CN104361338A - Peat bog information extracting method based on ENVISAT ASAR, Landsat TM and DEM data - Google Patents

Abstract

A peat swamp information extraction method based on ENVISAT ASAR, Landsat TM and DEM data, the invention relates to a peat swamp information extraction method. The invention solves the problem that traditional methods are difficult to distinguish peat bogs from other types of swamps. The method uses 1 to preprocess Landsat TM data; 2 to preprocess ENVISAT ASAR data; 3 to resample ENVISAT ASAR data; 4 to obtain ENVISAT ASAR images; 5 to obtain Slope data; 6. Extract the backscatter coefficient; 7. Determine the best polarization mode band of ENVISAT ASAR image; 8. Obtain the segmentation unit; 9. Extract characteristic parameters; 10. Determine the best classification band; Vector file; 13 steps to make a peat bog map. The invention is applied to the field of peat swamp information extraction.

Description

A Peat Swamp Information Extraction Method Based on ENVISAT ASAR, Landsat TM and DEM Data

technical field

The invention relates to an information extraction method, in particular to a peat swamp information extraction method. the

Background technique

Peat swamp is one of the main types of wetlands, which plays an important role in maintaining regional ecological balance and sustainable development. In addition, due to the huge carbon storage in peat swamps, accounting for about 1/3 of the global terrestrial carbon pool, equivalent to 75% of the carbon content in the atmosphere, peat swamps play a pivotal role in global climate change and ecosystem balance status. In recent years, experts and scholars at home and abroad have used different remote sensing image data to study the spatial information extraction of wetland types such as herbaceous wetlands, forest wetlands, and coastal mangrove wetlands. Relatively small. the

Optical remote sensing data has the advantages and characteristics of rich spectral information, high cost performance, easy acquisition and simple data processing. However, due to the influence of vegetation on the ground surface, wetlands and non-wetlands can be distinguished using traditional medium and low resolution optical remote sensing images, but it is difficult to distinguish different types of swamps, and it is even more difficult to distinguish different types of peat swamps. Radar remote sensing data has a longer wavelength than optical remote sensing data, and its characteristics of penetrating clouds and fog make it free from time and meteorological constraints when monitoring areas where swamps are widely developed. Moreover, the backscattering of radar images is more sensitive to the dielectric properties (soil moisture, vegetation water content) and geometric characteristics (surface roughness) of the imaging surface. item information. Low-frequency radar bands (P-band and L-band) are more suitable for monitoring wetlands dominated by woodlands, while high-frequency radar bands (C-band) are suitable for studying herbaceous swamps and peat swamps. the

The object-oriented interpretation method not only takes into account the spectral information of the ground objects, but also takes into account the geometric and structural features of the ground objects. The smallest unit of image interpretation is to have the same characteristics (such as spectrum, texture and Features such as spatial combination relationship) Homogeneous and uniform objects. Compared with the traditional remote sensing interpretation method that interprets the characteristics of a single pixel of the image, this method breaks through the limitations of the traditional remote sensing classification method that uses pixels as the basic classification and processing unit to contain more semantic information The object composed of multiple adjacent pixels of the remote sensing image is a processing unit, which can realize higher-level remote sensing image classification and target feature extraction. This method is a remote sensing information extraction method based on a cognitive model, which is closer to the human cognitive process, and has become one of the main research directions in the field of remote sensing information extraction. The object-oriented interpretation method emerges at the historic moment in response to the shortcomings of the previous pixel-oriented interpretation methods. The shortcomings of the object-oriented interpretation method are not mentioned in the current research. the

Landsat is a series of land resource satellites launched by NASA since 1972. The sensor TM carried by Landsat5 contains 7 wavebands (0.45～0.53μm, 0.52～0.60μm, 0.63～0.69μm, 0.76～0.90μm, 1.55～1.75μm, 10.40～12.50μm, 2.08～2.35μm), orbital height 705km , the spatial resolution is 30m, and the revisit period is 16 days. ENVISAT is a giant environmental monitoring satellite launched by the European Space Agency in March 2002. ASAR (Advanced Synthetic Aperture Radar) is an advanced synthetic aperture radar carried by ENVISAT. It has multi-mode, multi-polarization, large width, and multiple incident angles. and other characteristics. The digital elevation model (Digital Elevation Model, referred to as DEM), which is a solid ground model that expresses the ground elevation in the form of a set of ordered numerical arrays. the

