CN113537150B - Hyperspectral image target anomaly detection method, system, terminal and storage medium - Google Patents
Hyperspectral image target anomaly detection method, system, terminal and storage medium Download PDFInfo
- Publication number
- CN113537150B CN113537150B CN202110921317.9A CN202110921317A CN113537150B CN 113537150 B CN113537150 B CN 113537150B CN 202110921317 A CN202110921317 A CN 202110921317A CN 113537150 B CN113537150 B CN 113537150B
- Authority
- CN
- China
- Prior art keywords
- hyperspectral image
- weight
- image target
- hyperspectral
- maximum cross
- 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
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 69
- 238000000034 method Methods 0.000 claims abstract description 29
- 230000006870 function Effects 0.000 claims abstract description 17
- 230000002159 abnormal effect Effects 0.000 claims abstract description 16
- 230000005856 abnormality Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 12
- 230000008569 process Effects 0.000 claims description 9
- 238000001228 spectrum Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 3
- 230000008407 joint function Effects 0.000 abstract description 2
- 230000000694 effects Effects 0.000 description 5
- 238000013459 approach Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 241001466077 Salina Species 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
-
- 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
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10036—Multispectral image; Hyperspectral image
-
- 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Abstract
A hyperspectral image target anomaly detection method, a hyperspectral image target anomaly detection system, a hyperspectral image target anomaly detection terminal and a hyperspectral image target anomaly detection storage medium, wherein the hyperspectral image target anomaly detection method comprises the following steps: giving a proper weight to each wave band of the hyperspectral image through the maximum cross-correlation entropy cooperation representation; for hyperspectral images with proper weight in each band, pixel anomaly scores are obtained through maximum cross-correlation entropy cooperative representation, and anomaly pixels are selected through proper threshold value setting. According to the method, different weights are self-adaptively given to different wavebands of the hyperspectral image, so that the influence of the waveband noise and the abnormal pixel point segments on the detection accuracy can be effectively avoided, and the robustness of an algorithm is improved. The objective function is constructed by the joint function, the wave band weight and the neighborhood pixel point weight are solved at the same time, and the iterative solution algorithm is used for efficiently solving, so that the problem of suboptimal solution possibly caused by independent learning of the wave band weight and the neighborhood pixel point weight is avoided. The invention can effectively improve the accuracy.
Description
Technical Field
The invention belongs to the field of remote sensing image processing, and particularly relates to a hyperspectral image target anomaly detection method, a hyperspectral image target anomaly detection system, a hyperspectral image target anomaly detection terminal and a hyperspectral image target anomaly detection storage medium.
Background
The hyperspectral image has a plurality of spectrum domains, can provide a large amount of spectrum information and space information of the earth surface, and is widely applied to the fields of scene classification, mixing, clustering, change detection, target detection and the like due to the good detection performance of the hyperspectral image. Hyperspectral anomaly detection, one of the applications of hyperspectral images, is one of the hot spots of current research, and is widely applied to the military and civil fields at present, and has good effects. Hyperspectral anomaly detection is an unsupervised two-classification problem, requiring no prior knowledge. In hyperspectral anomaly detection, an object whose spectrum is significantly different from the spectrum around it is regarded as anomaly, and objects other than anomaly in the entire image are defined as background.
Many methods exist today to accomplish hyperspectral anomaly detection. The Reed-Xiaoli (RX) algorithm is based on Mahalanobis distance between pixel and background and assumes that a multivariate gaussian background distribution is used to identify anomalies, which in turn is divided into GRX modeling the background using the whole image and LRX modeling using a local double window. Later, a learner proposed a non-linear version of RX (KRX), which extends to a feature space related to the original input space by a non-linear mapping function. Since KRX has a problem of target pollution background, and a learner improves an algorithm based on the problem, a robust nonlinear anomaly detection algorithm (RNAD) using robust regression analysis in kernel space is proposed. Detectors (CRDs) based on collaborative representations have been proposed later and widely used. The CRD assumes that every pixel in the background can be approximated using its neighborhood pixels, while outlier pixels cannot. The background pixels are assumed to be linear combinations of neighboring pixels and by minimizationNorms to enhance the collaborative representation. Currently there are RX-based detectors that assume that the background is uniform and that are modeled using a multivariate gaussian distribution. But in practice this assumption is often not met, which leads to a reduced accuracy.
