CN107203790A - Utilize the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of two stage sampling model - Google Patents
Utilize the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of two stage sampling model Download PDFInfo
- Publication number
- CN107203790A CN107203790A CN201710488761.XA CN201710488761A CN107203790A CN 107203790 A CN107203790 A CN 107203790A CN 201710488761 A CN201710488761 A CN 201710488761A CN 107203790 A CN107203790 A CN 107203790A
- Authority
- CN
- China
- Prior art keywords
- noctilucence
- classification
- result
- remote sensing
- precision
- 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
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
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
- G06F18/2193—Validation; Performance evaluation; Active pattern learning techniques based on specific statistical tests
-
- 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
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model, its this method specifically includes following steps：Step S1, realizes in NPP/VIIRS series datas and the built-up areas of Chinese land area is extracted, obtain not verified extraction result；Step S2, sets up two stage sampling model, completes the sampling to evaluation region, obtains believable inspection data collection；Step S3, using error matrix verification method, the preliminary extraction to noctilucence data carries out subregion checking, and finally gives the precision evaluation of all noctilucence data classification results.Utilize the noctilucence Classification in Remote Sensing Image Accuracy Assessment of two stage sampling model, complete the system to noctilucence remotely-sensed data, effectively realize the precision evaluation to large area noctilucence Classification in Remote Sensing Image result, outcome quality is extracted to noctilucence remote sensing cities and towns to examine, it can solve the problem that conventional checking collection is very few, the incomplete problem of evaluating, has reached a kind of comprehensively accurate the result.
Description
Technical field
It is a kind of utilization two stage sampling model specifically the present invention relates to Remote Sensing Image Quality inspection technology field
Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment.
Background technology
Noctilucence remotely-sensed data can detect the low-intensity light that even small-scale settlement place of urban lighting etc. is sent, and by its
Made a distinction with dark rural background, indirectly reflect the distribution in cities and towns, provided for the city detection research of large scale
New approaches.But the factors such as the spills-over effects due to noctilucence, may cause cities and towns area extraction result when extracting cities and towns according to noctilucence
The problems such as more than actual value, therefore must carry out quality verification to extracting result before application.
Traditional verification method only extracts city such as in the research of the Yangtze River Delta urbanization general layout area attribute of result
Contrast verification has been carried out with national statistics data；Chosen in Chinese city expansion research using Landsat data as checking collection
Systematically verified using error matrix to extracting result as training set 15 big and medium-sized cities；Referred to using landscape shape
The shape of number, concentration class index, edge area than extracting result with 4 quantitative target checking noctilucence remotely-sensed datas of connectivity index
Area similarity degree；1014 sample points are extracted by the way of random sampling result progress is extracted to the noctilucence data in Zhejiang Province
Checking.
Conventional noctilucence remote sensing light area, which is extracted in result precision evaluation, has subproblem：(1) checking collection chooses very few, point
Mode for cloth is unreasonable, lacks representativeness compared with totally extracting region；(2) the checking evaluation index chosen is very few, not comprehensive, no
Extraction result can really be reflected；(3) evaluation method not system, it is only relative with national statistics numerical value to illustrate to extract result
Spatial distribution problem.
Chinese patent literature CN201110230310.9, the applying date 20110812, patent name is：One kind analysis landscape is special
Levy on remote sensing classification patch accuracy influence method, including Step 1: obtain data, including to raw video carry out data standard
Change is handled；Step 2: image recognition, including classification and post-classification comparison are carried out to the data that step one is obtained, wherein sorted
Include the determination of classification patch in journey；Step 3: step 2 is identified obtained data carries out classification patch and true value
Correlation calculations between data figure spot, carry out regression curve fitting and set up regression model, including expression Landscape Characteristics
Landscape index is defined；Step 4: statistical check is carried out to the regression model that step 3 is set up, while dividing landscape index expression
Class error is evaluated.
