CN104331698B - Remote sensing type urban image extracting method - Google Patents
Remote sensing type urban image extracting method Download PDFInfo
- Publication number
- CN104331698B CN104331698B CN201410662642.8A CN201410662642A CN104331698B CN 104331698 B CN104331698 B CN 104331698B CN 201410662642 A CN201410662642 A CN 201410662642A CN 104331698 B CN104331698 B CN 104331698B
- Authority
- CN
- China
- Prior art keywords
- city
- remote sensing
- sensing images
- aster
- data
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention provides a remote sensing type urban image extracting method. The method comprises the following steps: S1, performing characteristics extracting for altitude data of an ASTER VNIR satellite remote sensing image and derived products and a sample of a PALSAR HH/HV satellite remote sensing image; S2, extracting obvious urban and non-urban points based on the spectral characteristics of the urban and non-urban part; S3, performing confidence spreading by the LLGC using the obvious urban and non-urban points as the initial information based on the characteristics distribution feature of data to be classified, so as to obtain an urban confidence map; S4, obtaining the confidence of the whole remote sensing image, weighting and randomly sampling to obtain a training sample; S5, classifying the urban based on SVM, namely, classifying by the SVM method on the basis of the characteristic vectors extracted in the step S1 and the sample data extracted in step S4, and then obtaining a urban map subjected to binarization according to a classification label. With the adoption of the method, the problems of high cost and high time consumption and the like caused by manual sampling in the prior art can be solved.
Description
Technical field
The present invention relates to Remote Sensing Image Processing Technology, more particularly, to a kind of remote sensing images city extracting method.
Background technology
In population in the world, the ratio sustainable growth of urban population.The inhabitants live of half has been had more than at present in city.
The correlational study of urbanization is in aspect importants such as urban planning, social formation analysis, environmental change, Disaster controls.
The geographically relevant information in city is typically grasped by administrative unit, but has certain limitation, including as follows：
Definition between-various countries with regard to " city " has differences；
- more delayed to the dynamic change description in city；
The information sharing of-city is difficult；
The accurate city acquisition of information of-developing country is difficult.
In the current analysis to city map, urbanization correlational study is become based on the city cartographic analysis of remote sensing images
A kind of main way.Remote sensing analysis method is analyzed based on the multispectral characteristic in city, to obtain high-resolution city map.
It does not rely on administrative division definition, remains to being consistent property, be conducive to carrying out Global Urbanization phase between country variant, region
Close analysis.
But, current problems faced is that the city of Different Climatic Zones often differs greatly in multispectral characteristic, Fig. 1-
The 3 pseudo- coloured silk data of ASTER (b1～b3) spectrum for showing three different cities, the region in figure in circle substantially represents city.
As can be seen from the figure the distribution of color characteristic in city differs greatly in three width figures, and non-city is also such.
For the problem, supervised classification method is the main stream approach of the urban area recognition for being currently based on remote sensing images.The method
Training sample is used as by the manual city pixel and non-city pixel chosen in image range, then by supervised classification
Method (such as support vector machines, artificial neural network, decision tree etc.), learns the spectral characteristic in city/non-city and sets up mould
Type, then carries out classification and obtains city map to remotely-sensed data.Fig. 4 shows a training sample collection schematic diagram, and the figure comes
From " the carrying out the automated process of earth city mapping by with reference to ASTER satellite images and GIS data " of document Miyazaki et al.
(Miyazaki et al.,An Automated Method for Global Urban Area Mapping by
Integrating ASTER Satellite Images and GIS Data), the figure shows for researching and developing ground truth number
Understand according to the visualization of collection.According to the figure, by specific human-computer interaction interface, sample point is chosen by hand by research worker.
