CN110046674A - Classification of hyperspectral remote sensing image method based on feedback arrangement - Google Patents
Classification of hyperspectral remote sensing image method based on feedback arrangement Download PDFInfo
- Publication number
- CN110046674A CN110046674A CN201910337636.8A CN201910337636A CN110046674A CN 110046674 A CN110046674 A CN 110046674A CN 201910337636 A CN201910337636 A CN 201910337636A CN 110046674 A CN110046674 A CN 110046674A
- Authority
- CN
- China
- Prior art keywords
- classification
- image
- iteration
- hyperspectral
- indicate
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000012706 support-vector machine Methods 0.000 claims abstract description 13
- 238000001914 filtration Methods 0.000 claims abstract description 4
- VMXUWOKSQNHOCA-UKTHLTGXSA-N ranitidine Chemical compound [O-][N+](=O)\C=C(/NC)NCCSCC1=CC=C(CN(C)C)O1 VMXUWOKSQNHOCA-UKTHLTGXSA-N 0.000 claims description 6
- 238000001228 spectrum Methods 0.000 description 11
- 230000003595 spectral effect Effects 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 235000008331 Pinus X rigitaeda Nutrition 0.000 description 1
- 235000011613 Pinus brutia Nutrition 0.000 description 1
- 241000018646 Pinus brutia Species 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
Classification of hyperspectral remote sensing image method based on feedback arrangement, belongs to electronic information technical field, and the present invention is solves the problems, such as existing hyperspectral image classification method there are niceties of grading that the low, training time is grown.The present invention is classified in hyperspectral image data using support vector machines;Judge whether classification results meet stop condition, if it is terminate iteration, classification results at this time are final classification as a result, if otherwise generating bianry image block;Space filtering is carried out to bianry image block;It is fed back, is stacked with the hyperspectral image data of input;Then the number of iterations is added one, return re-starts classification, until terminating iteration using classification results as final classification result.The present invention is used for remote sensing technology.
Description
Technical field
The present invention relates to a kind of classification methods of high-spectrum remote sensing, belong to electronic information technical field.
Background technique
With the development of remote sensing technology and imaging spectral technology, high-spectrum remote sensing becomes the mankind and observes earth surface
One of important tool.High light spectrum image-forming technology combines imaging technique and spectrographic detection technology, for each space pixel several
Hundred so thousands of continuous narrow-bands on be imaged.The maximum difference of itself and traditional remote sensing images is embodied in two o'clock: first is that
The coverage area of spectral coverage is wider, high spectrum image may include ultraviolet, visible light, near-infrared and in infrared multiple spectral coverage ranges,
And traditional remote sensing images only include visible-range;Second is that wave band is narrower, the waveband width of general high spectrum image is on the left side 10nm
The right side, and traditional remote sensing images generally only include three wave bands of red, green, blue, and width is generally several hundred nanometers.Therefore, EO-1 hyperion
Image had both included spatial surface information, also included spectral information.
Due to different atural objects on different-waveband different, the more traditional remote sensing images of high spectrum image to the absorption of spectrum
More terrestrial object informations are capable of providing, there is better atural object identification capability.High spectrum image terrain classification is high spectrum image
One of important application, be divided into Supervised classification and unsupervised segmentation two major classes, wherein Supervised classification is the presence of training number
According to the classification under the conditions of i.e. prior information, parameter training first generally is carried out to classifier with training data, then with trained
Classifier classifies to the data newly obtained, and unsupervised data are the blind classification without using prior information.Due to being imaged
The complexity of journey and the diversity of atural object mainly use supervised classification method in practical application.
In recent years, researcher proposes the hyperspectral image classification method for much having supervision, including classical supporting vector
Machine method and the convolutional neural networks method developed rapidly in recent years.It is asked however, these methods still have following two
Topic:
1) in the case where training sample is less, nicety of grading is lower;
2) generally longer in convolutional neural networks method as the training time of the classification method of representative, it cannot be effectively in reality
It is used in the application of border.
Summary of the invention
That there are niceties of grading the invention aims to solving existing hyperspectral image classification method is low, the training time is long
Problem provides a kind of Classification of hyperspectral remote sensing image method based on feedback arrangement.