Above-mentioned patent document carries out the theoretical foundation of precision evaluation and analysis using the space characteristics of Classification in Remote Sensing Image, using distant
The Landscape Characteristics description and expression that are carried out to nicety of grading of sense classification patch, be the application related to land cover pattern thematic map and
Research provides foundation and guidance.But on a kind of noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model, it is complete
The system of paired noctilucence remotely-sensed data, the precision evaluation to large area noctilucence Classification in Remote Sensing Image result is effectively realized, it is distant to noctilucence
Feel cities and towns and extract outcome quality inspection, can solve the problem that conventional checking collects very few, the incomplete problem of evaluating, reached one
Plant accurate the result comprehensively.Technical scheme then without corresponding open.
In summary, need a kind of noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model badly, complete to night
The system of light remotely-sensed data, the precision evaluation to large area noctilucence Classification in Remote Sensing Image result is effectively realized, to noctilucence remote sensing cities and towns
Extract outcome quality to examine, can solve the problem that conventional checking collects very few, the incomplete problem of evaluating, reached a kind of comprehensive
Accurate the result.And had not been reported at present on this method.
The content of the invention
The purpose of the present invention is that there is provided a kind of noctilucence remote sensing of utilization two stage sampling model for deficiency of the prior art
Nicety of grading evaluation method, completes the system to noctilucence remotely-sensed data, effectively realizes to large area noctilucence Classification in Remote Sensing Image result
Precision evaluation, to noctilucence remote sensing cities and towns extract outcome quality examine, can solve the problem that it is conventional checking collection it is very few, evaluating is not
Comprehensively the problem of, a kind of comprehensively accurate the result is reached.
To achieve the above object, the present invention is adopted the technical scheme that：
The Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of a kind of utilization two stage sampling model, it is characterised in that should
Method specifically includes following steps：
Step S1, realizes in NPP/VIIRS series datas and the built-up areas of Chinese land area is extracted, obtain not verified
Extract result；
Step S2, sets up two stage sampling model, completes the sampling to evaluation region, obtains believable inspection data collection；
Step S3, using error matrix verification method, the preliminary extraction progress subregion checking to noctilucence data, and final
To the precision evaluation of all noctilucence data classification results.
As a kind of perferred technical scheme, the step S2 comprises the following steps：
Step S21：Divided with reference to the East Coastal of China, middle part, western three main economic region, on the Chinese land of covering
Partial image is extracted in the images of Landsat 8, the sampling collection in three regions is respectively obtained, first order sampling is completed；
Step S22：Chosen respectively in East Coastal, middle part, the corresponding each scape image in western three regions and examine pixel,
The result set of second level sampling is obtained, second level sampling is completed；
Step S23：The property value of checking pixel is extracted in the high-precision images of Landsat 8, judges that the pixel belongs to
Noctilucence area or non-noctilucence area.
As a kind of perferred technical scheme, the step S21 first order methods of samplings are：It is first according to this grade sampling inspection
Model is tested, width figure is extracted from the images of Landsat 8 of covering China's Mainland as sample, in China east coastal waters, middle part, west
It is sampled in portion's three main economic region by different weights, when the map sheet for being generally n verifies collection, by examining each sample
The quality of map sheet infers the quality level of atlas by the gross.
As a kind of perferred technical scheme, also include in the step S22 according to NDVI, NDBI to Landsat images
Classified, the urban area in a wide range of interior determination image；The region having a common boundary simultaneously to cities and towns and non-cities and towns is using visual
The mode of interpretation is accurately determined, accurate extraction result is finally reached in Landsat images, second level sampling is obtained.
As a kind of perferred technical scheme, comprise the following steps in step S3：
Step S31：The checking collection that the noctilucence classification of remote-sensing images result and two stage sampling model of stacked vector quantization are obtained；Step
Rapid S32：Using error matrix verification method, respectively three regions are carried out with the precision evaluation of classification results；Step S33：With reference to
The result of three regions, carries out Extrapolation, finally gives credible reliable global classification precision evaluation the result.
As a kind of perferred technical scheme, in step S3 also by choose overall accuracy, kappa coefficients, user's precision,
Producer's precision, misclassification error, leakage point error parameter, systematically evaluate noctilucence Remotely sensed acquisition result.