In addition, since the hyperspectral image is not completely shielded from noise and abnormal interference at the time of acquisition, noise and abnormal pixels may exist in a certain band in the hyperspectral image. The CRD and its derivative algorithm assume that each band has the same importance, and such an assumption will be less robust to noise bands, thereby reducing the effectiveness of anomaly detection.
Related literature:
【1】Reed I S,Yu X.Adaptive multiple-band CFAR detection of an optical pattern with unknown spectral distribution[J].IEEE Transactions on Acoustics,Speech,and Signal Processing,1990,38(10):1760-1770.
【2】Molero J M,Garzón E M,García I,et al.Analysis and optimizations of global and local versions of the RX algorithm for anomaly detection in hyperspectral data[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2013,6(2):801-814.
【3】Kwon H,Nasrabadi N M.Kernel RX-algorithm:Anonlinear anomaly detector for hyperspectral imagery[J].IEEE transactions on Geoscience and Remote Sensing,2005,43(2):388-397.
【4】Zhao R,Du B,Zhang L.Arobust nonlinear hyperspectral anomaly detection approach[J].IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing,2014,7(4):1227-1234.
【5】Li W,Du Q.Collaborative representation for hyperspectral anomaly detection[J].IEEE Transactions on Geoscience and Remote Sensing,2014,53(3):1463-1474.
disclosure of Invention
The invention aims to solve the problems in the prior art and provide a hyperspectral image target anomaly detection method, a hyperspectral image target anomaly detection system, a hyperspectral image target anomaly detection terminal and a hyperspectral image target anomaly detection storage medium, which can effectively avoid the influence of noise and abnormal wavebands on hyperspectral anomaly detection.
In order to achieve the above purpose, the present invention has the following technical scheme:
a hyperspectral image target abnormality detection method comprises the following steps:
giving a proper weight to each wave band of the hyperspectral image through the maximum cross-correlation entropy cooperation representation;
for hyperspectral images with proper weight in each band, pixel anomaly scores are obtained through maximum cross-correlation entropy cooperative representation, and anomaly pixels are selected through proper threshold value setting.
As a means ofThe hyperspectral image is defined asRepresenting a hyperspectral image having n pixels, the dimension being d-dimension; partial approximation using sliding double windows, the outer and inner windows being sized by w, respectively out And w in To indicate, therefore, the pixels of the neighborhood are in +.> Expressed, wherein s=w out ×w out -w in ×w in 。
As a preferable scheme of the hyperspectral image target abnormality detection method, the maximum cross-correlation entropy cooperation of the hyperspectral image represents the following objective function:
wherein,representing the weight of each neighborhood pixel used, λ being a regularization parameter;
in the method, in the process of the invention,larger represents larger difference of collaborative representation, then +.>The smaller.
As a preferable scheme of the hyperspectral image target abnormality detection method, the maximum cross-correlation entropy cooperative representation target function is reconstructed through a half-quadratic technology to obtain the following expression:
in the method, in the process of the invention,is a diagonal matrix, each element on the diagonal +.>A weight representing the band;
w k is defined as follows:
as a preferable scheme of the hyperspectral image target anomaly detection method, the method utilizes a diagonal regularization matrix to solve the problem that the coefficient is smaller corresponding to the pixel point with larger central pixel difference
Wherein Γ is i For Tikhonov regularization matrix, the expression is as follows:
as a preferable scheme of the hyperspectral image target abnormality detection method, the proposed objective function is optimized in an iterative optimization mode, and the unsupervised hyperspectral image target abnormality detection is completed.
The invention also provides a hyperspectral image target anomaly detection system, which comprises:
the spectrum band weighting module is used for giving a proper weight to each band of the hyperspectral image through the maximum cross-correlation entropy cooperative representation;
and the abnormal pixel screening module is used for obtaining a pixel abnormal score through maximum cross-correlation entropy cooperative representation for the hyperspectral image with a proper weight of each wave band, and selecting abnormal pixels through setting a proper threshold.
The invention also provides a terminal device which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the steps of the hyperspectral image target abnormality detection method when executing the computer program.
The invention also provides a computer readable storage medium storing a computer program which when executed by a processor implements the steps of the hyperspectral image target anomaly detection method.