In general, the multi-spectral remote sensing image city maps processing flow process based on supervised classification method is as follows：
(1) characteristic vector is extracted from multispectral image
Multispectral image typically by multi-spectral remote sensing image dataset representation, for example, is expressed as：
{Imgb, b=1,2 ..., B, B be wave band sum
For the pixel i in image, correlated characteristic is extracted by various methods, generate characteristic vector
(2) city disaggregated model is learnt by training sample
For supervised classification method, known training sample set is needed：
Generic labels (Label) of the L for sample point
(3) model after study is applied to into entire image, obtains the class label (city/non-city) of each pixel,
So as to obtain the city map of city binaryzation.
For city classification problem, the codomain of L is set { -1,1 }.Value represents city for 1, and -1 represents non-city.
There are the following problems for said method：
● the acquisition of training sample is wasted time and energy
The method needs have the research worker of abundant process experience to come hand-manipulated to remote sensing images, and pointwise is chosen.And
And, due to the change of different geographical city spectral characteristic, sample point is typically only suitable for present image and adjacent domain.
● for global city is analyzed, need to sample each city place, manpower and materials consume huge, into
This is high.
The content of the invention
For problems of the prior art, the present invention proposes a kind of new remote sensing images city extracting method, including：
S1), for ASTER VNIR satellite remote sensing images and its sample of derived product altitude data and PALSAR satellite remote sensing images
This, carries out feature extraction；S2), the spectral characteristic priori based on city and non-city, carries out notable city and non-city point
Extract；S3), using the notable city in this part and non-city point as initial information, with reference to the feature distribution characteristic of data to be sorted,
By LLGC methods, confidence level diffusion is carried out, obtain city confidence map；S4), after obtaining the confidence level of whole remote sensing images, with
This weighting carries out stochastical sampling, generates training sample；S5), city classification is carried out based on SVM, including：To be extracted in step S1
Characteristic vector, and based on the sample data extracted in step S4, classified by traditional SVM methods, according to classification
Label obtains the city map of binaryzation.
The method of the present invention is a kind of adaptive remote sensing images city extracting method.Which is based on city in remote sensing images
Spectrum essential information, with reference to the spectral distribution property of data to be sorted, by way of confidence level spreads, realizes training sample
Automatically choose, be then based on the training sample for generating, by the support vector machines method in supervised classification method, to remote sensing
Data are classified, and generate city map.
Description of the drawings
Fig. 1-3 shows the pseudo- coloured silk data of ASTER (b1～b3) spectrum of three different cities；
Fig. 4 shows a training sample collection schematic diagram；
Flow charts of the Fig. 5-6 for the method for the present invention；
Experimental result pictures of the Fig. 7 for the method for the present invention.
Specific embodiment
The handling process of the method for the present invention is as seen in figs. 5-6.The method of the present invention comprehensively uses two kinds of remote sensing
Data：
ASTER (advanced spaceborne heat radiation and reflection measuring set, Advanced Spaceborne Thermal Emission
And Reflection Radiometer) in VNIR (visible ray and near-infrared radiometer, visible and near-
Infrared radiometer) acquired in satellite remote sensing images 4 wave bands, be designated as Aster respectivelyb1～Asterb4, frequency spectrum
Scope be respectively 0.52-0.60 μm, 0.63-0.69 μm, 0.76-0.86 μm (nadir view), 0.76-0.86 μm
(backward scan), and related derivative product altitude data (slope data, be designated as slope)；And
PALSAR (L-band phased array synthetic aperture radar, Phased Array L-band Synthetic Aperture
Radar) HH, HV satellite remote sensing images (2 wave bands, be designated as hh, hv), and the HH ripples after processing are corrected to angle of incidence
Segment data (is designated as hhcor)。
With reference to Fig. 6, in step S1, for ASTER VNIR satellite remote sensing images and its derived product altitude data and
The sample of PALSAR HH/HV satellite remote sensing images, carries out feature extraction.