As a kind of perferred technical scheme, will be respectively in Chinese three economic zones by choosing in the step S3
Overall accuracy, kappa coefficients, user's precision, producer's precision, misclassification error, leakage point error parameter, systematically evaluate noctilucence distant
The noctilucence Remotely sensed acquisition precision that result method tries to achieve each region is extracted in sense, and Chinese land is gone out in conjunction with the weight computing in each region
The precision result of ground region noctilucence Remotely sensed acquisition.
The invention has the advantages that：
1st, the side that a kind of Chinese land noctilucence Classification in Remote Sensing Image precision using two stage sampling model of the invention is evaluated
Method, the theory of two stage sampling is utilized for great Qu surface areas, the images of all Landsat 8 from the Chinese land area of covering
In, the checking data that data are appropriate, be reasonably distributed are have chosen, the reliability of subsequent authentication result can be effectively ensured.
2nd, the evaluation method chooses error matrix and carries out statistical analysis, have chosen overall accuracy (oa), kappa coefficients, uses
Family precision, producer's precision, misclassification error, leakage point error are as evaluating, from different perspectives, comprehensive without demand system
Classification results are evaluated.
3rd, this method is less for conventional night-light checking research, often carries out numerical value contrast not with national statistics data
Foot, from ground angle, sampling is sampled in space angle, more accurately precision evaluation is realized.
Brief description of the drawings
Accompanying drawing 1 is the flow chart of two stage sampling verification method of the present invention.
Accompanying drawing 2 is noctilucence classification of remote-sensing images result result of the present invention.
Accompanying drawing 3 is the Chinese land area Landsat8 distribution maps of the present invention.
Accompanying drawing 4 is first order sampling verification collection distribution map of the present invention.
Accompanying drawing 5 is sampling verification collection distribution map in the second level of the present invention.
Embodiment
The embodiment that the present invention is provided is elaborated below in conjunction with the accompanying drawings.
Embodiment 1
Noctilucence remotely-sensed data is relative to data such as general visible remote sensing, when being extracted to cities and towns, only as an inter-species
The remotely-sensed data connect.This research is directed to the phenomenon that application is gently verified again in conventional noctilucence remote sensing application, to noctilucence in noctilucence remote sensing
Region is extracted, while having carried out complete precision test research for extracting result, finally proposes a whole set of complete
Extraction, research model.
The flow chart of this research method such as Fig. 1, mainly includes 3 parts composition：
(1) noctilucence classification of remote-sensing images.
Chinese land NPP/VIIRS noctilucence remote sensing images are classified using support vector machine classifier, precision is obtained
Noctilucence area to be verified and the class classification results of non-noctilucence area two；
(2) two stage sampling model is set up.
1. East Coastal, middle part, the western three main economic region for combining China are divided, on the Chinese land of covering
Partial image is extracted in the images of Landsat 8, the sampling collection in three regions is respectively obtained, first order sampling is completed；Wherein, three
Macro-economic region domain is the East Coastal region that the Chinese government divides：It is Liaoning, Beijing, Tianjin, Hebei, Jiangsu, Zhejiang, Fujian, wide
East, Taiwan, Hong Kong, Macao, Guangxi；Central region：Heilungkiang, Jilin, the Inner Mongol, Shanxi, Henan, Hubei, Hunan, Jiangxi；
Western area：Shaanxi, Gansu, Ningxia, Sichuan, Guizhou, Yunnan, Qinghai, Xinjiang, Tibet；
Because light area reflects the activity of the mankind in noctilucence remote sensing image, therefore model need to be examined simultaneously in primary sampling
Consider the factor of space and the density of population.It is that each region sets a weights that this research, which combines spatial area, the density of population, is carried out
The sampling of different proportion.
The images of Landsat 8 on the Chinese land of covering are distributed by world's wide-angle reference system -2, and the Chinese land of covering there are about
536 scapes.