Classification of hyperspectral remote sensing image method of the present invention based on feedback arrangement, detailed process are as follows:
S1, hyperspectral image data conduct input number of the format for a × b × N, comprising M atural object classification will be formed substantially
According to Ω (k), and enable the number of iterations k=0;
S2, in k iteration, classified on hyperspectral image data Ω (k) using support vector machines, classification results
ForWherein SVM indicates support vector machines, C presentation class, and MAP indicates Maximun Posterior Probability Estimation Method,It indicates to carry out classifying using support vector machines for a variable and utilizes Maximun Posterior Probability Estimation Method generation
Classification results;
S3, as k > 0, judgementWhether meet stop condition, if it is terminates iteration, at this timeFor final classification as a result, if otherwise executing S4;
S4, basisGenerate bianry image block
In j-th of image, position is that the value of the pixel of (x, y) and the pixel existOn classification it is corresponding, then
It is 1 that position, which is the value of the pixel of (x, y), in j-th of image, is otherwise 0;
S5, the bianry image block that S4 is generatedSpace filtering is carried out, filtered gray level image block is
S6, the filtered gray level image block for obtaining S5It is fed back, the high spectrum image with input
Data Ω (k) is stacked:
Then k=k+1 is enabled, and returns and executes S2, until terminating iteration willAs final classification result.
Preferably, using Tanimoto coefficient value TI as the condition for differentiating stopping:
And
Wherein,Indicate setIn element number,
Indicate setIn element number,
Indicate the classification results for carrying out the jth class of classification acquisition in kth time iteration by support vector machines,
Indicate the classification results for carrying out the jth class of classification acquisition in -1 iteration of kth by classifier,
TI(k)Indicate the Tanimoto coefficient value in kth time iteration,
Indicate the Tanimoto coefficient value of jth width classification results image in kth time iteration.
Advantages of the present invention: the Classification of hyperspectral remote sensing image method proposed by the present invention based on feedback arrangement, it can be most
Posterior information is converted prior information by the extraction spatial information of limits, and each iteration all increases a part of spatial information, real
Making full use of for spatial information is showed.This method can be obviously improved classification compared with existing hyperspectral image classification method
Precision, to existing public data collection such as Purdue ' s Indian Pine, in the condition that 10% sample point of selection is trained
Under, nicety of grading reaches 97% or more.In addition, existing nicety of grading is preferably based on convolutional Neural under identical conditions
The method of network generally requires the training time of a few hours, and the training time of this method greatly shortens, and about 5-10 minutes.
Detailed description of the invention
Fig. 1 is the schematic diagram of the Classification of hyperspectral remote sensing image method of the present invention based on feedback arrangement.
Specific embodiment
Specific embodiment 1: illustrating present embodiment below with reference to Fig. 1, based on feedback arrangement described in present embodiment
Classification of hyperspectral remote sensing image method, detailed process are as follows:
S1, hyperspectral image data conduct input number of the format for a × b × N, comprising M atural object classification will be formed substantially
According to Ω (k), and enable the number of iterations k=0;
S2, in k iteration, classified on hyperspectral image data Ω (k) using support vector machines, classification results
ForWherein SVM indicates support vector machines, C presentation class, and MAP indicates Maximun Posterior Probability Estimation Method,Point that carries out classifying using support vector machines and generate using Maximun Posterior Probability Estimation Method is indicated for a variable
Class result;
S3, as k > 0, judgementWhether meet stop condition, if it is terminates iteration, at this timeFor final classification as a result, if otherwise executing S4;
S4, basisGenerate bianry image block
In j-th of image, position is that the value of the pixel of (x, y) and the pixel existOn classification it is corresponding, then
It is 1 that position, which is the value of the pixel of (x, y), in j-th of image, is otherwise 0;
S5, the bianry image block that S4 is generatedSpace filtering is carried out, filtered gray level image block is
S6, the filtered gray level image block for obtaining S5It is fed back, the high spectrum image with input
Data Ω (k) is stacked:
Then k=k+1 is enabled, and returns and executes S2, until terminating iteration willAs final classification result.
Stop condition described in S3 are as follows:
Using Tanimoto coefficient value TI as the condition for differentiating stopping:
And
Wherein,Indicate setIn element number,
Indicate setIn element number,
Indicate the classification results for carrying out the jth class of classification acquisition in kth time iteration by support vector machines,
Indicate the classification results for carrying out the jth class of classification acquisition in -1 iteration of kth by classifier,
TI(k)Indicate the Tanimoto coefficient value in kth time iteration,
Indicate the Tanimoto coefficient value of jth width classification results image in kth time iteration.
In present embodiment, each classification corresponds to a classification results image, therefore, the j in classification results and j-th
J in image is a variable.