In the present invention, 12 features are calculated altogether and have used, wherein 8 original class input data (Asterb1~
Asterb4、slope、hh、hv、hhcor) be also feature a part, this is 8 features.In addition to original input data, the present invention
In also use extra 4 kinds of features：NDVI (normalized differential vegetation index, Normalized Difference Vegetation
Index), NDWI (normalization aqua index, Normalized Difference Water Index), hhsubAnd hhent.Wherein,
NDVI and NDWI are two kinds of general features, it is adaptable to distinguish vegetation and water body, all in accordance with ASTER VNIR satellite remote sensing images
Data Asterb1～Asterb4It is calculated.hhsubAccording to PALSAR HH satellite remote sensing images data hh and hhcorCalculate
Arrive, hhsubThere is certain effect to distinguishing mountain range (non-city).hhentCalculated according to PALSAR HH satellite remote sensing images data hh
Obtain, hhentFor describing the abundant degree of texture information, the texture information of general significantly non-city part is all less.This 4 kinds
The computational methods of feature such as formula (1)-(4)：
hhsub=| hh-hhcor| (3)
hhent=EntropyFilt (hh) (4)
Wherein, EntropyFilt () represents entropy filtering (referring to Eddins, S.L.；Gonzalez,R.；
Woods,R.Digital image processing using Matlab.Princeton Hall Pearson
2004), in the present invention, EntropyFilt () acts on (analysis on hh images to Education Inc., New Jersey
Window size is 15x15 pixels).
With reference to Fig. 6, in step S2, for ASTER VNIR satellite remote sensing images and PALSAR HH satellite remote sensing images, base
In the spectral characteristic in city/non-city, carry out notable city/non-city point and extract.
According to priori, so-called notable city/non-city point is referred to, which has more clearly spectral characteristic, can
By the in addition clear and definite discrimination of simple many assembled classifiers.Significantly the spectral characteristic of city/non-city point has preferably accurately
Property, and suitable for the data of different geographical.
But the point of significance negligible amounts in general, extracting, can not represent the real features of city/non-city data
Distribution.Therefore, point of significance can not also be needed further to be combined and complete with the feature distribution of input data directly as training sample
It is kind.
The present invention to point of significance extract principle be, based on calculate satellite remote sensing images in binaryzation mask image, so
Notable city and non-city point are determined with two to the form opening and closing operation of binaryzation mask image afterwards, wherein, binaryzation mask figure
The comparison threshold value of picture NDVI, DNWI, hh according to the satellite remote sensing imagessub, hh and/or hhentIt is calculated.
The concrete grammar that the point of significance of the present invention is extracted is referring to equation below (5)～(18).For non-city, 6 are produced
Mask (binaryzation mask image, for representing the non-city point of significance for meeting specified conditions).For mask1～mask6If,
Condition meets, then the value of mask is 1, is otherwise 0.mask1And mask2Set according to NDVI and NDWI respectively.If the value of pixel
It is higher than meansigma methodss, then the pixel is marked, vegetation region and pool is represented.mask3Before and after recognizing the correction of hh wave bands angle of incidence
Non- city indicated by difference, can effectively recognize mountain range (non-city) part.Further, since the building in city is with higher
Reflectance, rather than city partial reflectance is relatively low, therefore the value in non-city will be less than equal in PALSAR HH images
Value, mask4Setting i.e. be based on this.According to analysis slope data and the abundant degree of texture, mask5And mask6For with give
Fixed threshold value compares.Based on experience, threshold value thresh5And thresh6It is respectively set to 15 and 4.5.
Then, with specific morphology operations template (10x10 sizes, it is complete 1) to mask1～mask6The two-value form for carrying out
Opening and closing operation (is designated as MorphFilt ()), for refining mask.Its objective is that removing isolated non-city (that includes a small amount of
Pixel), and by merge these refinement after mask obtain masknonurban, such as shown in formula (15).