2. chosen respectively in East Coastal, middle part, the corresponding each scape image in western three regions and examine pixel, obtain the
The result set of two stage sampling, completes second level sampling；
3. the property value of checking pixel is extracted in the high-precision images of Landsat 8, judges that the pixel belongs to noctilucence area
Or non-noctilucence area；
For examining the attribute of pixel to determine, according to first according to NDVI, NDBI in the high-precision images of Landsat 8
Landsat images are classified, the urban area in a wide range of interior determination image.Cities and towns and non-cities and towns are had a common boundary simultaneously
Region is accurately determined by the way of visual interpretation, accurate extraction result is finally reached in the images of Landsat 8, judging should
Pixel belongs to noctilucence area or non-noctilucence area.
(3) classification results precision evaluation.
1. it is stacked the checking collection that the noctilucence classification of remote-sensing images result and two stage sampling model of vector quantization are obtained；2. using mistake
Three regions are carried out the precision evaluation of classification results by poor matrix verification method respectively；3. the result of three regions is combined,
Extrapolation is carried out, credible reliable global classification precision evaluation the result is finally given.Concrete model is designed and decision rule
It is as follows：
(1) two stage sampling model
In this paper second-level model Samplings, first order sampling inspection model is to cover China's Mainland region Landsat 8
" map sheet " be sampling unit, sampling inspection model in the second level is single for sampling with " light area and non-light area area " in map sheet
Member.
1) selection of first order sampling model
The sampling unit of first order sampling inspection model is " map sheet " (Map).In first order sampling inspection, it is first according to
This grade of sampling inspection model, extracts n width figure as sample, by examining from the images of Landsat 8 of covering China's Mainland
The quality of each Sample map area infers the quality level of atlas by the gross.
Because light area reflects the activity of the mankind in noctilucence remote sensing image, therefore model need to be examined simultaneously in primary sampling
Consider the factor of space and the density of population.The principle that the Chinese government is combined according to economic technology development level and geographical position, will
The whole nation is divided into three-piece bogies, i.e.,：East Coastal region, central region, western area.This research sets for each region
One weights λi, carry out the sampling of different proportion.
λi=θ (ai)·θ(di) (1)
Wherein it is θ (ai) accounting of the region in national land area, θ (di) it is region accounting in the density of population,
The sample size that each region is extracted is
ni=λi·N (2)
Wherein N is the image for all Landsat 8 that China's Mainland is covered by WRS-2 systems.
2) selection of second level sampling model
The pixel (Pixel) of image is defined as sampling unit in second level sampling inspection model, samples and examines in this level
Test in model, all pixels in Sample map area are counted as examining overall, according to second level sampling inspection model, select in right amount
Sample size n', is tested by the property value (noctilucence or non-noctilucence) to each inspection unit, judges the quality of the map sheet.
3) acquisition of Landsat 8/OLI images checking collection
Practically thing in the light region of noctilucence remote sensing image land mainly includes in cities and towns light, therefore Landsat 8
Main downtown areas of extracting carries out checking research.Landsat series remotely-sensed data band class informations are enriched, and rationally effectively utilize ripple
Segment information can rapidly and accurately extract construction land area.For the remote sensing images of a width Landsat 8, downtown areas extraction is carried out
When, according to being classified first according to NDVI, NDBI to Landsat images, the urban area in determining image in a wide range of.
The region that cities and towns and non-cities and towns have a common boundary accurately is determined by the way of visual interpretation simultaneously, finally in Landsat 8/OLI shadows
Accurate extraction result is reached as in.
(2) verification method
Error matrix can be evaluated effectively remote sensing image classification result.This programme is first to classification results vector
Change, by the Overlap Analysis to vector data, the checking collection obtained by two stage sampling sets up error matrix, and chooses overall essence
(oa), kappa coefficients, user's precision, producer's precision, misclassification error, leakage point error are spent to the progress system evaluation of its result.
The distribution relation of error matrix such as table 1,
Table 1：Error matrix and its Elemental redistribution
Wherein ωiRepresent that (producer) to be verified extracts the class categories of result, ωjRepresent the classification of checking (user) collection
Classification, consistent, the ρ of both classifications correspondenceijRepresent classification ωiIt is divided into ωjPixel number, ρi+And ρ+iRepresent respectively the producer and
Pixel sum of the user on without classification, m is all synthesis for verifying pixels.