In the present invention, in the selection of spatial filter, there are several types of:
1, Gaussian filter:
It is the most common spatial filter based on Gaussian Profile:
2, Steerable filter device:
Comprising a reference image R and input picture I to be filtered, enabling output image is O, then:
O(xi,yi)=∑ W(x,y)(R(xi,yi))I(xi,yi);
Wherein, (xi,yi) and (x, y) be respectively the coordinate of pixel in correspondence image, W(x,y)(R(xi,yi)) it is by with reference to figure
It is mutually indepedent with I as the kernel function that R is obtained.
3, Gabor filter:
It is a kind of for simulating the filter of human visual system, is generally used for the extraction and identification of texture information.Two
On dimension space domain, which is made of the Gaussian function modulated by sine curve:
G (x, y)=exp (- π [(x-x0)2α2+(y-y0)2β2])·exp(-πi[μ0(x-x0)2+υ0(y-y0)2]) wherein,
(x0,y0)、(μ0,υ0) and (α, β) be respectively location parameter, modulation parameter and scale parameter.
Claims (2)
1. the Classification of hyperspectral remote sensing image method based on feedback arrangement, which is characterized in that its detailed process are as follows:
S1, it will form that format is a × b × N, the hyperspectral image data comprising M atural object classification is as input data Ω substantially
(k), and the number of iterations k=0 is enabled;
S2, in k iteration, classified on hyperspectral image data Ω (k) using support vector machines, classification results areWherein SVM indicates support vector machines, C presentation class, and MAP indicates Maximun Posterior Probability Estimation Method,Point that carries out classifying using support vector machines and generate using Maximun Posterior Probability Estimation Method is indicated for a variable
Class result;
S3, as k > 0, judgementWhether meet stop condition, if it is terminates iteration, at this timeFor final classification as a result, if otherwise executing S4;
S4, basisGenerate bianry image block
In j-th of image, position is that the value of the pixel of (x, y) and the pixel existOn classification it is corresponding, then jth
It is 1 that position, which is the value of the pixel of (x, y), in a image, is otherwise 0;
S5, the bianry image block that S4 is generatedSpace filtering is carried out, filtered gray level image block is
S6, the filtered gray level image block for obtaining S5It is fed back, the hyperspectral image data with input
Ω (k) is stacked:
Then k=k+1 is enabled, and returns and executes S2, until terminating iteration willAs final classification result.
2. the Classification of hyperspectral remote sensing image method according to claim 1 based on feedback arrangement, which is characterized in that S3 institute
State stop condition are as follows:
Using Tanimoto coefficient value TI as the condition for differentiating stopping:
And
Wherein,Indicate setIn element number,
Indicate setIn element number,
Indicate the classification results for carrying out the jth class of classification acquisition in kth time iteration by support vector machines,
Indicate the classification results for carrying out the jth class of classification acquisition in -1 iteration of kth by classifier,
TI(k)Indicate the Tanimoto coefficient value in kth time iteration,
Indicate the Tanimoto coefficient value of jth width classification results image in kth time iteration.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910337636.8A CN110046674A (en) | 2019-04-24 | 2019-04-24 | Classification of hyperspectral remote sensing image method based on feedback arrangement |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910337636.8A CN110046674A (en) | 2019-04-24 | 2019-04-24 | Classification of hyperspectral remote sensing image method based on feedback arrangement |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110046674A true CN110046674A (en) | 2019-07-23 |
Family
ID=67279279
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910337636.8A Pending CN110046674A (en) | 2019-04-24 | 2019-04-24 | Classification of hyperspectral remote sensing image method based on feedback arrangement |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110046674A (en) |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103473557A (en) * | 2013-09-02 | 2013-12-25 | 杭州电子科技大学 | Spatial-spectral integrated hyperspectral remote sensing image classification method |
CN104268579A (en) * | 2014-10-14 | 2015-01-07 | 哈尔滨工业大学 | Hyperspectral remote sensing image classifying method based on hierarchy ensemble learning |
CN105631480A (en) * | 2015-12-30 | 2016-06-01 | 哈尔滨工业大学 | Hyperspectral data classification method based on multi-layer convolution network and data organization and folding |
CN106056157A (en) * | 2016-06-01 | 2016-10-26 | 西北大学 | Hyperspectral image semi-supervised classification method based on space-spectral information |