Prediction to city is based on and mask4Similar mode, with masknonurbanMerge and carry out identical form opening and closing
Mask is obtained after computingurban, such as shown in formula (16)-(18).
mask1=NDVI ＞ thresh1 (5)
thresh1=mean (NDVI)+std (NDVI) (6)
mask2=NDWI ＞ thresh2 (7)
thresh2=mean (NDWI)+std (NDWI) (8)
mask3=hhsub＞ thresh3 (9)
thresh3=mean (hhsub)+std(hhsub) (10)
mask4=hh ＜ thresh4 (11)
thresh4=mean (hh)-std (hh) (12)
mask5=slope ＞ thresh5 (13)
mask6=hhent＜ thresh6 (14)
mask7=hh ＞ thresh7 (16)
thresh7=mean (hh)+std (hh) (17)
maskurban=MorphFilt (mask7∩Not(masknonurban)) (18)
Here, mean () represents the meansigma methodss of input picture, and std () represents standard variance, and Not () represents binary value
Reciprocal value.Wherein threshold value thresh5And thresh6Set for empirical value, other thresh are to be calculated according to present analysis image
The adaptive threshold for drawing.
By above-mentioned steps, you can obtain notable city/non-city point.
With reference to Fig. 6, in step S3, using the notable city in this part/non-city point as initial information, with reference to data to be sorted
Feature distribution characteristic, by LLGC (Learning with Local and Global Consistency) method (referring to
Based on local and the learning method of global coherency, Zhou, D.；Bousquet,O.；Lal,T.N.；Weston,J.；B.Learning with local and global consistency.Advances in neural
Information processing systems 2004,16,321-328.), confidence level diffusion is carried out, city confidence is obtained
Figure.So-called unfiled data refer to also there be not the data that determination is city or non-city point.
In the present invention, 3 wave band Aster of ASTER VNIR satellite remote sensing imagesb1～Asterb3Image by conduct
Cluster feature is simultaneously merged into a coloured image Asterrgb。
In one embodiment, based on LLGC methods, using notable city/non-city point as initial value, feature based arrow
Amount similarity between any two builds diffusion strength criterion, and the confidence information in city/non-city is diffused to unfiled data
Feature space in.Specifically include：
1) LLGC methods input is image AsterrgbSample set X={ x1, x2..., xN, wherein xiFor vector, represent
The feature of sample.N is sample size.In addition label matrix F is set as Nx2 matrixes, 1,2 elements of the i-th row are represented i-th respectively
Sample belongs to city, the confidence level in non-city.
2) initial value of label matrix F is set by the notable city/non-city point for extracting.City point is be expert to be set to
[10], non-city is set to [01], and other are set to [00].
3) dispersion relation two-by-two between sample, such as formula are provided by diffusion matrix W and normalization diffusion matrix S
(19) shown in-(21)：
S=D-1/2WD-1/2 (21)
Wherein, dist () is scalar function, represents characteristic vector xiAnd xjDifference. constant σ represents diffusion kernel size.D
For the diagonal matrix of NxN, i-th row element sum of i-th diagonal element for W.
4) the final label matrix F by LLGC methods, after diffusion*It is expressed as shown in formula (22)：
F*=(1- α) (I- α S)-1F(0) (22)
But, original LLGC methods due to when the reason of operand is actually used efficiency comparison it is low.The remote sensing analyzed
The pixel number of image full figure is typically not less than 10,000,000, i.e. (N>107), W, S are the non-sparse matrix of NxN sizes, it is difficult to
It is directly realized by by conventional method and (generates the 10 of a double type in theory7x107Matrix needs about 700TB memory spaces).
Therefore, the present invention proposes a kind of improved LLGC methods, and input picture is carried out quantization index process, and amendment is corresponding
Computing and increase correlation mapping process.Operand can be greatly reduced and preferable diffusion effect is kept.
In a preferred embodiment, above-mentioned LLGC methods are improved, is specifically included：
1) by image AsterrgbQuantization transform is carried out, is changed into thumbnail, maximum color index number M is less than N, usually
Far below N, in one example, for example M is set to 300.
2) sample set is changed into X={ (xi, ni), i=1,2 ..., M }, wherein niFor the quantity of i-th sample.With identical
The pixel of color index is considered as same sample, niPixel then corresponding to the index is total.