Important the result can be obtained from error matrix：
Overall accuracy (overall accuracy) is (correctly classified of the elements in a main diagonal sum in error matrix
Number) divided by total number of samples：
Producer's precision (producer accuracy) and user's precision (user accuracy) can represent a certain list
The precision of individual classification.Producer's precision is that (such row are total for the correct classification number of certain classification divided by such total number of samples
With)：
pai=ρii/ρ+i, (5)
And user's definition of accuracy is such number for correctly classifying divided by is divided into such number of samples (such row is total
With)：
uai=ρii/ρi+, (6)
Except any of the above descriptive precision measure, in addition to misclassification error (commission error) and leakage point are missed
Poor (omission error)：
cei=1-uai, (7)
oei=1-pai, (8)
It can be used for the different sorting technique of comparison using various statistical analysis techniques on the basis of error matrix in addition, its
In most commonly Kappa analytical technologies：
(3) inverting of precision result
The noctilucence Remotely sensed acquisition precision in each region that the method for (2) obtains is utilized in Chinese 3 economic zones respectively
ai, the precision result A of Chinese land area noctilucence Remotely sensed acquisition is gone out in conjunction with the weight computing in each region.
By global precision A, user can intuitively draw the extraction accuracy of the noctilucence image.
Verification the verifying results
(1) noctilucence Remotely sensed acquisition result
Confirmatory experiment uses noctilucence remotely-sensed data mentioned above, and with SVM extracting methods, 2016 eve of the lunar New Year light datas are carried out
Classification.Herein by envi5.2 softwares, noctilucence data are exercised supervision classification using its integrated svm classifier algorithm, obtain night
Light area and the class classification information of non-noctilucence area two, as a result such as Fig. 2：
Violet region is extracts obtained downtown areas (pixel number 512284) in Fig. 2, and green area is not have light at night
The region of bright (pixel number 13339719), downtown areas accounts for the 3.7% of the national land gross area.
(2) foundation of two stage sampling model
According to first order sample mode, (Fig. 3) extracts 10% from the images of 536 scape Landsat 8 on the Chinese land of covering
That is 54 scape images collect as checking, and sample distribution presses Chinese three main economic region and divides distribution.Table 2 is three regions area and people
The statistical form of mouth：
The three regions area of table 2 and census returns
It can be calculated by formula (1) (2) and obtain East Coastal region, central region, the trizonal sampling fraction of western area
For 8.94：6.13：4.03, it is contemplated that the integrality of sampling map sheet, use 8 in actual sampling：6：4 sampling fraction, each region point
Not Chou Qu 24,18,12 scape images, sample distribution is as shown in Figure 4：Yellow area is the image that first order sampling is chosen in Fig. 4
Position and the region of covering.
According to second level sampling inspection model, the categorical attribute for examining pixel and the pixel need to be obtained, spectrum is utilized herein
The method of analysis and border visual identification carries out 54 scape images of selection of sampling the extraction of cities and towns area respectively, and a scape image is total to
There is 7701*7831 pixel point, all verified as checking collection.The Extraction of Image results of Landsat 8 and noctilucence data are carried
Result is taken to enter Overlap Analysis, it is the night obtained by being extracted in Landsat 8/OLI images to be stacked shadow region in effect such as Fig. 5, figure
Light region, noctilucence image zooming-out result can be judged out by the reference data classification results of contrast and the classification results of noctilucence remote sensing
Quality.
(3) the result is counted
This research is extracted region to noctilucence data using the method for error matrix and verified, according to the extraction face of two types
Product (Km2) error matrix statistical result table 3 is obtained, extract non-noctilucence region accounting in result and be far longer than noctilucence region, therefore
Non- noctilucence region is bigger to the image of result in follow-up evaluation.