Compared with the prior art, the invention has at least the following beneficial effects:
in the hyperspectral image target anomaly detection process, different weights are self-adaptively given to different wave bands of the hyperspectral image, the hyperspectral image with a proper weight for each wave band is obtained through maximum cross-correlation entropy cooperative expression, abnormal pixels are selected through setting a proper threshold, influence of wave band noise and abnormal pixel point segments on detection accuracy can be effectively avoided, and robustness of an algorithm is improved. Compared with the current main stream hyperspectral image target anomaly detection method, the method provided by the invention has obvious advantages, and the accuracy rate on the public data set is greatly improved over other methods.
Furthermore, the objective function is constructed as a joint function, the wave band weight and the neighborhood pixel point weight are solved at the same time, and the iterative solving algorithm is used for efficiently solving, so that the problem of suboptimal solution possibly caused by independent learning of the wave band weight and the neighborhood pixel point weight is avoided.
Drawings
FIG. 1 is a graph comparing ROC curves on an AVIRIS-I dataset with other methods of the present invention;
FIG. 2 is a graph comparing ROC curves of the present invention with other methods on an ABU-air-1 dataset;
FIG. 3 AVIRIS-I dataset pseudo-color map;
FIG. 4 AVIRIS-I dataset tab map;
FIG. 5 AVIRIS-I dataset CRD detection result graph;
FIG. 6 shows a graph of the results of the detection of the present invention for an AVIRIS-I dataset;
FIG. 7 AVIRIS-II dataset pseudo-color map;
FIG. 8 AVIRIS-II dataset tab map;
FIG. 9 AVIRIS-II dataset CRD detection result graph;
FIG. 10 shows a graph of the results of the inventive test for the AVIRIS-II dataset.
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Let the hyperspectral image be defined asRepresenting a hyperspectral image having n pixels, the dimension being d-dimension; partial approximation using sliding double windows, the outer and inner windows being sized by w, respectively out And w in To indicate, therefore, the pixels of the neighborhood are in +.>Expressed, wherein s=w out ×w out -w in ×w in 。
S1, giving a proper weight to each wave band of the hyperspectral image through maximum cross-correlation entropy cooperative representation;
s2, obtaining pixel anomaly scores through maximum cross-correlation entropy cooperative representation for hyperspectral images with proper weights of each band, and selecting anomaly pixels through setting proper thresholds.
The hyperspectral image target anomaly detection method provided by the invention constructs a detector based on the maximum entropy theory (MCC) proposed in Information Theory Learning (ITL). ITL provides a unified approach to improving machine learning approaches to outlier robustness. The MCC can obtain more information from the data for adjustment, and thus can have more information when processing non-Gaussian signalsGreat advantage. Entropy introduces a new metric that is equal to two points when they are closeNorms. As the two points get farther apart, the metric will approach +.>Norms, eventually approaching->Norms. Therefore, MCC is robust against outliers.
The maximum cross-correlation entropy cooperation of the hyperspectral image represents the objective function as follows:
wherein,representing the weight of each neighborhood pixel used, λ being a regularization parameter;
in the method, in the process of the invention,larger represents larger difference of collaborative representation, then +.>The smaller.
Reconstructing the maximum cross-correlation entropy cooperative representation objective function through a half-quadratic technology to obtain the following expression:
in the method, in the process of the invention,is a diagonal matrix, each element on the diagonal +.>A weight representing the band;
w k is defined as follows:
the smaller coefficient corresponding to the pixel point with larger difference with the central pixel is solved by using the diagonal regularization matrix Wherein Γ is i For Tikhonov regularization matrix, the expression is as follows:
and optimizing the proposed objective function by adopting an iterative optimization mode to finish the objective anomaly detection of the unsupervised hyperspectral image.
The hyperspectral image target anomaly detection method provided by the invention can effectively avoid the influence of noise and abnormal wave bands on hyperspectral anomaly detection. The noise band is given a sufficiently small weight by the method of the present invention so as not to affect the reasoning result, while the band contributing much more is given more weight so that it can provide more effective information.
The invention also provides a hyperspectral image target anomaly detection system, which comprises:
the spectrum band weighting module is used for giving a proper weight to each band of the hyperspectral image through the maximum cross-correlation entropy cooperative representation;
and the abnormal pixel screening module is used for obtaining a pixel abnormal score through maximum cross-correlation entropy cooperative representation for the hyperspectral image with a proper weight of each wave band, and selecting abnormal pixels through setting a proper threshold.