3) to the sample set application LLGC algorithms after conversion, confidence level diffusion is carried out, obtains city confidence map.
Specifically, each calculation procedure in Jing originals LLGC algorithms is transformed to formula (23)-(25)：
WI, j=exp [- dist (xi, xj)/2σ2] (23)
Now incidence matrix W, S is the matrix of MxM sizes, i.e., less than 300x300 sizes, therefore operand no longer structure
It is a problem.Meanwhile, set up index and only cause to diffusion process slight impact, therefore performance to be close to former algorithm.
Based on maskurban, find the connected city of maximum and subgraph is selected according to its bounding rectangles.In the subgraph
The quantity of city/non-city pixel be averaged, and apply LLGC algorithms in the region.According to the picture with same index color
Element also has the rule of identical confidence level, the city confidence map of subgraph is marked again and is returned on whole figure.In this way, it is improved
LLGC algorithms can efficiently and effectively generate city confidence map.
With reference to Fig. 6, in step S4, after the confidence level of all data of full figure for obtaining whole remote sensing images, with this weight into
Row stochastical sampling, generates training sample.
In step S4, the training data (such as city/non-city picture for further classifying is obtained by weight sampling
Element), wherein the confidence level of each pixel is used as weight.Specific method of weighting includes step S401-S403：
S401) the confidence level C in the city to each pixel/non-cityu、CnuNormalization, such as so that this two sums are
1。
S402) for city, choose CuMore than first threshold (such as 0.5) pixel as Candidate Set, then using with
The machine method of sampling (a kind of such as document " sigma-t method of diffusion for visual tracking ", Isard M, Blake
A.Condensation—conditional density propagation for visual tracking[J]
.International journal of computer vision,1998,29(1):5-28.), with CuAs weight, choose
City training sample (sample points are, for example, 500).
S403) for non-city, choose CnuMore than Second Threshold (such as 0.5) pixel as Candidate Set, then with Cnu
As weight, non-city training sample (sample points are, for example, 300) is obtained using identical step.
With reference to Fig. 6, in step S5, based on SVM (support vector machine, Support Vector Machine) method
(Cristianini,N.；Shawe-Taylor,J.An introduction to support vector machines and
other kernel-based learning methods；Cambridge university press, 2000) carry out city
Classification, specifically includes：10 category features extracted using in step S1 are used as characteristic vector, and the training sample extracted in step S4
Based on data, classified by traditional SVM methods, the city map of binaryzation is obtained according to tag along sort.It is used
10 category features be Asterb1～Asterb4、slope、NDVI、NDWI、hh、hv、hhent。
For the method for the present invention, also tested, experimental design is as follows：
● to verify the effectiveness of this method, the city of 75 Different Climatic Zones in global range is analyzed；
● total 100 points or so city/non-city point is chosen in each city by hand to be used to verify as true value；
● map compares；
● compare with Remote Sensing Products MCD12Q1 whole world city map；
● using half true value point as training sample, city map is obtained by the sorting technique in SVM then and is carried out
Relatively.
Evaluation criterion adopts general quantitatively evaluating standard, formula (26)-(29) to represent user's precision, Producer essence respectively
Degree, overall accuracy and kappa coefficients：
In formula, each symbol implication is as shown in table 1 below, and table 1 illustrates the structure for city/non-city classification results matrix.
1 city of table/non-city classification results matrix
Experimental result is as shown in table 2 and Fig. 7.Table 2 shows the estimation accuracy of city map.
The estimation accuracy of 2 city map of table
Fig. 7 shows the comparative result of the city map in four places, and (a) in Fig. 7 is ASTER/VNIR pseudo color coding hologram figures,
B () is PALSAR pseudo color coding hologram figures, (c) the city map obtained for the method for the present invention, is (d) MCD cities map, is (e) SVM
City map.As a result show, the city Map quality for being generated is better than MCD, be close to supervised classification SVM methods performance.
The method of the present invention combines priori and confidence level method of diffusion, can between good conformity different geographical city light
Spectrum change.