Error matrix statistical form (the unit of table 3：Km2)
It can respectively be calculated according to formula (4) (9) and obtain three region overall accuracy oa and kappa coefficients, as a result such as table 4
Table 4：Overall accuracy and kappa index contrast table (units：%)
User's precision, production precision, misclassification error, leakage point error result such as table 4：
Table 4：Evaluating contrast table (unit：%)
It can show that the overall accuracy oa of this noctilucence extraction result is higher up to precision from the result, but kappa coefficients are only
Have 37.54%, the numerical value difference of two comment parameters is larger；Particularly, user's precision in noctilucence region only has in classification results
25.43%, misclassification error has reached 74.57%, illustrates that the classifying quality in noctilucence area is poor；On the other hand, the use in night region
Family precision has reached 99.88%, and misclassification error 0.12%, leakage point error 6.98% illustrates that the classification in this region is more accurate.
Thus judge that the nicety of grading in noctilucence region in this noctilucence Remotely sensed acquisition result is relatively low, night region classification knot
Really high, because area of the area in noctilucence region much smaller than night region, noctilucence region is shared in overall accuracy evaluation
Ratio is relatively low so that overall accuracy oa is higher；From kappa coefficients, the result of this noctilucence Classification in Remote Sensing Image result and checking
The similitude of collection is relatively low, and the nicety of grading for being mainly manifested in noctilucence region is relatively low, and the region at most of attribute night is divided into by mistake
Noctilucence region.Thus show that the overall accuracy of this classification results is higher, but wherein the precision in noctilucence region is less than non-noctilucence
The nicety of grading in region.
Described above is only the preferred embodiment of the present invention, it is noted that for the ordinary skill people of the art
Member, on the premise of the inventive method is not departed from, can also make some improvement and supplement, and these are improved and supplement also should be regarded as
Protection scope of the present invention.
Claims (7)
1. a kind of Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model, it is characterised in that the party
Method specifically includes following steps：
Step S1, realizes in NPP/VIIRS series datas and the built-up areas of Chinese land area is extracted, obtain not verified extraction
As a result；
Step S2, sets up two stage sampling model, completes the sampling to evaluation region, obtains believable inspection data collection；
Step S3, using error matrix verification method, the preliminary extraction to noctilucence data carries out subregion checking, and finally gives complete
The precision evaluation of body noctilucence data classification results.
2. the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model according to claim 1,
Characterized in that, the step S2 comprises the following steps：
Step S21：Divided with reference to the East Coastal of China, middle part, western three main economic region, on the Chinese land of covering
Partial image is extracted in the images of Landsat 8, the sampling collection in three regions is respectively obtained, first order sampling is completed；
Step S22：Chosen respectively in East Coastal, middle part, the corresponding each scape image in western three regions and examine pixel, obtained
The result set of second level sampling, completes second level sampling；
Step S23：The property value of checking pixel is extracted in the high-precision images of Landsat 8, judges that the pixel belongs to noctilucence
Area or non-noctilucence area.
3. the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model according to claim 1,
Characterized in that, the step S21 first order methods of samplings are：This grade of sampling inspection model is first according to, China is big from covering
Width figure is extracted in the images of Landsat 8 in land as sample, is pressed in China east coastal waters, middle part, western three main economic region
Different weights are sampled, when the map sheet for being generally n verifies collection, by examining the quality of each Sample map area whole to infer
Criticize the quality level of atlas.
4. the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model according to claim 1,
Characterized in that, also including classifying to Landsat images according to NDVI, NDBI in the step S22, a wide range of interior true
Urban area in fixing；The region that cities and towns and non-cities and towns have a common boundary accurately is determined by the way of visual interpretation simultaneously, most
Accurate extraction result is reached in Landsat images eventually, second level sampling is obtained.
5. the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model according to claim 1,
Characterized in that, comprising the following steps in step S3：
Step S31：The checking collection that the noctilucence classification of remote-sensing images result and two stage sampling model of stacked vector quantization are obtained；
Step S32：Using error matrix verification method, respectively three regions are carried out with the precision evaluation of classification results；
Step S33：With reference to the result of three regions, Extrapolation is carried out, credible reliable global classification precision is finally given
Evaluate the result.