The invention also provides a terminal device which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor realizes the steps of the hyperspectral image target abnormality detection method when executing the computer program.
The invention also provides a computer readable storage medium storing a computer program which when executed by a processor implements the steps of the hyperspectral image target anomaly detection method.
The computer program may be divided into one or more modules/units, which are stored in the memory and executed by the processor to perform the hyperspectral image object anomaly detection method of the present invention.
The terminal can be a computing device such as a desktop computer, a notebook computer, a palm computer, a cloud server and the like, and can also be a processor and a memory. The processor may be a central processing unit (CentralProcessingUnit, CPU), but may also be other general purpose processors, digital signal processors (DigitalSignalProcessor, DSP), application specific integrated circuits (ApplicationSpecificIntegratedCircuit, ASIC), off-the-shelf programmable gate arrays (Field-ProgrammableGateArray, FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, or the like. The memory may be used to store computer programs and/or modules and the processor may implement various functions of the hyperspectral image target anomaly detection system by running or executing the computer programs and/or modules stored in the memory and invoking data stored in the memory.
Examples
A hyperspectral image target abnormality detection method comprises the following steps:
step one: inputting hyperspectral imageRegular term parameters λ, σ and sliding window size w out And w in 。
Step two: solving the objective function, fixing W i (giving an initial value I) updating alpha i The objective function will translate into:
let the above formula be the same as the formula i The derivative is derived and the derivative is set to 0, and the following steps can be obtained:
according to this formula, alpha can be calculated i And updating.
Step three: fix alpha i Updating W by formula i 。
Step four: and repeating the second step and the third step until convergence.
Step five: the sliding window is moved and the next pixel is processed. And performing the operation of the second step, the third step and the fourth step.
Step six: by passing throughObtaining the hyperspectral full-pixel anomaly score, and setting a threshold value to obtain the anomaly pixel.
Table 1 is a comparison of AUC values of the present mainstream method with AUC values of the disclosed dataset for hyperspectral anomaly detection based on the maximum entropy criterion, wherein the larger the AUC values, the better.
TABLE 1
D a taset | GRX | LRX | CRD | LSMAD | SWCRD | ERCRD | Ours |
AVIRIS-I | 0.9111 | 0.8584 | 0.9911 | 0.9758 | 0.9936 | 0.9678 | 0.9985 |
AVIRIS-II | 0.9403 | 0.8466 | 0.9696 | 0.9703 | 0.9798 | 0.9614 | 0.9855 |
AVIRIS-III | 0.8771 | 0.8411 | 0.9828 | 0.9318 | 0.9837 | 0.9349 | 0.9904 |
ABU-airport - 1 | 0.8221 | 0.8886 | 0.923 | 0.8806 | 0.9273 | 0.897 | 0.9631 |
ABU-airport-2 | 0.8404 | 0.9706 | 0.9735 | 0.9315 | 0.9692 | 0.9693 | 0.9879 |
Salinas | 0.8872 | 0.9172 | 0.9606 | 0.9351 | 0.951 | 0.9447 | 0.9643 |
FIGS. 1 and 2 are graphs comparing the ROC curves of the present invention with the current mainstream method on the AVIRIS-I and ABU-air-1 datasets, the curves in the graphs being better the closer to the upper left corner. Fig. 3, fig. 4, fig. 5 and fig. 6 are respectively a pseudo color image, a label image, a detection effect image of CRD and a detection effect image of the present invention of the aviis-I dataset. Fig. 6, 7, 8 and 9 are respectively a pseudo-color image, a label image, a CRD detection effect image and a detection effect image of the present invention of an aviis-II dataset, and the brighter the pixel in these detection images is, the higher the possibility that the pixel is an outlier. Compared with the current main stream hyperspectral image target abnormality detection method, the method provided by the invention has obvious advantages, and the accuracy rate on the public data set (AVIRIS-I data set and AVIRIS-II data set) is greatly higher than that of other methods.
The foregoing description of the preferred embodiment of the present invention is not intended to limit the technical solution of the present invention in any way, and it should be understood that the technical solution can be modified and replaced in several ways without departing from the spirit and principle of the present invention, and these modifications and substitutions are also included in the protection scope of the claims.
