CN114049562A - Method for fusing and correcting land cover data - Google Patents
Method for fusing and correcting land cover data Download PDFInfo
- Publication number
- CN114049562A CN114049562A CN202111445778.XA CN202111445778A CN114049562A CN 114049562 A CN114049562 A CN 114049562A CN 202111445778 A CN202111445778 A CN 202111445778A CN 114049562 A CN114049562 A CN 114049562A
- Authority
- CN
- China
- Prior art keywords
- spatial resolution
- source data
- remote sensing
- grid
- index
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000012549 training Methods 0.000 claims abstract description 65
- 238000005070 sampling Methods 0.000 claims abstract description 27
- 238000012937 correction Methods 0.000 claims abstract description 23
- 238000012847 principal component analysis method Methods 0.000 claims abstract description 5
- 230000003595 spectral effect Effects 0.000 claims description 40
- 239000002689 soil Substances 0.000 claims description 21
- 230000009467 reduction Effects 0.000 claims description 20
- 238000012545 processing Methods 0.000 claims description 19
- 238000004422 calculation algorithm Methods 0.000 claims description 12
- 238000007781 pre-processing Methods 0.000 claims description 10
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 10
- 229930002875 chlorophyll Natural products 0.000 claims description 7
- 235000019804 chlorophyll Nutrition 0.000 claims description 7
- ATNHDLDRLWWWCB-AENOIHSZSA-M chlorophyll a Chemical compound C1([C@@H](C(=O)OC)C(=O)C2=C3C)=C2N2C3=CC(C(CC)=C3C)=[N+]4C3=CC3=C(C=C)C(C)=C5N3[Mg-2]42[N+]2=C1[C@@H](CCC(=O)OC\C=C(/C)CCC[C@H](C)CCC[C@H](C)CCCC(C)C)[C@H](C)C2=C5 ATNHDLDRLWWWCB-AENOIHSZSA-M 0.000 claims description 7
- 238000012216 screening Methods 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000008030 elimination Effects 0.000 claims description 3
- 238000003379 elimination reaction Methods 0.000 claims description 3
- 230000000873 masking effect Effects 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 238000004364 calculation method Methods 0.000 claims 1
- 230000004927 fusion Effects 0.000 abstract description 15
- 238000000605 extraction Methods 0.000 abstract description 4
- 230000006872 improvement Effects 0.000 abstract description 4
- 230000008569 process Effects 0.000 abstract description 4
- 238000005516 engineering process Methods 0.000 abstract description 2
- 238000001914 filtration Methods 0.000 abstract description 2
- 239000000523 sample Substances 0.000 description 51
- 238000010276 construction Methods 0.000 description 4
- 238000012952 Resampling Methods 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 2
- 238000007499 fusion processing Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 235000016626 Agrimonia eupatoria Nutrition 0.000 description 1
- 241000196323 Marchantiophyta Species 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 239000012468 concentrated sample Substances 0.000 description 1
- 238000009432 framing Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007500 overflow downdraw method Methods 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000000513 principal component analysis Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 238000002310 reflectometry Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Images
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
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- 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
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (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 Processing (AREA)
Abstract
The invention discloses a method for fusing and correcting land cover data, which realizes uncertain information filtering from a coarse spatial resolution multi-source data consistent region to a fine spatial resolution multi-source data consistent region and fine spatial resolution information extraction by combining a generated super-pixel object with a principal component analysis method, and obtains a reliable training sample by a local sampling technology, corrects the fine spatial resolution multi-source data inconsistent region, and realizes the fusing and correcting of the land cover data. Compared with the prior art, the method can only realize the fusion correction of the multi-source data with the coarse spatial resolution, the technical scheme of the invention realizes the fine improvement from the coarse resolution to the fine resolution, has transportability, simultaneously solves the difficulty of simultaneously improving the resolution and the precision in the fusion correction process of the multi-source data product, and effectively solves the difficult problem of improving the operation efficiency of mass data.
Description
Technical Field
The invention relates to the field of land cover information, in particular to a method for fusing and correcting land cover data.
Background
Information is obtained through a single source of surface covering products, and the reliability is often weaker than the fusion of multiple products. The consistent area created by the various products retains information about each of the surface covering products, and although the data sources and classification methods used in the production of each product may not be the same, the consistent area is maintained under the superposition of the various data, usually with a high degree of confidence. There is then a study of the consistent and inconsistent areas created by the superposition of multiple surface covering products, and screening high confidence samples in the consistent areas. Research has shown that, whether optimization of algorithm parameters or improvement of classifiers, the contribution to classification accuracy is still smaller than that of effective, accurate and sufficient training samples. Therefore, methods of obtaining training samples from existing land cover products typically require a quality control mechanism to select reliable representative samples.
Although various classification algorithms are continuously improved, the large-scale ground cover products mostly mainly have medium-low resolution at present, uniform areas of multiple products still need to be uniformly resampled to relatively low resolution, and the resampling of the low-resolution products to fine resolution can increase the uncertainty of map information, so that the ground cover classification result obtained by the method is also low in resolution, more mixed pixels are formed, and the accuracy is reduced. Therefore, research on an extraction method for acquiring more effective information of fine resolution and fine precision of multi-source land cover products as soon as possible is urgently needed.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the super-resolution land cover data fusion correction method is provided to solve one or more technical problems that the resolution of a result generated in the existing multi-source data fusion process is too coarse, and a large number of mixed pixels exist.
In order to solve the technical problem, the invention provides a method for fusing and correcting the coverlet data, which comprises the following steps:
establishing a unified land cover classification system, re-classifying the obtained land cover products of the area to be treated, and preprocessing each land cover product to obtain a coarse spatial resolution land cover product;
carrying out region division on the land cover product with the coarse spatial resolution to obtain a coarse spatial resolution multi-source data consistent region and a coarse spatial resolution multi-source data inconsistent region;
obtaining a remote sensing image feature set of the region to be processed, establishing a plurality of grids, obtaining and performing dimension reduction processing on the remote sensing image feature set corresponding to the coarse spatial resolution multi-source data consistent region in each grid, and generating a dimension reduction remote sensing image feature set;
carrying out segmentation clustering processing on the coarse spatial resolution multi-source data consistent region to generate a super-pixel object set, judging a dimensionality reduction remote sensing image feature set corresponding to each super-pixel object, obtaining a fine spatial resolution multi-source data inconsistent region, eliminating an unreliable region in the coarse spatial resolution multi-source data consistent region to obtain a fine spatial resolution multi-source data consistent region, and taking the fine spatial resolution multi-source data consistent region as an original training sample set;
calculating and purifying an original training sample set according to a spectral index corresponding to each ground cover type in a fine spatial resolution multi-source data consistent region in each grid to obtain a fine-screened training sample set;
taking each preset grid as a center, forming a local sampling sample area by combining adjacent grids, and performing layered random sampling on the training sample set subjected to fine screening so as to enable the preset grids and the adjacent grids to respectively obtain a preset number of samples and obtain a training sample set corresponding to each grid for remote sensing interpretation;
constructing a classifier for the training sample set, interpreting the area with inconsistent multi-source data according to the classifier, and acquiring a fine spatial resolution land cover interpreted correction area;
and combining the fine spatial resolution multi-source data consistent region and the fine spatial resolution land cover interpreted correction region to generate land cover fused correction data.