The Map quality that this method is generated is better than MCD, is very close to supervised classification SVM method performances.
Present method solves the problems such as high cost of artificial sample needed for supervised classification method, height take, automatically enters
Row is processed and excellent performance, is expected to be widely applied in the related applications such as global city drawing and its Dynamic profiling description.
Although describing the present invention already in connection with the illustrative embodiments for being presently considered to feasible, it will be understood that, this
Invention is not limited to disclosed illustrative embodiments, but on the contrary, is included in claims it is contemplated that covering
Spirit and scope in various modifications and equivalent arrangements.
Claims (7)
1. a kind of remote sensing images city extracting method, it is characterised in that include：
S1), it is distant for ASTER VNIR satellite remote sensing images and its derived product altitude data and PALSAR HH, HV satellites
The sample of sense image, carries out feature extraction, and the feature of extraction includes：Input data Asterb1～Asterb4、slope、hh、hv、
hhcor、NDVI、NDWI、hhsubAnd hhent, wherein, Asterb1～Asterb4Represent 4 of ASTER VNIR satellite remote sensing images
The data of wave band, slope represent the derived product altitude data of ASTER satellite remote sensing images, and hh and hv represents PALSAR respectively
The data of the HH and HV wave bands of satellite remote sensing images, hhcorRepresent that the angle of incidence of PALSAR satellite remote sensing images is corrected process
HH wave band datas afterwards, NDVI represent normalized differential vegetation index, and NDWI represents normalization aqua index, hhsubAccording to PALSAR HH
Satellite remote sensing images data hh and hhcorIt is calculated, for distinguishing the data in non-city, hhentIt is distant according to PALSAR HH satellites
Sense view data hh is calculated, for describing texture information；
S2), for ASTER VNIR satellite remote sensing images and PALSAR HH satellite remote sensing images, based on city and non-city
Spectral characteristic, carry out notable city and non-city point and extract；
S3), using the notable city in this part and non-city point as initial information, with reference to the feature distribution characteristic of data to be sorted,
By LLGC methods, confidence level diffusion is carried out, obtain city confidence map；
S4), after obtaining the confidence level of whole remote sensing images, being weighted with this carries out stochastical sampling, generates training sample；
S5), city classification is carried out based on SVM, including：With the characteristic vector extracted in step S1, and extract in step S4
Based on sample data, classified by traditional SVM methods, the city map of binaryzation is obtained according to tag along sort.
2. remote sensing images city according to claim 1 extracting method, it is characterised in that in step s 2, based on calculating
Binaryzation mask image in satellite remote sensing images, then determines with the two-value form opening and closing operation of binaryzation mask image notable
City and non-city point, wherein, the comparison threshold value of binaryzation mask image according to the NDVI of the satellite remote sensing images, DNWI,
hhsub, hh and/or hhentIt is calculated.
3. remote sensing images city according to claim 1 extracting method, it is characterised in that step S3 also includes：
S301) by 3 wave band Aster of ASTER VNIR satellite remote sensing imagesb1～Asterb3Image by as cluster feature
And it is merged into a coloured image Asterrgb；
S302) based on LLGC methods, using notable city/non-city point as initial value, feature based vector between any two similar
Property build diffusion strength criterion, the confidence information in city/non-city is diffused in the feature space of unfiled data.
4. remote sensing images city according to claim 3 extracting method, it is characterised in that step S3 also includes：
S304) by image AsterrgbQuantization transform is carried out, is changed into thumbnail, setting maximum color index quantity M is less than as former state
This quantity N；
S305) for the new samples collection with new samples quantity after conversion in S304, labelling has the picture that same color is indexed
It is same sample that element is；
S306) to the new samples collection in S305, feature based vector similarity between any two builds diffusion strength criterion, is based on
LLGC methods, the confidence information in city/non-city is diffused in the feature space of unfiled data.