Claims (6)
1. The hyperspectral image target abnormality detection method is characterized by comprising the following steps of:
giving a proper weight to each wave band of the hyperspectral image through the maximum cross-correlation entropy cooperation representation;
obtaining pixel anomaly scores through maximum cross-correlation entropy cooperative representation for hyperspectral images with proper weights in each band, and selecting anomaly pixels through setting proper thresholds;
the maximum cross-correlation entropy cooperation of the hyperspectral image represents the objective function as follows:
wherein,representing the weight of each neighborhood pixel used, λ being a regularization parameter;
in the method, in the process of the invention,larger represents larger difference of collaborative representation, then +.>The smaller;
reconstructing the maximum cross-correlation entropy cooperative representation objective function through a half-quadratic technology to obtain the following expression:
in the method, in the process of the invention,is a diagonal matrix, each element on the diagonal +.>A weight representing the band;
w k is defined as follows:
the smaller coefficient corresponding to the pixel point with larger difference with the central pixel is solved by using the diagonal regularization matrix
Wherein Γ is i For Tikhonov regularization matrix, the expression is as follows:
2. the hyperspectral image target abnormality detection method according to claim 1, characterized in that: the hyperspectral image is defined asRepresenting a hyperspectral image having n pixels, the dimension being d-dimension; partial approximation using sliding double windows, the outer and inner windows being sized by w, respectively out And w in To indicate, therefore, the pixels of the neighborhood are in +.>Expressed, wherein s=w out ×w out -w in ×w in 。
3. The hyperspectral image target abnormality detection method according to claim 1, characterized in that: and optimizing the proposed objective function by adopting an iterative optimization mode to finish the objective anomaly detection of the unsupervised hyperspectral image.
4. A hyperspectral image target abnormality detection system, characterized by comprising:
the spectrum band weighting module is used for giving a proper weight to each band of the hyperspectral image through the maximum cross-correlation entropy cooperative representation;
the abnormal pixel screening module is used for obtaining pixel abnormal scores through maximum cross-correlation entropy cooperative representation for hyperspectral images with proper weight of each wave band, and selecting abnormal pixels through setting proper thresholds;
the maximum cross-correlation entropy cooperation of the hyperspectral image represents the objective function as follows:
wherein,representing the weight of each neighborhood pixel used, λ being a regularization parameter;
in the method, in the process of the invention,larger represents larger difference of collaborative representation, then +.>The smaller;
reconstructing the maximum cross-correlation entropy cooperative representation objective function through a half-quadratic technology to obtain the following expression:
in the method, in the process of the invention,is a diagonal matrix, each element on the diagonal +.>A weight representing the band;
w k is defined as follows:
the smaller coefficient corresponding to the pixel point with larger difference with the central pixel is solved by using the diagonal regularization matrix
Wherein Γ is i For Tikhonov regularization matrix, the expression is as follows:
5. a terminal device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: the steps of the hyperspectral image target abnormality detection method according to any one of claims 1 to 3 are realized when the processor executes the computer program.
6. A computer-readable storage medium storing a computer program, characterized in that: the computer program when executed by a processor implements the steps of the hyperspectral image target abnormality detection method as claimed in any one of claims 1 to 3.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110921317.9A CN113537150B (en) | 2021-08-11 | 2021-08-11 | Hyperspectral image target anomaly detection method, system, terminal and storage medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110921317.9A CN113537150B (en) | 2021-08-11 | 2021-08-11 | Hyperspectral image target anomaly detection method, system, terminal and storage medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN113537150A CN113537150A (en) | 2021-10-22 |
CN113537150B true CN113537150B (en) | 2024-04-02 |
Family
ID=78090891
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110921317.9A Active CN113537150B (en) | 2021-08-11 | 2021-08-11 | Hyperspectral image target anomaly detection method, system, terminal and storage medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113537150B (en) |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2010132135A1 (en) * | 2009-05-14 | 2010-11-18 | Raytheon Company | Adaptive spatial-spectral processing (assp) |