Further, the obtaining of the remote sensing image feature set of the to-be-processed area specifically includes:
acquiring and masking low-quality pixels in remote sensing image data according to the remote sensing image data of the to-be-processed area in a preset year;
and simultaneously extracting and calculating the spectral index and the numerical standard deviation of the preset year according to the high-quality multi-waveband data in the remote sensing image data.
Further, the extracting and calculating the spectral index according to the high-quality multi-waveband data in the remote sensing image data specifically comprises:
extracting 6 wave bands from the 1 st wave band to the 5 th wave band and the 7 th wave band of Landsat 7 in the remote sensing image data, and 6 wave bands from the 2 nd wave band to the 7 th wave band of Landsat 8;
calculating a spectral index according to the six wave bands of the Landsat 7 and the six wave bands of the Landsat 8; the spectral indexes comprise a vegetation index, a green chlorophyll vegetation index, a modified soil adjustment vegetation index, a soil adjustment total vegetation index, a building index, a bare soil index, a burning index, a snow index and a water body index, and the formula of each spectral index is as follows:
vegetation index: NDVI ═ (Nir-Red)/(Nir + Red);
building index: NDBI ═ (Swir 1-ner)/(Swir 1+ ner);
bare soil index: BSI ═ ((Swir2+ Red) - (NIR + Blue))/((Swir2+ Red) + (NIR + Blue));
burning index: NBR (Swir1-Swir2)/(Swir1+ Swir 2);
snow index: NDSI ═ Green-Swir1)/(Green + Swir 1;
water body index: NDWI ═ Green-Nir)/(Green + Nir);
in the formula, Blue, Green, Red, Nir, Swir1 and Swir2 are respectively the waveband 1-Blue, the waveband 2-Green, the waveband 3-Red, the waveband 4-near infrared, the waveband 5-short wave infrared 1, the waveband 7-short wave infrared 2 of Landsat 7ETM + images or the waveband 2-Blue, the waveband 3-Green, the waveband 4-Red, the waveband 5-near infrared, the waveband 6-short wave infrared 1 and the waveband 7-short wave infrared 2 of Landsat 8 OLI images.
Further, the establishing a plurality of grids specifically includes:
and establishing a plurality of grids of 5 degrees multiplied by 5 degrees in the coarse spatial resolution multi-source data consistent area so as to enable the grids to uniformly cover the coarse spatial resolution multi-source data consistent area.
Further, the obtaining and dimension reduction processing is performed on the remote sensing image feature set corresponding to the coarse spatial resolution multi-source data consistent region in each grid to generate a dimension reduction remote sensing image feature set, which specifically comprises:
by adopting a principal component analysis method, performing data reconstruction on the remote sensing image feature set corresponding to the coarse spatial resolution multi-source data consistent area in each grid so as to obtain a first principal component and a second principal component of the remote sensing image feature set corresponding to each grid, and combining the first principal component and the second principal component to generate a dimension-reduction remote sensing image feature set.
Further, the method for judging the feature set of the dimension-reduced remote sensing image corresponding to each super-pixel object, obtaining the inconsistent area of the fine spatial resolution multi-source data, and eliminating the unreliable area in the consistent area of the coarse spatial resolution multi-source data specifically comprises the following steps:
dividing the coarse spatial resolution multi-source data consistent region into a plurality of super-pixel objects according to a preset super-pixel algorithm, performing feature statistical judgment on a reduced-dimension remote sensing image feature set corresponding to each super-pixel object, acquiring a fine spatial resolution multi-source data inconsistent region, and removing an outlier in the coarse spatial resolution multi-source data consistent region;
for each kind of land cover, distinguishing and removing the super-pixel object set of each kind of land cover by setting an outlier of the obvious deviation of the dimension-reduced remote sensing image characteristics in the super-pixel object set as a distinguishing condition; the formula of the discrimination condition is as follows:
where i represents the type of groundcover, j represents each superpixel object, PC1i,j,PC2i,jRespectively representing the average value of the first principal component and the second principal component of each super-pixel object dimension-reduction remote sensing image characteristic of each category,the average value of the first principal component and the second principal component of all the dimension-reduced remote sensing image characteristics representing each land cover category,and the standard deviation of the first principal component and the second principal component of the dimension-reduced remote sensing image characteristic of all the superpixel objects in each category is represented.
Further, the calculating and purifying the original training sample set according to the spectral index corresponding to each land cover type in the fine spatial resolution multi-source data consistent region in each grid to obtain a fine-screened training sample set specifically comprises:
extracting a spectral index corresponding to each cover type in a fine spatial resolution multi-source data consistent region in each grid for each cover type, obtaining a mean value and a standard deviation of the spectral index, and removing the original training sample set according to the mean value and the standard deviation;
and sequencing the original training sample set subjected to the elimination processing according to the spectral index, establishing a data histogram, and purifying data in the data histogram according to a preset percentage to obtain a fine-screened training sample set.
Further, with each preset grid as the center, combine its adjacent grid to form local sampling appearance district, right the training sample set of fine screening carries out layering random sampling, so that preset grid with adjacent grid acquire the sample of predetermineeing quantity respectively, obtain the training sample set of the remote sensing interpretation that each grid corresponds, specifically do:
establishing a local sampling sample area by an established grid-by-grid local adaptive sampling method by taking each preset grid as a center and 8 grids adjacent to the preset grid as boundaries;
in the local sampling sample area, performing layered random sampling on the training sample set finely screened in each grid, respectively obtaining half samples of the preset grid and the adjacent grids, and obtaining a training sample set of remote sensing interpretation corresponding to each grid, wherein the training sample set of each cover type in the training sample set of remote sensing interpretation is that the number of the preset grid is 1600, and the total number of the training sample sets in the adjacent grids is 1600.