5. remote sensing images city according to claim 1 extracting method, it is characterised in that step S4 also includes：
S401 city), to each pixel and the confidence level C in non-cityu、CnuNormalization；
S402), for city, choose CuMore than first threshold pixel as Candidate Set, then using stochastical sampling method, with
CuAs weight, city training sample is chosen；
S403), for non-city, choose CnuMore than Second Threshold pixel as Candidate Set, then with CnuAs weight, obtain
Non- city training sample.
6. remote sensing images city according to claim 5 extracting method, it is characterised in that in step S401 so that put
Reliability Cu、CnuTwo sums are 1, and in step S402 and S403, the first threshold and Second Threshold are set to 0.5.
7. remote sensing images city according to claim 1 extracting method, it is characterised in that in step s 5：Used
Vector characteristic is Asterb1～Asterb4, slope, NDVI, NDWI, hh, hv and hhent。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410662642.8A CN104331698B (en) | 2014-11-19 | 2014-11-19 | Remote sensing type urban image extracting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410662642.8A CN104331698B (en) | 2014-11-19 | 2014-11-19 | Remote sensing type urban image extracting method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104331698A CN104331698A (en) | 2015-02-04 |
| CN104331698B true CN104331698B (en) | 2017-05-03 |
Family
ID=52406420
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410662642.8A Active CN104331698B (en) | 2014-11-19 | 2014-11-19 | Remote sensing type urban image extracting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104331698B (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104794478B (en) * | 2015-05-04 | 2017-12-19 | 福建师范大学 | A kind of building extracting method for being used in remote sensing image have uniform spectral characteristic |
| CN105335697A (en) * | 2015-09-18 | 2016-02-17 | 黄山学院 | Method and device for measuring and calculating ancient dwelling similarity on basis of satellite images |
| CN105388475B (en) * | 2015-10-27 | 2018-01-19 | 中国热带农业科学院橡胶研究所 | A kind of method that high-biomass sugarcane is removed in PALSAR forest classified results |
| CN106127121B (en) * | 2016-06-15 | 2019-03-08 | 四川省遥感信息测绘院 | A kind of built-up areas intelligence extracting method based on nighttime light data |
| CN106897674B (en) * | 2017-01-20 | 2019-07-26 | 北京理工大学 | A kind of in-orbit remote sensing images city detection method based on JPEG2000 code stream |
| CN107067003B (en) * | 2017-03-09 | 2021-09-21 | 百度在线网络技术（北京）有限公司 | Region-of-interest boundary extraction method, device, equipment and computer storage medium |
| CN109858450B (en) * | 2019-02-12 | 2020-10-23 | 中国科学院遥感与数字地球研究所 | Ten-meter-level spatial resolution remote sensing image town extraction method and system |
| CN110175638B (en) * | 2019-05-13 | 2021-04-30 | 北京中科锐景科技有限公司 | Raise dust source monitoring method |
| CN110189328B (en) * | 2019-06-11 | 2021-02-23 | 北华航天工业学院 | Satellite remote sensing image processing system and processing method thereof |
| CN110321808B (en) * | 2019-06-13 | 2021-09-14 | 浙江大华技术股份有限公司 | Method, apparatus and storage medium for detecting carry-over and stolen object |
| CN111521191A (en) * | 2020-04-20 | 2020-08-11 | 中国农业科学院农业信息研究所 | Mobile phone user moving path map matching method based on signaling data |
| CN111598101B (en) * | 2020-05-25 | 2021-03-23 | 中国测绘科学研究院 | Urban area intelligent extraction method, system and equipment based on remote sensing image scene segmentation |
| CN111832628A (en) * | 2020-06-23 | 2020-10-27 | 东南大学 | Old city edge area range defining method based on double data source fusion |
| CN113076963B (en) * | 2021-06-07 | 2021-08-13 | 腾讯科技（深圳）有限公司 | Image recognition method and device and computer readable storage medium |
| CN115620170A (en) * | 2022-12-16 | 2023-01-17 | 自然资源部第三航测遥感院 | Multi-source optical remote sensing image water body extraction method based on SegFormer |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1472634A (en) * | 2003-05-16 | 2004-02-04 | 上海交通大学 | High spectrum remote sensing image combined weighting random sorting method |