CN105989597A (en) * | 2015-02-13 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Hyperspectral image anomaly target detection method based on pixel selection process |
CN109447272A (en) * | 2018-09-29 | 2019-03-08 | 西安交通大学 | A kind of extreme learning machine method based on center of maximum cross-correlation entropy criterion |
WO2019218313A1 (en) * | 2018-05-17 | 2019-11-21 | 五邑大学 | Progressive dynamic hyperspectral image classification method |
CN110705600A (en) * | 2019-09-06 | 2020-01-17 | 西安交通大学 | Cross-correlation entropy based multi-depth learning model fusion method, terminal device and readable storage medium |
CN111160310A (en) * | 2020-01-02 | 2020-05-15 | 西北工业大学 | Hyperspectral abnormal target detection method based on self-weight collaborative representation |
-
2021
- 2021-08-11 CN CN202110921317.9A patent/CN113537150B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2010132135A1 (en) * | 2009-05-14 | 2010-11-18 | Raytheon Company | Adaptive spatial-spectral processing (assp) |
CN105989597A (en) * | 2015-02-13 | 2016-10-05 | 中国科学院西安光学精密机械研究所 | Hyperspectral image anomaly target detection method based on pixel selection process |
WO2019218313A1 (en) * | 2018-05-17 | 2019-11-21 | 五邑大学 | Progressive dynamic hyperspectral image classification method |
CN109447272A (en) * | 2018-09-29 | 2019-03-08 | 西安交通大学 | A kind of extreme learning machine method based on center of maximum cross-correlation entropy criterion |
CN110705600A (en) * | 2019-09-06 | 2020-01-17 | 西安交通大学 | Cross-correlation entropy based multi-depth learning model fusion method, terminal device and readable storage medium |
CN111160310A (en) * | 2020-01-02 | 2020-05-15 | 西北工业大学 | Hyperspectral abnormal target detection method based on self-weight collaborative representation |
Non-Patent Citations (2)
Title |
---|
周千 ; 马文涛 ; 桂冠 ; .基于l_1-范数约束的递归互相关熵的稀疏系统辨识.信号处理.2016,(第09期),全文. * |
朱德辉 ; 杜博 ; 张良培 ; .波段选择协同表达高光谱异常探测算法.遥感学报.2020,(第04期),全文. * |
Also Published As
Publication number | Publication date |
---|---|
CN113537150A (en) | 2021-10-22 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Yang et al. | Image analysis by Gaussian–Hermite moments | |
Sobral et al. | Online stochastic tensor decomposition for background subtraction in multispectral video sequences | |
Zhang et al. | An adaptive bilateral filter based framework for image denoising | |
Lefkimmiatis et al. | Nonlocal structure tensor functionals for image regularization | |
US9865037B2 (en) | Method for upscaling an image and apparatus for upscaling an image | |
Xu et al. | Spectral–spatial classification of hyperspectral image based on low-rank decomposition | |
Xiang et al. | Hyperspectral anomaly detection by local joint subspace process and support vector machine | |
CN111476813A (en) | Image change detection method, image change detection device, electronic equipment and storage medium | |
Ma et al. | Multiscale superpixelwise prophet model for noise-robust feature extraction in hyperspectral images | |
Sun et al. | A general non-local denoising model using multi-kernel-induced measures | |
CN114998750B (en) | Method, device, equipment and medium for removing random length strips of remote sensing image | |
Bui et al. | Projection-free kernel principal component analysis for denoising | |
Tu et al. | Hyperspectral image classification by combining local binary pattern and joint sparse representation | |
Yang et al. | A new TLD target tracking method based on improved correlation filter and adaptive scale | |
Wei et al. | Hyperspectral anomaly detection based on graph regularized variational autoencoder | |
Han et al. | Deep low-rank graph convolutional subspace clustering for hyperspectral image | |
Rajabi et al. | A Modified adaptive hysteresis smoothing approach for image denoising based on spatial domain redundancy | |
CN113537150B (en) | Hyperspectral image target anomaly detection method, system, terminal and storage medium | |
Shire et al. | A review paper on: agricultural plant leaf disease detection using image processing | |
Xiao et al. | Robust Tensor Low-Rank Sparse Representation with Saliency Prior for Hyperspectral Anomaly Detection | |
Bi et al. | l 2, p-norm sequential bilateral 2DPCA: a novel robust technology for underwater image classification and representation | |
Wang et al. | Dynamic super-pixel normalization for robust hyperspectral image classification | |
Kim et al. | ADOM: ADMM-based optimization model for stripe noise removal in remote sensing image | |
CN108492283B (en) | Hyperspectral image anomaly detection method based on band-constrained sparse representation | |
CN113222924B (en) | Hyperspectral image anomaly detection system based on FPGA |
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 |