Further, performing region division on the coarse spatial resolution land cover product to obtain a coarse spatial resolution multi-source data consistent region and a coarse spatial resolution multi-source data inconsistent region, specifically:
and carrying out spatial overlapping processing on the coarse spatial resolution land cover product, extracting grid units of the same land cover type in all products as coarse spatial resolution multi-source data consistent regions, and taking the rest regions as coarse spatial resolution multi-source data inconsistent regions, so that the coarse spatial resolution land cover product is divided into the coarse spatial resolution multi-source data consistent regions and the coarse spatial resolution multi-source data inconsistent regions.
Compared with the prior art, the method for fusing and correcting the land cover data has the following beneficial effects that:
1. the technical scheme of the invention establishes a fusion method for reliable information extraction and resolution improvement of a consistent region of a multi-source land cover product, successfully realizes fine resolution and high-precision fusion correction of multi-source data, and solves the problems of too coarse resolution and a large amount of mixed pixels of a land cover result generated in the existing data fusion process;
2. according to the technical scheme, based on the super-pixel object, principal component analysis and local sampling processing are carried out on the feature set of the dimensionality reduction remote sensing image, a method for extracting fine resolution and high-precision information of a multi-source land cover product is established, the difficulty that the resolution and the precision are improved while the fusion correction process of the multi-source data product is solved, and the difficulty that the efficiency of mass data operation is improved is effectively solved;
3. the technical scheme of the invention can quickly and accurately generate the multi-source data fusion product with fine spatial resolution in a large-scale area, realizes the fine promotion of the land cover fusion result from the coarse resolution to the fine resolution in the prior art, and has transportability.
Drawings
Fig. 1 is a schematic flow chart of an embodiment of a method for fusing and correcting cover-soil data provided by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Example 1
Referring to fig. 1, fig. 1 is a schematic flow chart of an embodiment of a method for fusing and correcting coverlet data according to the present invention, as shown in fig. 1, the method includes steps 101 to 108, which are as follows:
step 101: and establishing a uniform land cover classification system, reclassifying the obtained land cover products of the area to be treated, and preprocessing each land cover product to obtain a coarse spatial resolution land cover product.
In this embodiment, before reclassifying the obtained coverlet products for the area to be treated, the method includes: and carrying out framing, merging, inlaying and mask extracting on each multi-source land cover product, acquiring the altitude data of the area to be processed, and calculating the gradient of the area to be processed according to the altitude data. The land cover products include, but are not limited to, comprehensive land cover products such as CGLS-LC, CCI-LC, FROM-GLC and MCD12Q1, and special land cover products such as arable land product GFSAD-30, forest land product PALSAR, water product GSWD, and construction land product GHS-build.
In this embodiment, the types of the land cover include, but are not limited to, cultivated land, forest land, grassland, shrub, water, construction land, bare land, glacier snow, liverwort, and wetland, and the obtained land cover products of the area to be treated are divided according to the land cover classification system.
In the embodiment, each divided multi-source land cover product is preprocessed to obtain a coarse spatial resolution land cover product; wherein, the preprocessing comprises but is not limited to spatial registration preprocessing, data format preprocessing and spatial resolution unified preprocessing;
in the embodiment, the spatial registration preprocessing is to perform unified processing on the spatial range and the reference coordinate system of the ground cover product; preprocessing the data format into a format which unifies the formats of the land cover products into a grid format; and (3) uniformly preprocessing the spatial resolution, selecting the medium spatial resolution in the land cover products as a standard, dividing the land cover products according to an optimal area principle, and uniformly processing the spatial resolution of the residual land cover products. In this embodiment, the divided grid type of land cover is determined according to the type of land cover with the largest area in the grid unit.
As an example in this embodiment, the reference coordinate system of each cover product is unified into WGS84 coordinate system, and grid resampling is performed on each cover product to 300m by comparing the resolutions of CCI-LC products with the resolutions of each cover product being centered as a reference.
Step 102: and carrying out region division on the land cover product with the coarse spatial resolution to obtain a coarse spatial resolution multi-source data consistent region and a coarse spatial resolution multi-source data inconsistent region.
In this embodiment, the spatial overlay processing is performed on the coarse spatial resolution ground cover product, the grid units in all the products, which are of the same ground cover type, are extracted as coarse spatial resolution multi-source data consistent regions, and the remaining regions are extracted as coarse spatial resolution multi-source data inconsistent regions, so that the coarse spatial resolution multi-source data consistent regions and the coarse spatial resolution multi-source data inconsistent regions are obtained. Specifically, all selected land cover products are spatially superposed and subjected to consistency analysis, grid units of which all products are considered to be of the same land cover type are extracted, the grid units are divided into coarse spatial resolution multi-source data consistency areas of the multi-source land cover products, and the rest areas are divided into coarse spatial resolution multi-source data inconsistency areas.
Step 103: and obtaining a remote sensing image feature set of the region to be processed, establishing a plurality of grids, obtaining and performing dimension reduction processing on the remote sensing image feature set corresponding to the coarse spatial resolution multi-source data consistent region in each grid, and generating a dimension reduction remote sensing image feature set.
In this embodiment, according to the specified year of the land cover product, mass remote sensing image data of the to-be-processed area in the specified year is acquired, and mass remote sensing image data of the to-be-processed area in two years before and after the specified year is acquired, wherein the mass remote sensing image data are from Landsat 7ETM + images and Landsat 8 OLI images on a *** earth engine platform.
In this embodiment, a low-quality pixel in remote sensing image data is masked by an FMASK algorithm provided on a GEE platform, and the masking process is completed by a QA band of the image itself, where the low-quality pixel includes a cloud, a snow and a cloud shadow pixel. Specifically, Landsat image data are acquired based on a Google Earth engine platform, altitude data SRTM DEM is acquired in the step 101, pixels with poor quality such as cloud, snow and cloud shadow in the remote sensing image are masked according to QA wave band information of the Landsat image, and gradient is calculated by applying the altitude data. And the Landsat 7ETM + spectral features are linearly converted into Landsat 8 OLI spectral features by applying a conversion coefficient.
In this embodiment, the spectral index, the numerical value of the preset year, and the standard deviation are calculated according to the high-quality multi-band data in the remote sensing image data, and the remote sensing image data feature set is generated according to the multi-band data and the spectral data. The high-quality multi-band data in the remote sensing image data comprise 6 bands from 1 st band to 5 th band and 7 th band of Landsat 7, and 6 bands from 2 nd band to 7 th band of Landsat 8.
In this embodiment, a spectral index is calculated according to six bands of the Landsat 7 and six bands of the Landsat 8; the spectral indexes comprise a vegetation index, a green chlorophyll vegetation index, a modified soil adjustment vegetation index, a soil adjustment total vegetation index, a building index, a bare soil index, a burning index, a snow index and a water body index, and the formula of each spectral index is as follows:
vegetation index: NDVI ═ (Nir-Red)/(Nir + Red);
building index: NDBI ═ (Swir 1-ner)/(Swir 1+ ner);
bare soil index: BSI ═ ((Swir2+ Red) - (NIR + Blue))/((Swir2+ Red) + (NIR + Blue));
burning index: NBR (Swir1-Swir2)/(Swir1+ Swir 2);
snow index: NDSI ═ Green-Swir1)/(Green + Swir 1;
water body index: NDWI ═ Green-Nir)/(Green + Nir);
in the formula, Blue, Green, Red, Nir, Swir1 and Swir2 are respectively the waveband 1-Blue, the waveband 2-Green, the waveband 3-Red, the waveband 4-near infrared, the waveband 5-short wave infrared 1, the waveband 7-short wave infrared 2 of Landsat 7ETM + images or the waveband 2-Blue, the waveband 3-Green, the waveband 4-Red, the waveband 5-near infrared, the waveband 6-short wave infrared 1 and the waveband 7-short wave infrared 2 of Landsat 8 OLI images.
In the embodiment, the time-sharing segment number value of the remote sensing image characteristic of the region to be processed in 1 year comprises the median composition of remote sensing image data in each segment of 4 segments in one year and corresponding spectral indexes, wherein the 4 segments are respectively 1-3 months, 4-6 months, 7-9 months and 10-12 months; and finally, acquiring median results of 6 wave bands and 9 spectral indexes in 4 periods of a specified year, acquiring standard deviation results of a year before and after the specified year, and collecting the results to generate a remote sensing image feature set with 75 layers of coarse spatial resolution multi-source data consistent regions and 30m spatial resolution.
In this embodiment, in order to increase the distinguishing capability of the remote sensing image features, the interpretation precision is improved by setting the difference of the annual fluctuation of different land cover types, the numerical standard deviation of the corresponding wave band or index between the specified year and the years before and after the specified year is calculated, and the remote sensing image feature set corrected by land cover data fusion is established.
In this embodiment, a plurality of grids of 5 ° × 5 ° are established in the coarse spatial resolution multi-source data consistency region, so that the plurality of grids uniformly cover the coarse spatial resolution multi-source data consistency region. The grid is established in the area with the consistent multi-source data of the coarse spatial resolution, so that the problem that the image interpretation precision is restricted due to overlarge reflectivity difference of the same ground object caused by overlarge latitude and longitude span of the area is solved, and the partition correction is facilitated.
In this embodiment, dimension reduction processing is performed on the remote sensing image feature sets of 75 layers corresponding to each grid, and data reconstruction is performed on the remote sensing image feature sets of 75 layers corresponding to the coarse spatial resolution multi-source data consistent area in each grid mainly by using a principal component analysis method, so that a first principal component and a second principal component of the remote sensing image feature set of 30m spatial resolution corresponding to each grid are obtained, and the dimension reduction remote sensing image feature set is generated by combining the first principal component and the second principal component.
In this embodiment, it is difficult to detect outliers of various indexes because there is no single image feature to completely distinguish various features. Therefore, dimension reduction can be performed by a principal component method, and more spatial information can be represented by less data dimension information, so that each place can be distinguished for subsequent processing analysis.
Step 104: and carrying out segmentation clustering processing on the coarse spatial resolution multi-source data consistent region to generate a super-pixel object set, judging a dimensionality reduction remote sensing image feature set corresponding to each super-pixel object, obtaining a multi-source data inconsistent region, eliminating unreliable regions in the coarse spatial resolution multi-source data consistent region to obtain a fine spatial resolution multi-source data consistent region, and taking the fine spatial resolution multi-source data consistent region as an original training sample set.
In the embodiment, the grids are clustered through a superpixel algorithm based on the similarity between the dimensionality reduction remote sensing image characteristics of the grids in the coarse spatial resolution multi-source data consistent region, the coarse spatial resolution multi-source data consistent region is divided into a plurality of superpixel objects, and a superpixel object set is generated, wherein the size parameter of a superpixel algorithm seed interval is set to be 3, and the compactness parameter is set to be 0; and taking the median synthesis result of the 6 original wave bands of the Landsat image in four time periods in a preset year and the acquired first principal component and second principal component of the image features after dimensionality reduction as input wave bands. After segmentation by using a superpixel algorithm, calculating an average value of the first principal component and the second principal component in each superpixel object as the characteristic of the superpixel object.
In this embodiment, the feature set of the dimension-reduced remote sensing image corresponding to each superpixel object is discriminated, and an unreliable region in the coarse spatial resolution multi-source data consistent region is removed, specifically, the feature set of the dimension-reduced remote sensing image corresponding to each superpixel object is subjected to image feature statistical discrimination, a fine spatial resolution multi-source data inconsistent region is obtained, and an outlier in the coarse spatial resolution multi-source data consistent region is removed, where the outlier is the unreliable region; for each land cover variety, judging and removing the super-pixel object set of each land cover variety by setting an outlier of the obvious deviation of the dimension-reduced remote sensing image characteristics in the super-pixel object set as a judging condition; the formula of the discrimination condition is as follows:
where i represents the type of groundcover, j represents each superpixel object, PC1i,j,PC2i,jRespectively representing the average value of the first principal component and the second principal component of each super-pixel object dimension-reduction remote sensing image characteristic of each category,the average value of the first principal component and the second principal component of all the dimension-reduced remote sensing image characteristics representing each land cover category,and the standard deviation of the first principal component and the second principal component of the dimension-reduced remote sensing image characteristic of all the superpixel objects in each category is represented.
As an example in this embodiment, based on an image feature set with a spatial resolution of 30m, a result obtained by performing superpixel segmentation on a coarse spatial resolution multi-source data consistent region and performing discrimination and rejection satisfying an outlier is set as a fine spatial resolution multi-source data consistent region with a spatial resolution of 30m, which is used as an original training sample set for land cover interpretation, and an outlier range and a coarse spatial resolution multi-source data consistent region which are statistically discriminated are uniformly classified as multi-source data inconsistent regions.
Different from the prior art, for the selection of the spatial resolution of the data, the coarsest resolution of all products is often selected as the target of resampling to control the product precision, so that the spatial resolution of the preprocessed multi-source land cover product is 300m in step 101, the data set is subjected to fusion correction, and finally the result with coarser resolution is still achieved. In this embodiment, the area with the relatively coarse resolution consistency obtained initially is segmented, screened, and rejected by combining the Landsat image with the spatial resolution of 30m, so as to obtain the multi-source data effective information with the spatial resolution of 30 m.
And 105, calculating and purifying the original training sample set according to the spectral index corresponding to each ground cover type in the fine spatial resolution multi-source data consistent region in each grid to obtain a fine-screened training sample set.
In this embodiment, a preset training sample screening method is set for types of land cover, for each type of land cover, a spectral index corresponding to each type of land cover in a fine spatial resolution multisource data consistent region in each grid is calculated, specifically, the type of land cover corresponding to each grid is obtained, the calculated corresponding spectral index of each type of land cover is as follows, cultivated land is a vegetation index NDVI, forest land, grassland and shrubs are green chlorophyll vegetation indexes GCVI, water and wetland are water indexes NDWI, construction land is a construction index NDBI, bare land is a bare soil index BSI, and glacier snow and fur are snow indexes NDSI.
In the embodiment, by setting the corresponding spectral indexes aiming at different types of the land cover, the problem that the identification capability of directly adopting a remote sensing image spectral band or a single spectral index to different types of the land cover is insufficient can be effectively solved, and the abnormal value in the original training sample set of each type of the land cover can be effectively judged.
In this embodiment, after extracting the spectral index corresponding to each grid, and establishing statistical characteristic judgment, the original training sample set established in the fine spatial resolution multi-source data consistent region is purified to obtain a fine-screened training sample, specifically, based on the fine spatial resolution multi-source data consistent region with the spatial resolution of 30m extracted preliminarily, for each land cover type, the mean value and the standard deviation of the spectral index corresponding to the land type in each grid are obtained, grid units in the numerical range in which the spectral index exceeds the positive and negative standard deviations of the grid mean value are removed, and based on the mean value and the standard deviations, the original training sample set is removed.
In this embodiment, the original training sample sets after being subjected to the elimination processing are sorted according to each land cover type and the corresponding spectral index, a data histogram is established, and sorting is performed in a sequence from low to high. And purifying the data in the data histogram according to a preset percentage to obtain a fine-screened training sample set, wherein as an example in the embodiment, 2.5% of samples are removed from the head end and the tail end of the histogram respectively, and 95% of samples in the center are reserved. By eliminating the height difference between the head end and the tail end, the purity of the concentrated samples of the training samples is improved, and the training results of the subsequent samples are more stable. Step 106: and taking each preset grid as a center, combining adjacent grids to form a local sampling sample area, and performing layered random sampling on the finely screened training sample set so that the preset grids and the adjacent grids respectively obtain samples with preset quantity to obtain the training sample set corresponding to each grid for remote sensing interpretation.
In this embodiment, a local sampling sample area is established by a preset grid-by-grid local adaptive sampling method, with each grid as a center and 8 grids adjacent to the center grid as boundaries; in the local sampling sample area, performing layered random sampling on the training sample set finely screened in each grid according to the types of the ground cover, and respectively obtaining half of samples from a preset grid and an adjacent grid to obtain a training sample set remotely interpreted by each grid, wherein as an example in the embodiment, the number of the samples of each ground cover type in the local sampling sample area is 1600, and the number of the samples of 8 grids adjacent to the central grid is respectively 200, so that the data of the training sample set of each ground cover type of each grid is 3200; if the quantity of training samples is insufficient due to the fact that a certain type of consistency region is lacked or the number of consistency regions is small in the corresponding grid, the lacked samples are supplemented from the nearest grids around, and a training sample set used for remote sensing interpretation of each grid is obtained.
In this embodiment, the local sampling sample area is established to obtain a training sample interpreted by remote sensing of the preset grid, and the sample integrates information of all peripheral grids, thereby controlling differences between different grids to the maximum extent.
Step 107: and constructing a classifier for the training sample set, interpreting the area with inconsistent multi-source data according to the classifier, and acquiring a fine spatial resolution land cover interpreted correction area.
In this embodiment, according to a 75-layer remote sensing image feature set with a spatial resolution of 30m and a spectral index and terrain auxiliary data, a classifier is constructed for a remote sensing image interpretation training sample set corresponding to each grid through a machine learning algorithm, an inconsistent area of the multi-source data is interpreted, and a fine spatial resolution multi-source data classification result with a corrected spatial resolution of 30m is obtained. Wherein the terrain assistance data comprises elevation and grade data; the machine learning algorithm is a random forest algorithm.
In the embodiment, a characteristic set for remote sensing image land cover interpretation is generated based on a Google earth engine platform, a random forest algorithm is applied, samples purified in a fine spatial resolution multi-source data consistent region are used as a training sample set of the algorithm, a parameter tree number is set to be 300, the selected characteristic quantity of each tree is the square root of the total wave band and the remote sensing index, the training quantity input into each tree is 63% of the total sample quantity, and correction is performed on the multi-source data inconsistent region.
Step 108: and combining the fine spatial resolution multi-source data consistent region and the fine spatial resolution land cover interpreted correction region to generate land cover fused correction data.
In this embodiment, the land cover correction result of the multisource data inconsistency area and the land cover result of the fine spatial resolution multisource data consistency area are spatially embedded by combining the multisource data inconsistency area obtained in step 106 and the fine spatial resolution multisource data consistency area obtained in step 104, so as to generate fine resolution and high-precision land cover fusion correction data. And the spatial resolution corresponding to the fine spatial resolution multi-source data consistent area and the multi-source data inconsistent area is 30 m.
In summary, the invention provides a super-resolution land cover data fusion correction method, which realizes uncertain information filtering from a coarse spatial resolution multi-source data consistent region to a fine spatial resolution consistent region and fine spatial resolution multi-source data information extraction by combining a generated super-pixel object with a principal component analysis method, and corrects the fine spatial multi-source data inconsistent region by a method of obtaining a reliable training sample by a local sampling technology, thereby realizing the fusion correction of the land cover data. Compared with the prior art, the method and the device can only realize the fusion correction of the multi-source data with the coarse spatial resolution, the technical scheme of the invention realizes the fine improvement from the coarse resolution to the fine resolution, has transportability, simultaneously solves the difficult points of simultaneously improving the resolution and the precision in the fusion correction process of the multi-source data product, and effectively solves the problem of improving the efficiency of mass data operation.
The above description is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and substitutions can be made without departing from the technical principle of the present invention, and these modifications and substitutions should also be regarded as the protection scope of the present invention.
Claims (9)
1. A method for fusing and correcting cover data of the land is characterized by comprising the following steps:
establishing a unified land cover classification system, re-classifying the obtained land cover products of the area to be treated, and preprocessing each land cover product to obtain a coarse spatial resolution land cover product;
carrying out region division on the land cover product with the coarse spatial resolution to obtain a coarse spatial resolution multi-source data consistent region and a coarse spatial resolution multi-source data inconsistent region;
obtaining a remote sensing image feature set of the region to be processed, establishing a plurality of grids, obtaining and performing dimension reduction processing on the remote sensing image feature set corresponding to the coarse spatial resolution multi-source data consistent region in each grid, and generating a dimension reduction remote sensing image feature set;
carrying out segmentation clustering processing on the coarse spatial resolution multi-source data consistent region to generate a super-pixel object set, judging a dimensionality reduction remote sensing image feature set corresponding to each super-pixel object, obtaining a fine spatial resolution multi-source data inconsistent region, eliminating an unreliable region in the coarse spatial resolution multi-source data consistent region to obtain a fine spatial resolution multi-source data consistent region, and taking the fine spatial resolution multi-source data consistent region as an original training sample set;
calculating and purifying an original training sample set according to a spectral index corresponding to each ground cover type in a fine spatial resolution multi-source data consistent region in each grid to obtain a fine-screened training sample set;
taking each preset grid as a center, forming a local sampling sample area by combining adjacent grids, and performing layered random sampling on the training sample set subjected to fine screening so as to enable the preset grids and the adjacent grids to respectively obtain a preset number of samples and obtain a training sample set corresponding to each grid for remote sensing interpretation;
constructing a classifier for the training sample set, interpreting the area with inconsistent multi-source data according to the classifier, and acquiring a fine spatial resolution land cover interpreted correction area;
and combining the fine spatial resolution multi-source data consistent region and the fine spatial resolution land cover interpreted correction region to generate land cover fused correction data.
2. The method for fusing and correcting the land cover data according to claim 1, wherein the obtaining of the remote sensing image feature set of the area to be processed specifically comprises:
acquiring and masking low-quality pixels in remote sensing image data according to the remote sensing image data of the to-be-processed area in a preset year;
and simultaneously extracting and calculating the spectral index and the numerical standard deviation of the preset year according to the high-quality multi-waveband data in the remote sensing image data.
3. The method for fusing and correcting the land cover data according to claim 2, wherein the spectral index is extracted and calculated according to high-quality multi-waveband data in the remote sensing image data, and specifically comprises the following steps:
extracting 6 wave bands from the 1 st wave band to the 5 th wave band and the 7 th wave band of Landsat 7 in the remote sensing image data, and 6 wave bands from the 2 nd wave band to the 7 th wave band of Landsat 8;
calculating a spectral index according to the six wave bands of the Landsat 7 and the six wave bands of the Landsat 8; the spectral indexes comprise a vegetation index, a green chlorophyll vegetation index, a modified soil adjustment vegetation index, a soil adjustment total vegetation index, a building index, a bare soil index, a burning index, a snow index and a water body index, and the formula of each spectral index is as follows:
vegetation index: NDVI ═ (Nir-Red)/(Nir + Red);
building index: NDBI ═ (Swir 1-ner)/(Swir 1+ ner);
bare soil index: BSI ═ ((Swir2+ Red) - (NIR + Blue))/((Swir2+ Red) + (NIR + Blue));
burning index: NBR (Swir1-Swir2)/(Swir1+ Swir 2);
snow index: NDSI ═ Green-Swir1)/(Green + Swir 1;
water body index: NDWI ═ Green-Nir)/(Green + Nir);
in the formula, Blue, Green, Red, Nir, Swir1 and Swir2 are respectively the waveband 1-Blue, the waveband 2-Green, the waveband 3-Red, the waveband 4-near infrared, the waveband 5-short wave infrared 1, the waveband 7-short wave infrared 2 of Landsat 7ETM + images or the waveband 2-Blue, the waveband 3-Green, the waveband 4-Red, the waveband 5-near infrared, the waveband 6-short wave infrared 1 and the waveband 7-short wave infrared 2 of Landsat 8 OLI images.
4. The method for fusing and correcting cover data of land according to claim 1, wherein the establishing of a plurality of grids specifically comprises:
and establishing a plurality of grids of 5 degrees multiplied by 5 degrees in the coarse spatial resolution multi-source data consistent area so as to enable the grids to uniformly cover the coarse spatial resolution multi-source data consistent area.
5. The method for fusing and correcting the land cover data according to claim 4, wherein the obtaining and performing the dimension reduction processing on the remote sensing image feature set corresponding to the coarse spatial resolution multi-source data consistent area in each grid to generate the dimension reduction remote sensing image feature set specifically comprises:
by adopting a principal component analysis method, performing data reconstruction on the remote sensing image feature set corresponding to the coarse spatial resolution multi-source data consistent area in each grid so as to obtain a first principal component and a second principal component of the remote sensing image feature set corresponding to each grid, and combining the first principal component and the second principal component to generate a dimension-reduction remote sensing image feature set.
6. The method for fusing and correcting the land cover data according to claim 1, wherein the step of distinguishing the feature set of the dimensionality reduction remote sensing image corresponding to each super-pixel object, the step of obtaining the area with the inconsistent fine spatial resolution multi-source data, and the step of eliminating the unreliable area in the area with the consistent coarse spatial resolution multi-source data comprises the following steps:
dividing the coarse spatial resolution multi-source data consistent region into a plurality of super-pixel objects according to a preset super-pixel algorithm, performing feature statistical judgment on a reduced-dimension remote sensing image feature set corresponding to each super-pixel object, acquiring a fine spatial resolution multi-source data inconsistent region, and removing an outlier in the coarse spatial resolution multi-source data consistent region;
for each kind of land cover, distinguishing and removing the super-pixel object set of each kind of land cover by setting an outlier of the obvious deviation of the dimension-reduced remote sensing image characteristics in the super-pixel object set as a distinguishing condition; the formula of the discrimination condition is as follows:
where i represents the type of groundcover, j represents each superpixel object, PC1i,j,PC2i,jRespectively representing the average value of the first principal component and the second principal component of each super-pixel object dimension-reduction remote sensing image characteristic of each category,the average value of the first principal component and the second principal component of all the dimension-reduced remote sensing image characteristics representing each land cover category,and the standard deviation of the first principal component and the second principal component of the dimension-reduced remote sensing image characteristic of all the superpixel objects in each category is represented.
7. The method for fusing and correcting cover data according to claim 1, wherein the original training sample set is purified according to the spectral index corresponding to each cover type in the fine spatial resolution multi-source data consistent region in each grid by calculation to obtain a fine-screened training sample set, specifically:
extracting a spectral index corresponding to each cover type in a fine spatial resolution multi-source data consistent region in each grid for each cover type, obtaining a mean value and a standard deviation of the spectral index, and removing the original training sample set according to the mean value and the standard deviation;
and sequencing the original training sample set subjected to the elimination processing according to the spectral index, establishing a data histogram, and purifying data in the data histogram according to a preset percentage to obtain a fine-screened training sample set.
8. The method for fusing and correcting cover-soil data according to claim 1, wherein each preset grid is taken as a center, a local sampling sample area is formed by combining adjacent grids, and the finely screened training sample set is subjected to layered random sampling, so that the preset grid and the adjacent grids respectively obtain a preset number of samples, and a remotely sensed and interpreted training sample set corresponding to each grid is obtained, specifically:
establishing a local sampling sample area by an established grid-by-grid local adaptive sampling method by taking each preset grid as a center and 8 grids adjacent to the preset grid as boundaries;
in the local sampling sample area, performing layered random sampling on the training sample set finely screened in each grid, respectively obtaining half samples of the preset grid and the adjacent grids, and obtaining a training sample set of remote sensing interpretation corresponding to each grid, wherein the training sample set of each cover type in the training sample set of remote sensing interpretation is that the number of the preset grid is 1600, and the total number of the training sample sets in the adjacent grids is 1600.
9. The method for fusing and correcting the land cover data according to claim 1, wherein the land cover product with the coarse spatial resolution is divided into regions to obtain a coarse spatial resolution multi-source data consistent region and a coarse spatial resolution multi-source data inconsistent region, and specifically comprises the following steps:
and carrying out spatial overlapping processing on the coarse spatial resolution land cover product, extracting grid units of the same land cover type in all products as coarse spatial resolution multi-source data consistent regions, and taking the rest regions as coarse spatial resolution multi-source data inconsistent regions, so that the coarse spatial resolution land cover product is divided into the coarse spatial resolution multi-source data consistent regions and the coarse spatial resolution multi-source data inconsistent regions.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111445778.XA CN114049562B (en) | 2021-11-30 | 2021-11-30 | Method for fusing and correcting land cover data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111445778.XA CN114049562B (en) | 2021-11-30 | 2021-11-30 | Method for fusing and correcting land cover data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114049562A true CN114049562A (en) | 2022-02-15 |
CN114049562B CN114049562B (en) | 2022-06-21 |
Family
ID=80212014
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111445778.XA Expired - Fee Related CN114049562B (en) | 2021-11-30 | 2021-11-30 | Method for fusing and correcting land cover data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114049562B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114842356A (en) * | 2022-07-01 | 2022-08-02 | 江西师范大学 | High-resolution earth surface type sample automatic generation method, system and equipment |
CN115527123A (en) * | 2022-10-21 | 2022-12-27 | 河北省科学院地理科学研究所 | Land cover remote sensing monitoring method based on multi-source feature fusion |
CN116975784A (en) * | 2023-09-19 | 2023-10-31 | 四川省水利科学研究院 | High-space-time resolution MPDI data set construction method, system and storage medium |
CN118154994A (en) * | 2024-05-10 | 2024-06-07 | 中国科学院、水利部成都山地灾害与环境研究所 | Remote sensing image land classification precision evaluation sampling method and device and electronic equipment |
Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103035006A (en) * | 2012-12-14 | 2013-04-10 | 南京大学 | High-resolution aerial image partition method based on LEGION and under assisting of LiDAR |
CN106991676A (en) * | 2017-04-06 | 2017-07-28 | 哈尔滨理工大学 | A kind of super-pixel fusion method of local correlation |
CN107170000A (en) * | 2017-04-18 | 2017-09-15 | 武汉市工程科学技术研究院 | The stereopsis dense Stereo Matching method optimized based on global block |
CN109063663A (en) * | 2018-08-10 | 2018-12-21 | 武汉大学 | A kind of spissatus detection of timing remote sensing image and minimizing technology by slightly to essence |
CN109856056A (en) * | 2018-12-26 | 2019-06-07 | 北京林业大学 | A kind of Application of Remote Sensing Technique To Sandy Desertification method for quickly identifying |
CN110796667A (en) * | 2019-10-22 | 2020-02-14 | 辽宁工程技术大学 | Color image segmentation method based on improved wavelet clustering |
CN112800138A (en) * | 2021-02-04 | 2021-05-14 | 广东云曌医疗科技有限公司 | Big data classification method and system |
CN113160237A (en) * | 2021-03-02 | 2021-07-23 | 中国科学院地理科学与资源研究所 | Method for drawing earth cover |
CN113344013A (en) * | 2021-07-15 | 2021-09-03 | 西安电子科技大学 | Polarimetric SAR image feature mining method based on graph low-rank sparse decomposition |
CN113569905A (en) * | 2021-06-10 | 2021-10-29 | 西安电子科技大学 | Heterogeneous image fusion method based on multi-scale feature extraction and total variation |
-
2021
- 2021-11-30 CN CN202111445778.XA patent/CN114049562B/en not_active Expired - Fee Related
Patent Citations (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103035006A (en) * | 2012-12-14 | 2013-04-10 | 南京大学 | High-resolution aerial image partition method based on LEGION and under assisting of LiDAR |
CN106991676A (en) * | 2017-04-06 | 2017-07-28 | 哈尔滨理工大学 | A kind of super-pixel fusion method of local correlation |
CN107170000A (en) * | 2017-04-18 | 2017-09-15 | 武汉市工程科学技术研究院 | The stereopsis dense Stereo Matching method optimized based on global block |
CN109063663A (en) * | 2018-08-10 | 2018-12-21 | 武汉大学 | A kind of spissatus detection of timing remote sensing image and minimizing technology by slightly to essence |
CN109856056A (en) * | 2018-12-26 | 2019-06-07 | 北京林业大学 | A kind of Application of Remote Sensing Technique To Sandy Desertification method for quickly identifying |
CN110796667A (en) * | 2019-10-22 | 2020-02-14 | 辽宁工程技术大学 | Color image segmentation method based on improved wavelet clustering |
CN112800138A (en) * | 2021-02-04 | 2021-05-14 | 广东云曌医疗科技有限公司 | Big data classification method and system |
CN113160237A (en) * | 2021-03-02 | 2021-07-23 | 中国科学院地理科学与资源研究所 | Method for drawing earth cover |
CN113569905A (en) * | 2021-06-10 | 2021-10-29 | 西安电子科技大学 | Heterogeneous image fusion method based on multi-scale feature extraction and total variation |
CN113344013A (en) * | 2021-07-15 | 2021-09-03 | 西安电子科技大学 | Polarimetric SAR image feature mining method based on graph low-rank sparse decomposition |
Non-Patent Citations (5)
Title |
---|
ACHANTA R等: "Susstrunk S. Superpixels and polygons using simple non-iterative clustering", 《PROCEEDINGS OF THE IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION》 * |
CALDERÓN-LOOR M等: "High-resolution wall-to-wall land-cover mapping and land change assessment for Australia from 1985 to 2015", 《REMOTE SENSING OF ENVIRONMENT》 * |
JIN Q等: "A Fusion Method for Multisource Land Cover Products Based on Superpixels and Statistical Extraction for Enhancing Resolution and Improving Accuracy", 《REMOTE SENSING》 * |
LI KEWEI等: "Cropland data fusion and correction using spatial analysis techniques and the Google Earth Engine", 《GISCIENCE & REMOTE SENSING》 * |
宋宏利: "多源土地覆被遥感信息融合及数据重构研究", 《中国优秀博硕士学位论文全文数据库(博士)基础科学辑》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114842356A (en) * | 2022-07-01 | 2022-08-02 | 江西师范大学 | High-resolution earth surface type sample automatic generation method, system and equipment |
CN114842356B (en) * | 2022-07-01 | 2022-10-04 | 江西师范大学 | High-resolution earth surface type sample automatic generation method, system and equipment |
CN115527123A (en) * | 2022-10-21 | 2022-12-27 | 河北省科学院地理科学研究所 | Land cover remote sensing monitoring method based on multi-source feature fusion |
CN115527123B (en) * | 2022-10-21 | 2023-05-05 | 河北省科学院地理科学研究所 | Land cover remote sensing monitoring method based on multisource feature fusion |
CN116975784A (en) * | 2023-09-19 | 2023-10-31 | 四川省水利科学研究院 | High-space-time resolution MPDI data set construction method, system and storage medium |
CN116975784B (en) * | 2023-09-19 | 2023-12-29 | 四川省水利科学研究院 | High-space-time resolution MPDI data set construction method, system and storage medium |
CN118154994A (en) * | 2024-05-10 | 2024-06-07 | 中国科学院、水利部成都山地灾害与环境研究所 | Remote sensing image land classification precision evaluation sampling method and device and electronic equipment |
Also Published As
Publication number | Publication date |
---|---|
CN114049562B (en) | 2022-06-21 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114049562B (en) | Method for fusing and correcting land cover data | |
CN111709379B (en) | Remote sensing image-based hilly area citrus planting land plot monitoring method and system | |
CN113033670B (en) | Rice planting area extraction method based on Sentinel-2A/B data | |
Faisal et al. | IHDS: Intelligent harvesting decision system for date fruit based on maturity stage using deep learning and computer vision | |
Yang et al. | Evaluating high resolution SPOT 5 satellite imagery for crop identification | |
CN110991335A (en) | Visible light unmanned aerial vehicle remote sensing image forest tree species classification method based on multi-feature optimization | |
CN112102312B (en) | Moso bamboo forest remote sensing identification method based on satellite image and phenological difference containing red edge wave band | |
CN111345214A (en) | Xinjiang cotton region identification method and system based on satellite image data | |
Baofeng et al. | Digital surface model applied to unmanned aerial vehicle based photogrammetry to assess potential biotic or abiotic effects on grapevine canopies | |
CN113160237B (en) | Method for drawing earth cover | |
Homer et al. | Development of a circa 2000 landcover database for the United States | |
CN113869370A (en) | Tree species refined classification method based on deep learning algorithm and time sequence sentinel image | |
CN113205014A (en) | Time sequence data farmland extraction method based on image sharpening | |
CN114778483A (en) | Method for correcting terrain shadow of remote sensing image near-infrared wave band for monitoring mountainous region | |
CN115512237A (en) | Anhui winter wheat planting area extraction method based on multi-temporal satellite remote sensing data | |
CN111832480B (en) | Remote sensing identification method for rape planting area based on spectral characteristics | |
Fisette et al. | Methodology for a Canadian agricultural land cover classification | |
CN109446965B (en) | Automatic tobacco field identification method based on unmanned aerial vehicle data | |
CN113989668A (en) | Remote sensing crop automatic classification method based on time series characteristics | |
CN115147730A (en) | Remote sensing classification method combining full convolution neural network and ensemble learning | |
CN114694048A (en) | Sparse shrub species identification method and system based on unmanned aerial vehicle remote sensing technology | |
CN109886142B (en) | Crop interpretation method based on SAR technology | |
Ma et al. | Banana individual segmentation and phenotypic parameter measurements using deep learning and terrestrial LiDAR | |
Whiteside et al. | Vegetation map for Magela Creek floodplain using WorldView-2 multispectral image data | |
CN113705454A (en) | Method for extracting forest land containing infrared spectrum remote sensing image |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20220621 |