CN106408034A (en) * | 2016-10-17 | 2017-02-15 | 大连海事大学 | Hyperspectral image classification method based on spatial feature iteration |
CN108038448A (en) * | 2017-12-13 | 2018-05-15 | 河南理工大学 | Semi-supervised random forest Hyperspectral Remote Sensing Imagery Classification method based on weighted entropy |
US20180365820A1 (en) * | 2017-06-19 | 2018-12-20 | ImpactVision, Inc. | System and method for hyperspectral image processing to identify object |
-
2019
- 2019-04-24 CN CN201910337636.8A patent/CN110046674A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103473557A (en) * | 2013-09-02 | 2013-12-25 | 杭州电子科技大学 | Spatial-spectral integrated hyperspectral remote sensing image classification method |
CN104268579A (en) * | 2014-10-14 | 2015-01-07 | 哈尔滨工业大学 | Hyperspectral remote sensing image classifying method based on hierarchy ensemble learning |
CN105631480A (en) * | 2015-12-30 | 2016-06-01 | 哈尔滨工业大学 | Hyperspectral data classification method based on multi-layer convolution network and data organization and folding |
CN106056157A (en) * | 2016-06-01 | 2016-10-26 | 西北大学 | Hyperspectral image semi-supervised classification method based on space-spectral information |
CN106408034A (en) * | 2016-10-17 | 2017-02-15 | 大连海事大学 | Hyperspectral image classification method based on spatial feature iteration |
US20180365820A1 (en) * | 2017-06-19 | 2018-12-20 | ImpactVision, Inc. | System and method for hyperspectral image processing to identify object |
CN108038448A (en) * | 2017-12-13 | 2018-05-15 | 河南理工大学 | Semi-supervised random forest Hyperspectral Remote Sensing Imagery Classification method based on weighted entropy |
Non-Patent Citations (3)
Title |
---|
SHENGWEI ZHONG: "ITERATIVE SUPPORT VECTOR MACHINE FOR HYPERSPECTRAL IMAGE CLASSIFICATION", 《IEEE INTERNATIONAL CONFERENCE IMAGE PROCESSING ICIP》 * |
于纯妍: "基于目标约束与谱空迭代的高光谱图像分类方法", 《光学学报》 * |
马宏斌等: "基于MCMC的遥感图像分类算法研究", 《测绘与空间地理信息》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108573276B (en) | Change detection method based on high-resolution remote sensing image | |
Xia et al. | Spectral–spatial classification of hyperspectral images using ICA and edge-preserving filter via an ensemble strategy | |
CN104966085B (en) | A kind of remote sensing images region of interest area detecting method based on the fusion of more notable features | |
Neal et al. | Measuring shape | |
Ishak et al. | Weed image classification using Gabor wavelet and gradient field distribution | |
CN103971123B (en) | Hyperspectral image classification method based on linear regression Fisher discrimination dictionary learning (LRFDDL) | |
CN107239759B (en) | High-spatial-resolution remote sensing image transfer learning method based on depth features | |
Agrawal et al. | Grape leaf disease detection and classification using multi-class support vector machine | |
CN107392130A (en) | Classification of Multispectral Images method based on threshold adaptive and convolutional neural networks | |
CN106845418A (en) | A kind of hyperspectral image classification method based on deep learning | |
CN104268879B (en) | Building quantity of goods produced Damage assessment method based on remote sensing multispectral image | |
CN109492593A (en) | Hyperspectral image classification method based on principal component analysis network and space coordinate | |
CN107358260A (en) | A kind of Classification of Multispectral Images method based on surface wave CNN | |
CN106650606A (en) | Matching and processing method for face image and face image model construction system | |
Li et al. | A spectral-spatial kernel-based method for hyperspectral imagery classification | |
Ye et al. | Hyperspectral image classification using principal components-based smooth ordering and multiple 1-D interpolation | |
CN109242812A (en) | Image interfusion method and device based on conspicuousness detection and singular value decomposition | |
Ablin et al. | An investigation in satellite images based on image enhancement techniques | |
CN108734200A (en) | Human body target visible detection method and device based on BING features | |
Teng et al. | Adaptive morphological filtering method for structural fusion restoration of hyperspectral images | |
CN101533475A (en) | Method for extracting feature of shape-adaptive neighborhood based remote sensing image | |
CN108256557A (en) | The hyperspectral image classification method integrated with reference to deep learning and neighborhood | |
CN110046674A (en) | Classification of hyperspectral remote sensing image method based on feedback arrangement | |
Zubair | Machine learning based biomedical image analysis and feature extraction methods | |
Sebastian et al. | A novel MRI and PET image fusion in the NSST domain using YUV color space based on convolutional neural networks |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190723 |