| CN102110227A (en) * | 2010-11-24 | 2011-06-29 | 清华大学 | Compound method for classifying multiresolution remote sensing images based on context |
| CN104036294A (en) * | 2014-06-18 | 2014-09-10 | 西安电子科技大学 | Spectral tag based adaptive multi-spectral remote sensing image classification method |
-
2014
- 2014-11-19 CN CN201410662642.8A patent/CN104331698B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1472634A (en) * | 2003-05-16 | 2004-02-04 | 上海交通大学 | High spectrum remote sensing image combined weighting random sorting method |
| CN102110227A (en) * | 2010-11-24 | 2011-06-29 | 清华大学 | Compound method for classifying multiresolution remote sensing images based on context |
| CN104036294A (en) * | 2014-06-18 | 2014-09-10 | 西安电子科技大学 | Spectral tag based adaptive multi-spectral remote sensing image classification method |
Non-Patent Citations (1)
| Title |
|---|
| "基于支持向量机(SVM)特征加权/选择的光谱匹配算法";王圆圆 等;《光谱学与光谱分析》;20090331;第29卷(第3期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104331698A (en) | 2015-02-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104331698B (en) | Remote sensing type urban image extracting method | |
| CN111860612B (en) | Unsupervised hyperspectral image hidden low-rank projection learning feature extraction method | |
| CN110399909B (en) | Hyperspectral image classification method based on label constraint elastic network graph model | |
| Li et al. | Graph-feature-enhanced selective assignment network for hyperspectral and multispectral data classification | |
| Huang et al. | A comparative study of spatial approaches for urban mapping using hyperspectral ROSIS images over Pavia City, northern Italy | |
| Im et al. | A change detection model based on neighborhood correlation image analysis and decision tree classification | |
| Lu et al. | A survey of image classification methods and techniques for improving classification performance | |
| Zhuo et al. | Cloud classification of ground-based images using texture–structure features | |
| CN109063754B (en) | Remote sensing image multi-feature joint classification method based on OpenStreetMap | |
| CN111368712A (en) | Hyperspectral image disguised target detection method based on deep learning | |
| Chen et al. | The application of the tasseled cap transformation and feature knowledge for the extraction of coastline information from remote sensing images | |
| CN110458192B (en) | Hyperspectral remote sensing image classification method and system based on visual saliency | |
| CN109635726B (en) | Landslide identification method based on combination of symmetric deep network and multi-scale pooling | |
| Dong et al. | A shadow constrained conditional generative adversarial net for SRTM data restoration | |
| Kakooei et al. | A two-level fusion for building irregularity detection in post-disaster VHR oblique images | |
| Hasanlou et al. | A sub-pixel multiple change detection approach for hyperspectral imagery | |
| Tang et al. | Improving cloud type classification of ground-based images using region covariance descriptors | |
| Guan et al. | Partially supervised hierarchical classification for urban features from lidar data with aerial imagery | |
| Zhang et al. | Multi-feature embedded learning SVM for cloud detection in remote sensing images | |
| Luo et al. | Discriminative supervised neighborhood preserving embedding feature extraction for hyperspectral-image classification | |
| Xu et al. | Impervious surface extraction in imbalanced datasets: integrating partial results and multi-temporal information in an iterative one-class classifier | |
| Matsuoka et al. | Automatic detection of stationary fronts around Japan using a deep convolutional neural network | |
| Dong et al. | A cloud detection method for GaoFen-6 wide field of view imagery based on the spectrum and variance of superpixels | |
| Kathirvelu et al. | Spectral unmixing based random forest classifier for detecting surface water changes in multitemporal pansharpened Landsat image | |
| Menze et al. | Multitemporal fusion for the detection of static spatial patterns in multispectral satellite images—With application to archaeological survey |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |