrangian function; the lagrangian function constructed is as follows:
Figure FDA0003649158170000021
wherein Y is a Lagrange multiplier, rho controls the iteration step length of the ADMM algorithm, | S | | YpThe p is more than 0 and less than 1, the closer p is to 0, the more sparse S is;
step 3.2: performing distribution optimization on the Lagrangian function constructed in the step 3.1 so as to obtain a separated background b and a separated target S; the method comprises the following steps:
step 3.2.1: fixing { Z (t), S (t), Y (t) }, and updating { b (t +1) and S (t +1) };
neglecting the irrelevant item, dividing into two subproblems to solve, and simplifying the formula for solving the background b into:
Figure FDA0003649158170000022
namely solving:
Figure FDA0003649158170000023
solving the available b (t +1)
Solving the sparse foreground S, neglecting the irrelevant terms, and simplifying the formula as follows:
Figure FDA0003649158170000024
namely solving:
Figure FDA0003649158170000025
wherein the content of the first and second substances,
Figure FDA0003649158170000026
further, the above equation can be divided into n subproblem solutions:
Figure FDA0003649158170000027
wherein Sn，
Figure FDA0003649158170000031
Are respectively S,
Figure FDA0003649158170000032
The nth element in (1), and the obtained optimal solution is a sparse target part;
step 3.2.2: fix { b (t +1), S (t +1), Y (t) }, update { Z (t +1) };
ignoring the extraneous terms, the formula reduces to:
Figure FDA0003649158170000033
solving the above formula by using a fixed point iteration method to obtain Z (t + 1);
step 3.2.3: fix { b (t +1), S (t +1), Z (t +1) }, update { Y (t +1) }
Y(t+1)＝Y(t)+ρ(Z(t+1)-r+S(t+1)) (14)
Repeating the steps 3.2.1-3.2.3 until the maximum iteration number or the convergence threshold is reached, and solving to obtain a background b and a target S;
and 4, step 4: and (4) converting the background part and the target part obtained by the solution in the step (3) into a video form and outputting.
CN201911325125.0A 2019-12-20 2019-12-20 Small target detection method based on infrared video image Active CN111160181B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911325125.0A CN111160181B (en) 2019-12-20 2019-12-20 Small target detection method based on infrared video image

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911325125.0A CN111160181B (en) 2019-12-20 2019-12-20 Small target detection method based on infrared video image

Publications (2)

Publication Number Publication Date
CN111160181A CN111160181A (en) 2020-05-15
CN111160181B true CN111160181B (en) 2022-07-05

Family

ID=70557438

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911325125.0A Active CN111160181B (en) 2019-12-20 2019-12-20 Small target detection method based on infrared video image

Country Status (1)

Country Link
CN (1) CN111160181B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111784599B (en) * 2020-06-24 2022-04-29 西北工业大学 Method for eliminating infrared image stripe noise
CN113421279B (en) * 2021-07-07 2023-03-31 中国科学院光电技术研究所 Infrared weak and small target detection method based on weighted nuclear norm minimization

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105931264A (en) * 2016-04-14 2016-09-07 西安电子科技大学 Sea-surface infrared small object detection method
CN107991636A (en) * 2017-11-24 2018-05-04 哈尔滨工业大学 A kind of rapid magnetic-resonance image rebuilding method based on adaptability structure low-rank matrix
CN108520501A (en) * 2018-03-30 2018-09-11 西安交通大学 A kind of video and removes rain snow method based on multiple dimensioned convolution sparse coding
CN109215025A (en) * 2018-09-25 2019-01-15 电子科技大学 A kind of method for detecting infrared puniness target approaching minimization based on non-convex order
CN109934178A (en) * 2019-03-18 2019-06-25 电子科技大学 A kind of method for detecting infrared puniness target based on Kronecker base rarefaction representation
CN110135344A (en) * 2019-05-16 2019-08-16 电子科技大学 Based on the method for detecting infrared puniness target for weighting fixed order expression

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105931264A (en) * 2016-04-14 2016-09-07 西安电子科技大学 Sea-surface infrared small object detection method
CN107991636A (en) * 2017-11-24 2018-05-04 哈尔滨工业大学 A kind of rapid magnetic-resonance image rebuilding method based on adaptability structure low-rank matrix
CN108520501A (en) * 2018-03-30 2018-09-11 西安交通大学 A kind of video and removes rain snow method based on multiple dimensioned convolution sparse coding
CN109215025A (en) * 2018-09-25 2019-01-15 电子科技大学 A kind of method for detecting infrared puniness target approaching minimization based on non-convex order
CN109934178A (en) * 2019-03-18 2019-06-25 电子科技大学 A kind of method for detecting infrared puniness target based on Kronecker base rarefaction representation
CN110135344A (en) * 2019-05-16 2019-08-16 电子科技大学 Based on the method for detecting infrared puniness target for weighting fixed order expression

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
An Enhanced High-Order Variational Model Based on Speckle Noise Removal With G0 Distribution;Yunping Mu等;《IEEE Access》;20190729;第104365-104379页 *
Partial Occlusion Handling for Visual Tracking via Robust Part Matching;Tianzhu Zhang等;《2014 IEEE Conference on Computer Vision and Pattern Recognition》;20140925;第1258-1265页 *
区域弱相关自表示的显著目标检测方法;陈晓云等;《计算机辅助设计与图形学学报》;20181231;第30卷(第12期);第2300-2310页 *
基于结构和判别语义嵌入的图像表征学习;张正;《中国博士学位论文全文数据库 信息科技辑》;20190131;第2019年卷(第1期);I138-205 *

Also Published As

Publication number Publication date
CN111160181A (en) 2020-05-15

Similar Documents

Publication Publication Date Title
Wang et al. Infrared dim target detection based on total variation regularization and principal component pursuit
US10301041B2 (en) Systems and methods for tracking moving objects
Zhang et al. Algorithms for optical weak small targets detection and tracking
CN103873743B (en) Video de-noising method based on structure tensor and Kalman filtering
CN109325446B (en) Infrared weak and small target detection method based on weighted truncation nuclear norm
CN111160181B (en) Small target detection method based on infrared video image
CN109215025B (en) Infrared weak and small target detection method based on non-convex rank approach minimization
CN110988818A (en) Cheating interference template generation method for countermeasure network based on condition generation formula
CN110135344B (en) Infrared dim target detection method based on weighted fixed rank representation
CN108765460B (en) Hyperspectral image-based space-time joint anomaly detection method and electronic equipment
CN116486243A (en) DP-ViT-based sonar image target detection method
Chen et al. Space-based infrared aerial target detection method via interframe registration and spatial local contrast
Li et al. Weak moving object detection in optical remote sensing video with motion-drive fusion network
Gupta et al. Infrared small target detection enhancement using a lightweight convolutional neural network
Li et al. Dim moving target detection algorithm based on spatio-temporal classification sparse representation
Chawla et al. Image masking for robust self-supervised monocular depth estimation
CN114005018B (en) Small calculation force driven multi-target tracking method for unmanned surface vehicle
CN112417948B (en) Method for accurately guiding lead-in ring of underwater vehicle based on monocular vision
CN115439777A (en) Video satellite target tracking method based on multi-feature fusion and motion estimation
Wu et al. Infrared Dim and Small Target Detection Algorithm Combining Multiway Gradient Regularized Principal Component Decomposition Model
CN110443209B (en) Infrared small target detection method based on matrix recovery
Erokhin et al. Detection and tracking of moving objects with real-time onboard vision system
Oveis et al. Moving and Stationary Targets Separation in SAR Signal Domain Using Parallel Convolutional Autoencoders with RPCA Loss
Xing et al. Computationally efficient RGB-T UAV detection and tracking system
Pakfiliz A New Method for Surface-to-Air Video Detection and Tracking of Airborne Vehicles

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