6. the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model according to claim 5,
Characterized in that, in step S3 also by choose overall accuracy, kappa coefficients, user's precision, producer's precision, misclassification error,
Leakage point error parameter, systematically evaluates noctilucence Remotely sensed acquisition result.
7. the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of utilization two stage sampling model according to claim 1,
Characterized in that, in the step S3 will respectively in Chinese three economic zones by choose overall accuracy, kappa coefficients,
User's precision, producer's precision, misclassification error, leakage point error parameter, systematically evaluate noctilucence Remotely sensed acquisition result method and try to achieve
The noctilucence Remotely sensed acquisition precision in each region, Chinese land area noctilucence Remotely sensed acquisition is gone out in conjunction with the weight computing in each region
Precision result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710488761.XA CN107203790A (en) | 2017-06-23 | 2017-06-23 | Utilize the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of two stage sampling model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710488761.XA CN107203790A (en) | 2017-06-23 | 2017-06-23 | Utilize the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of two stage sampling model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107203790A true CN107203790A (en) | 2017-09-26 |
Family
ID=59907782
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710488761.XA Pending CN107203790A (en) | 2017-06-23 | 2017-06-23 | Utilize the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of two stage sampling model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107203790A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109670556A (en) * | 2018-12-27 | 2019-04-23 | 中国科学院遥感与数字地球研究所 | Global heat source heavy industry region recognizer based on fire point and noctilucence data |
| CN109740678A (en) * | 2019-01-07 | 2019-05-10 | 上海海洋大学 | Classification of remote-sensing images inspection method of accuracy based on multistage uneven Spatial sampling |
| CN109753916A (en) * | 2018-12-28 | 2019-05-14 | 厦门理工学院 | A kind of vegetation index spatial scaling model building method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005189099A (en) * | 2003-12-25 | 2005-07-14 | National Institute Of Information & Communication Technology | Method and device for removing noise in sar data processing |
| CN103955583A (en) * | 2014-05-12 | 2014-07-30 | 中国科学院城市环境研究所 | Method for determining threshold value of urban built-up area extracted through nighttime light data |
| CN104820774A (en) * | 2015-04-16 | 2015-08-05 | 同济大学 | Space complexity based mapsheet sampling method |
| CN105279521A (en) * | 2015-09-28 | 2016-01-27 | 上海海洋大学 | Remote-sensing image classification result precision examination method based on space sampling |
| CN106127121A (en) * | 2016-06-15 | 2016-11-16 | 四川省遥感信息测绘院 | A kind of built-up areas intellectuality extracting method based on nighttime light data |
-
2017
- 2017-06-23 CN CN201710488761.XA patent/CN107203790A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005189099A (en) * | 2003-12-25 | 2005-07-14 | National Institute Of Information & Communication Technology | Method and device for removing noise in sar data processing |
| CN103955583A (en) * | 2014-05-12 | 2014-07-30 | 中国科学院城市环境研究所 | Method for determining threshold value of urban built-up area extracted through nighttime light data |
| CN104820774A (en) * | 2015-04-16 | 2015-08-05 | 同济大学 | Space complexity based mapsheet sampling method |
| CN105279521A (en) * | 2015-09-28 | 2016-01-27 | 上海海洋大学 | Remote-sensing image classification result precision examination method based on space sampling |
| CN106127121A (en) * | 2016-06-15 | 2016-11-16 | 四川省遥感信息测绘院 | A kind of built-up areas intellectuality extracting method based on nighttime light data |
Non-Patent Citations (8)
| Title |
|---|
| KAIFANG SHI ET AL: "Evaluation of NPP-VIIRS night-time light composite data for extracting built-up urban areas", 《REMOTE SENSING LETTERS》 * |
| WENKAI LI ET AL: "A New Accuracy Assessment Method for One-Class Remote Sensing Classification", 《IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING》 * |
| 孟雯等: "基于空间抽样的区域地表覆盖遥感制图产品精度评估—以中国陕西省为例", 《地球信息科学》 * |
| 李松: "《乌鲁木齐居住空间分异及响应研究》", 28 February 2017 * |
| 柴宝惠等: "基于Landsat 数据和DMSP/OLS 夜间灯光数据的城市扩展提取: 以天津市为例", 《北京大学学报(自然科学版）》 * |
| 梁保平等: "桂林市主城区建设用地扩张时空动态特征分析", 《广西师范大学学报：自然科学版》 * |
| 王振华: "空间数据质量抽样检验与控制的理论、方法和应用", 《国家图书馆》 * |
| 陈珂: "遥感影像分类结果的空间抽样精度检验方法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109670556A (en) * | 2018-12-27 | 2019-04-23 | 中国科学院遥感与数字地球研究所 | Global heat source heavy industry region recognizer based on fire point and noctilucence data |
| CN109670556B (en) * | 2018-12-27 | 2023-07-04 | 中国科学院遥感与数字地球研究所 | Global heat source heavy industry area identification method based on fire point and noctilucent data |
| CN109753916A (en) * | 2018-12-28 | 2019-05-14 | 厦门理工学院 | A kind of vegetation index spatial scaling model building method and device |
| CN109740678A (en) * | 2019-01-07 | 2019-05-10 | 上海海洋大学 | Classification of remote-sensing images inspection method of accuracy based on multistage uneven Spatial sampling |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liao et al. | High-resolution remote sensing mapping of global land water | |
| CN102750703B (en) | Remote sensing image airport automatic detecting method based on linear cutting | |
| CN107680678A (en) | Based on multiple dimensioned convolutional neural networks Thyroid ultrasound image tubercle auto-check system | |
| CN111274918B (en) | River dryout and cutoff monitoring method and device based on multi-source remote sensing image | |
| CN106127204A (en) | A kind of multi-direction meter reading Region detection algorithms of full convolutional neural networks | |
| CN107230197B (en) | Tropical cyclone objective strength determination method based on satellite cloud image and RVM | |
| CN103632363A (en) | Object-level high-resolution remote sensing image change detection method based on multi-scale fusion | |
| CN109766936A (en) | Image change detection method based on information transmitting and attention mechanism | |
| CN111339827A (en) | SAR image change detection method based on multi-region convolutional neural network | |
| Li et al. | SPM-IS: An auto-algorithm to acquire a mature soybean phenotype based on instance segmentation | |
| CN107203790A (en) | Utilize the Chinese land noctilucence Classification in Remote Sensing Image Accuracy Assessment of two stage sampling model | |
| CN111597949A (en) | NPP-VIIRS night light data-based urban built-up area extraction method | |
| CN109635713A (en) | The shadow region glacier recognition methods of plateau mountainous region | |
| CN110852243A (en) | Improved YOLOv 3-based road intersection detection method and device | |
| Li et al. | Framework for unknown airport detection in broad areas supported by deep learning and geographic analysis | |
| CN116310853A (en) | Multi-source data-based extraction method for edge regions of medium and small cities | |
| Xu et al. | Hybrid attention-aware transformer network collaborative multiscale feature alignment for building change detection | |
| CN112347926B (en) | High-resolution image city village detection method based on building morphology distribution | |
| An et al. | Object-oriented urban dynamic monitoring—A case study of Haidian District of Beijing | |
| Li et al. | The extraction of built-up areas in Chinese mainland cities based on the local optimal threshold method using NPP-VIIRS images | |
| CN104915925A (en) | Dry valley extraction method by preparing high-precision mask file | |
| CN104809336B (en) | A kind of regional feature sampling approach for considering spatial coherence | |
| Zhang et al. | Semantic segmentation method accelerated quantitative analysis of the spatial characteristics of traditional villages | |
| CN109753896A (en) | A kind of unsupervised heterologous method for detecting change of remote sensing image based on general character autocoder | |
| Naixun et al. | Object-oriented open pit extraction based on convolutional neural network, a case study in yuzhou, China |
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: 20170926 |