The high-frequency radar band (C-band) is suitable for the study of herbaceous swamps and peat swamps. It is the conclusion that the existing research has drawn. Based on this conclusion, this study selected the C-band radar images when extracting peat swamps, that is, ENVI SAT used in the study. Because the research only involves radar images in the C-band, and does not involve radar images in other bands, it does not involve solving the problem that the high-frequency radar band (C-band) is suitable for studying herbaceous swamps and peat swamps. the

Contents of the invention

The purpose of the present invention is to solve the problem that it is difficult to distinguish peat swamps from other swamp types by using traditional medium and low resolution optical remote sensing images, and propose a peat swamp information extraction method based on ENVISAT ASAR, Landsat TM and DEM data . the

Above-mentioned purpose of the invention is achieved through the following technical solutions:

Step 1: Preprocessing the Landsat TM data;

Step 2: Preprocessing the ENVISAT ASAR data;

Step 3: Resample the ENVISAT ASAR data preprocessed in Step 2, and the resampled ENVISAT ASAR data has the same grid size as the Landsat TM data processed in Step 1;

Step 4: Use the function of adding control points provided by the Georeferencing module of ArcGIS software to select control points on the preprocessed Landsat TM data, and register the resampled ENVISAT ASAR data according to the control point space to obtain ENVISAT ASAR images;

Step 5: Extract the slope from the DEM data to obtain the slope data;

Step 6: Based on the land cover type survey sample points, extract the backscatter coefficients of radar images under different polarization modes for different land cover types from the ENVISAT ASAR image completed in step 4;

Step 7: Analyze the difference in radar backscatter coefficients between peat swamps and other different land cover types under different polarization modes, and determine the optimal polarization mode band for ENVISAT ASAR images, which is the optimal polarization mode for radar images extracted from peat swamps band;

Step 8: Carry out multi-layer and multi-scale segmentation on the preprocessed Landsat TM data, slope data and the optimal polarization mode band of ENVISATASAR image determined in step 7 to obtain a series of segmentation units;

Step 9: Extract feature parameters from a series of segmentation units completed in step 8; wherein, the feature parameters include the average value of each band, normalized vegetation index, normalized water index, TM2+TM3-TM4-TM5 and hue ;

Step 10: According to the feature parameters extracted in step 9, use the JM distance method to determine the best classification band;

Step 11: According to the optimal classification band determined in step 10, a classification decision tree is established with reference to the survey sample points of land cover types; among them, the survey sample points of reference land cover types include peat swamp, herbaceous swamp, residential area, traffic land, farmland , forest land, water body land cover type;

Step 12: Run the classification decision tree, export the land cover type classification results, and produce the land cover type vector file; where the land cover type vector file includes farmland, forest land, water body, residential traffic land, herbaceous swamp and peat swamp land cover types ;

Step 13: Make a peat swamp thematic map based on the land cover type vector file completed in step 12; that is, a peat swamp information extraction method based on ENVISAT ASAR, Landsat TM and DEM data is completed. the

Invention effect

Distinguish other swamp types that are easily confused with peat information, and automatically and quickly and accurately extract the spatial distribution information of peat swamps in medium-resolution remote sensing images (Landsat TM ), so as to realize the method of automatic extraction of peat swamp information for thematic mapping. the

Combining radar images with optical images, taking terrain factors as the control factors of peat swamps, comprehensively using object-oriented and decision tree remote sensing classification methods to obtain the characteristic parameters of objects, and selecting the best classification band by JM distance method to establish a decision tree Complete the classification and make a peat swamp thematic map. the

Based on Landsat TM data, ENVISAT ASAR images and DEM data, the present invention applies object-oriented and decision tree classification methods to the automatic extraction of peat swamp information, and merges independent pixels into homogeneous objects. In the process of object segmentation, not only Considering spectral features, texture features and topological features, and then by selecting the optimal band and establishing a decision tree, the spatial distribution information of peat swamps is gradually obtained. The accuracy of the obtained classification results is 93%, which is 5%-8% higher than that of the existing method of extracting peat bogs using medium-resolution remote sensing images. At the same time, considering the influence of topography on the distribution of peat bogs, the classification results have clear geographical significance. The invention overcomes the phenomenon of omission and misclassification that occurred when only optical images were used in the extraction of peat swamp information in the past, and also solves the "salt and pepper phenomenon" and "enclave phenomenon" in the peat swamp spatial information obtained by classification, and does not have a clear geographical location. questions of meaning. The invention has practical significance for the combination of optical image, radar image and DEM to quickly and automatically extract peat swamp information. the

Description of drawings

Fig. 1 is a flow chart of a method for extracting peat swamp information based on ENVISAT ASAR, Landsat TM and DEM data proposed by Embodiment 1;

Fig. 2 is a thematic map of peat swamps made from land cover type vector files proposed in Embodiment 1. the

Detailed ways

Specific embodiment one: a kind of peat swamp information extraction method based on ENVISAT ASAR, Landsat TM and DEM data of this embodiment is specifically prepared according to the following steps:

Step 1: Preprocessing the Landsat TM data;

Step 2: Preprocessing the ENVISAT ASAR data;

Step 3: Resample the ENVISAT ASAR data preprocessed in Step 2 in ArcGIS, and the resampled ENVISAT ASAR data has the same grid size as the Landsat TM data processed in Step 1;

Step 4: Based on the preprocessed Landsat TM data, compare the preprocessed Landsat TM data and the resampled ENVISAT ASAR data in the ArcGIS software, and use the function of adding control points provided by the Georeferencing module of the ArcGIS software Select the control points on the preprocessed Landsat TM data, and register the resampled ENVISAT ASAR data according to the control point space to obtain the ENVISAT ASAR image;

Step 5: Use the Aspect command in Surface Analysis under the Spatial Analyst module in ArcGIS software to extract the slope from the DEM data to obtain slope data;

Step 6: Based on the land cover type survey sample points, extract the backscatter coefficients of the radar image under different polarization modes for different land cover types in ArcGIS for the ENVISAT ASAR image completed in step 4;

Step 7: Analyze the difference in radar backscatter coefficients between peat swamps and other different land cover types under different polarization modes, and determine the optimal polarization mode band for ENVISAT ASAR images, which is the optimal polarization mode for radar images extracted from peat swamps band;

Step 8: Use eCognition software to perform multi-layer and multi-scale segmentation on the preprocessed Landsat TM data, slope data, and the optimal polarization band of the ENVISAT ASAR image determined in step 7, to obtain a series of segmentation units, and use each segmentation unit as an object;

Step 9: Use eCognition software to extract the characteristic parameters of a series of segmentation units completed in step 8; wherein, the characteristic parameters include the average value of each band, normalized difference vegetation index (NDVI), normalized difference water index (NDWI), TM2+TM3-TM4-TM5, hue(R:G:B＝TM5:TM4:TM3) and hue(R:G:B＝TM4:TM3:TM2);

Step 10: According to the feature parameters extracted in step 9, use the JM distance (Jeffreys Matusita Distance) method to determine the best classification band;

Step 11: According to the best classification band determined in step 10, refer to the survey sample points of land cover type, and use See5.0 software to establish a classification decision tree; among them, the reference land cover type survey sample points include peat swamp, herbaceous swamp, residential area , transportation land, farmland, forest land, water body and other land cover types;

Step 12: Run the classification decision tree in eCognition software, export the classification results of land cover types, and produce land cover type vector files; among them, the land cover type vector files include farmland, forest land, water body, residential traffic land, herbaceous swamp and peat swamp other land cover types;

Step 13: In the Layout View mode of the ArcGIS software, make a peat swamp thematic map according to the land cover type vector file completed in step 12 (the schematic diagram of the peat swamp thematic map is shown in Figure 2); that is, a ENVISAT-based The peat swamp information extraction method of ASAR, Landsat TM and DEM data is shown in Figure 1. the

The effect of this implementation mode:

Distinguish other swamp types that are easily confused with peat information, and automatically and quickly and accurately extract the spatial distribution information of peat swamps in medium-resolution remote sensing images (Landsat TM ), so as to realize the method of automatic extraction of peat swamp information for thematic mapping. the

Combining radar images with optical images, taking terrain factors as the control factors of peat swamps, comprehensively using object-oriented and decision tree remote sensing classification methods to obtain the characteristic parameters of objects, and selecting the best classification band by JM distance method to establish a decision tree Complete the classification and make a peat swamp thematic map. the

This embodiment is based on Landsat TM data, ENVISAT ASAR images and DEM data, and applies object-oriented and decision tree classification methods to the automatic extraction of peat swamp information, and merges independent pixels into homogeneous objects. Not only spectral features, but also texture features and topological features are considered, and then the spatial distribution information of peat swamps is gradually obtained by selecting the optimal band and establishing a decision tree. The accuracy of the obtained classification results is 93%, which is 5%-8% higher than that of the existing method of extracting peat bogs using medium-resolution remote sensing images. At the same time, considering the influence of topography on the distribution of peat bogs, the classification results have clear geographical significance. This embodiment overcomes the phenomenon of omission and misclassification that occurred when only optical images were used in the extraction of peat swamp information in the past, and also solves the "salt and pepper phenomenon" and "enclave phenomenon" in the classified peat swamp spatial information, which does not have a clear issues of geographical significance. This embodiment has practical significance for the combination of optical images, radar images and DEM to quickly and automatically extract peat swamp information. the

Specific implementation mode two: the difference between this implementation mode and specific implementation mode one is: the preprocessing process is carried out to Landsat TM data in step one:

(1) Within the peat swamp distribution range, determine the track number of the Landsat TM data of the peat swamp, and download the Landsat TM data covering the peat swamp distribution range according to the track number;

(2) In order to eliminate terrain distortion, the Landsat TM data is orthorectified by using the DEM data in the corresponding area of the Landsat TM data to obtain the Landsat TM data after orthorectification;

(3) In order to eliminate geometric distortion, using terrain data, select ground control points in ERDAS software, and perform geometric fine correction on Landsat TM data after orthorectification to obtain preprocessed Landsat TM data. Other steps and parameters are the same as those in Embodiment 1. the

Specific implementation mode three: the difference between this implementation mode and specific implementation mode one or two is: the preprocessing process is carried out to ENVISAT ASAR data in step two:

(1) Within the coverage of the Landsat TM data range, download the ENVISAT ASAR fine image first-level data (ENVISAT ASAR APP Level 1B data) used in the test (the polarization mode is HH and HV);

(2) Carrying out radiometric calibration on the ENVISAT ASAR fine image first-level data is to convert the DN value of the ENVISAT ASAR fine image first-level data into the backscatter coefficient (in dB) to obtain the radiation-corrected ENVISAT ASAR data; its radiometric calibration The standard formula is as follows:

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}^{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; log</span>}}_{\mathrm{<span\; class="notranslate">\; 10</span>}}<span\; class="notranslate">\; [</span>\frac{{\mathrm{<span\; class="notranslate">\; DN</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; K</span>}}\mathrm{<span\; class="notranslate">\; sin</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>$

in, is the backscattering coefficient of the pixel in row i and column j; DN _ij is the original intensity value of the pixel in row i and column j; θ _ij is the radar incident angle of the pixel in row i and column j; K is is the absolute calibration coefficient;

(3) In order to eliminate terrain distortion, use the DEM data of the corresponding area of ENVISAT ASAR data to use the Range-Doppler imaging algorithm (Range-Doppler) to perform terrain correction on the radiation-corrected ENVISAT ASAR data;

(4) In order to eliminate image noise, the Enhanced Lee filter (window size 3╳3 pixels) was applied to perform spatial filtering on the ENVISAT ASAR data after terrain correction. Other steps and parameters are the same as those in Embodiment 1 or Embodiment 2. the

Embodiment 4: This embodiment differs from Embodiments 1 to 3 in that: in Step 4, the spatial registration error is controlled within 0.5 pixels. Other steps and parameters are the same as those in Embodiments 1 to 3. the

Specific implementation mode five: the difference between this implementation mode and one of specific implementation modes one to four is that the optimal polarization mode band of the radar image extracted from the peat swamp in step seven is specifically: through statistics of different land cover types at the HV and HH poles The average value of the backscatter coefficients in the polarization mode is compared, and the difference between the backscatter coefficients is obvious as the optimal polarization mode band. Other steps and parameters are the same as in one of the specific embodiments 1 to 4. the

Embodiment 6: This embodiment differs from Embodiment 1 to Embodiment 5 in that each segmentation unit in Step 8 is composed of pixels that are adjacent in space and have a homogeneity of 80% to 100%. Other steps and parameters are the same as one of the specific embodiments 1 to 5. the

Specific embodiment seven: this embodiment is different from one of specific embodiments one to six in that: in step nine, utilize eCognition software to carry out the normalized normalized vegetation index (NDVI) and the normalized difference vegetation index (NDVI) that feature parameter extraction is carried out to a series of segmentation units that step eight obtains A chemical water body index (NDWI) is:

$\mathrm{<span\; class="notranslate">\; NDVI</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 3</span>}}{\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; DNWI</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 4</span>}}{\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 4</span>}}<span\; class="notranslate">\; ;</span>$

Among them, TM2 is the second band of the Landsat TM sensor, TM3 is the third band of the Landsat TM sensor, and TM4 is the fourth band of the Landsat TM sensor. Other steps and parameters are the same as those in Embodiments 1 to 6. the

Embodiment eight: this embodiment is different from one of embodiment one to seven: in step ten, the calculation formula of JM distance method (Jeffreys Matusita Distance) is as follows:

${\mathrm{<span\; class="notranslate">\; JM</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; =</span>\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ;</span>$

${\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 8</span>}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; t</span>}}\frac{{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; ln</span>}\frac{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; |</span>}{{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; ;</span>$

In the formula, i and j respectively represent any two different classification types; b _ij is the Bhattacharyya distance between the i-class classification type and the j-class classification type; m _i represents the mean vector of the i-class classification type, and m _j represents The mean value vector of category j; C _i represents the covariance matrix of category i, and C _j represents the covariance matrix of category j; the slope value (0°～42.43°) and normalization are determined according to the calculation results of JM The vegetation index (NDVI) (-1～1), TM2+TM3-TM4-TM5 (-200.67～37.77) and hue (R:G:B=TM5:TM4:TM3) (0～1) are the participating classification bands . Other steps and parameters are the same as one of the specific embodiments 1 to 7.

Adopt the following examples to verify the beneficial effects of the present invention:

Embodiment one:

In this embodiment, a peat swamp information extraction method based on ENVISAT ASAR, Landsat TM and DEM data is specifically prepared according to the following steps:

Step 1: Preprocessing Landsat TM data:

(1) Within the distribution range of the peat bog, determine the track number of the Landsat TM data of the peat bog, download the Landsat TM data covering the distribution range of the peat bog according to the track number, the track number is P119R26, and the time is June 11, 2010;

(2) In order to eliminate terrain distortion, the Landsat TM data is orthorectified by using the DEM data in the corresponding area of the Landsat TM data to obtain the Landsat TM data after orthorectification;

(3) In order to eliminate geometric distortion, use terrain data, select ground control points in ERDAS software, and perform geometric fine correction on Landsat TM data after orthorectification to obtain preprocessed Landsat TM data;

Step 2: Preprocessing the ENVISAT ASAR data;

(1) Within the coverage of the Landsat TM data range, download the ENVISAT ASAR fine image first-level data (ENVISAT ASAR APP Level 1B data) used in the test (the polarization mode is HH and HV), and the time is July 2, 2010;

(2) Carrying out radiometric calibration on ENVISAT ASAR fine image level-1 data is to convert the DN value of ENVISAT ASAR fine-image level-1 data into backscatter coefficient (in dB) to obtain radiometrically corrected ENVISAT ASAR data; its radiometric calibration The formula is as follows:

${\mathrm{<span\; class="notranslate">\; \&sigma;</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}^{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 10</span>}<span\; class="notranslate">\; \&Center\; Dot;</span>{\mathrm{<span\; class="notranslate">\; log</span>}}_{\mathrm{<span\; class="notranslate">\; 10</span>}}<span\; class="notranslate">\; [</span>\frac{{\mathrm{<span\; class="notranslate">\; DN</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\mathrm{<span\; class="notranslate">\; K</span>}}\mathrm{<span\; class="notranslate">\; sin</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; \&theta;</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ]</span>$

in, is the backscattering coefficient of the pixel in row i and column j; DN _ij is the original intensity value of the pixel in row i and column j; θ _ij is the radar incident angle of the pixel in row i and column j; K is is the absolute calibration coefficient;

(3) In order to eliminate terrain distortion, in the NEXT 4C software, use the DEM data of the corresponding area of ENVISAT ASAR data to use the Range-Doppler imaging algorithm (Range-Doppler) to perform terrain correction on the radiometric ENVISAT ASAR data;

(4) In order to eliminate image noise, apply the Enhanced Lee filter (window size 3╳3 pixels) to perform spatial filtering on the ENVISAT ASAR data after terrain correction;

Step 3: Resample the ENVISAT ASAR data preprocessed in Step 2 in ArcGIS. The resampled ENVISAT ASAR data has the same grid size as the Landsat TM data processed in Step 1. The grid size is 30m×30m;

Step 4: Based on the preprocessed Landsat TM data, compare the preprocessed Landsat TM data and the resampled ENVISAT ASAR data in the ArcGIS software, and use the function of adding control points provided by the Georeferencing module of the ArcGIS software Select the control points on the preprocessed Landsat TM data, and register the resampled ENVISAT ASAR data according to the control point space to obtain the ENVISAT ASAR image; the spatial registration error is controlled within 0.5 pixels;

Step 5: Use the Aspect command in Surface Analysis under the Spatial Analyst module in ArcGIS software to extract the slope from the DEM data to obtain slope data;

Step 6: Combined with the land cover type survey sample points, extract the backscatter coefficients of radar images under different polarization modes for different land cover types in ArcGIS from the ENVISAT ASAR image completed in step 4;

Step 7: Analyze the difference in radar backscatter coefficients between peat swamps and other different land cover types under different polarization modes, and determine the optimal polarization mode band for ENVISAT ASAR images, which is the optimal polarization mode for radar images extracted from peat swamps Band:

The average values of the backscatter coefficients of different land cover types under the HV and HH polarization modes are shown in Table 2:

Table 2

By comparison, it is found that the difference between the backscatter coefficients of peat bogs and other types of ground features under the HV polarization mode is more obvious than the difference between the backscatter coefficients of peat bogs and other types of ground features under the HH polarization mode. The best polarization mode band, so choose ENVISAT ASAR image HV polarization mode band. In addition, the backscattering coefficients of herbaceous swamps and peat bogs differ by 0.88dB under HH polarization mode, while the difference of backscattering coefficients between them is 3.44dB under HV polarization mode, so this study is suitable for studying herbaceous swamps;

Step 8: Use eCognition software to perform multi-layer and multi-scale segmentation on the preprocessed Landsat TM data, slope data, and the optimal polarization band of the ENVISAT ASAR image determined in step 7, to obtain a series of segmentation units, and use each segmentation unit as An object; each segmentation unit consists of spatially adjacent and homogeneous pixels of 80% to 100%. The parameter settings of multi-scale segmentation in the process of object-oriented classification are shown in Table 1:

Segmentation scale color factor form factor smoothness firmness 8 0.9 0.1 0.6 0.4

Step 9: Use eCognition software to extract the characteristic parameters of a series of segmentation units completed in step 8; wherein, the characteristic parameters include the average value of each band, normalized difference vegetation index (NDVI), normalized difference water index (NDWI), TM2+TM3-TM4-TM5, Hue(R:G:B＝TM5:TM4:TM3), Hue(R:G:B＝TM4:TM3:TM2);

The normalized normalized vegetation index (NDVI) and normalized normalized water index (NDWI) for feature parameter extraction of a series of segmentation units are:

$\mathrm{<span\; class="notranslate">\; NDVI</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 3</span>}}{\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 4</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 3</span>}}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; DNWI</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 4</span>}}{\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; tm</span>}\mathrm{<span\; class="notranslate">\; 4</span>}}<span\; class="notranslate">\; ;</span>$

Among them, TM2 is the second band of the Landsat TM sensor, TM3 is the third band of the Landsat TM sensor, and TM4 is the fourth band of the Landsat TM sensor;

Step 10: According to the feature parameters extracted in step 9, use the JM distance (Jeffreys Matusita Distance) method to determine the best classification band; the JM distance method (Jeffreys Matusita Distance) calculation formula is as follows:

${\mathrm{<span\; class="notranslate">\; JM</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; =</span>\sqrt{\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; e</span>}}^{<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}}<span\; class="notranslate">\; )</span>}<span\; class="notranslate">\; ;</span>$

${\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; ij</span>}}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 8</span>}{<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; t</span>}}\frac{{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}}{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; m</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; ln</span>}\frac{<span\; class="notranslate">\; |</span><span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; |</span>}{{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}}{<span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; c</span>}}_{\mathrm{<span\; class="notranslate">\; j</span>}}<span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; ;</span>$

In the formula, i and j respectively represent any two different classification types; b _ij is the Bhattacharyya distance between the i-class classification type and the j-class classification type; m _i represents the mean vector of the i-class classification type, and m _j represents The mean vector of the j category classification type; C _i represents the covariance matrix of the i category classification type, and C _j represents the covariance matrix of the j category classification type; the value range of the JM distance is between 0 and 2.0, when the value is greater than 1.9 It shows that the separability between ground features is good; according to the calculation results of JM, the slope value (0°～42.43°), normalized difference vegetation index (NDVI) (-1～1), TM2+TM3-TM4-TM5 (-200.67～37.77) and hue (R:G:B=TM5:TM4:TM3) (0～1) are bands involved in classification.

Step 11: According to the best classification band determined in step 10, refer to the survey sample points of land cover type, and use See5.0 software to establish a classification decision tree; among them, the reference land cover type survey sample points include peat swamp, herbaceous swamp, residential area , transportation land, farmland, forest land, water body and other land cover types;

Step 12: Run the classification decision tree in the eCognition software, export the land cover type classification results, and produce the land cover type vector file, the peat swamp extraction accuracy is 93%; among them, the land cover type vector file includes farmland, forest land, water body, residential Land cover types such as traffic land, herbaceous swamps and peat bogs;

Step 13: In the Layout View mode of the ArcGIS software, make a peat swamp thematic map based on the land cover type vector file completed in step 12 (the schematic diagram of the peat swamp thematic map is shown in Figure 2); that is, a ENVISAT-based Peat bog information extraction method from ASAR, Landsat TM and DEM data. the

The present invention can also have other various embodiments, without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and deformations according to the present invention, but these corresponding changes and deformations are all Should belong to the scope of protection of the appended claims of the present invention. the

Claims (8)

1. A peat bog information extraction method based on ENVISAT ASAR, Landsat TM and DEM data is characterized in that:

the method comprises the following steps: preprocessing Landsat TM data;

step two: ENVISAT ASAR preprocessing the data;

step three: resampling the ENVISAT ASAR data preprocessed in the step two, wherein the grid size of the ENVISAT ASAR data after resampling is consistent with that of the Landsat TM data processed in the step one;

step four: selecting a control point on the preprocessed Landsat TM data by utilizing a control point adding function provided by a Georeferencening module of ArcGIS software, and registering the resampled ENVISAT ASAR data according to a control point space to obtain a ENVISAT ASAR image;

step five: gradient extraction is carried out on the DEM data to obtain gradient data;

step six: combining with the land cover type survey sampling points, extracting the backward scattering coefficients of the radar images of different land cover types in different polarization modes from the ENVISAT ASAR images completed in the step four;

step seven: analyzing the difference of radar backscattering coefficients of the peat bogs and other different land cover types under different polarization modes, and determining ENVISAT ASAR image optimal polarization mode wave bands, namely radar image optimal polarization mode wave bands for extracting the peat bogs;

step eight: performing multilayer multi-scale segmentation on the preprocessed Landsat TM data, gradient data and the ENVISATASAR image optimal polarization mode wave band determined in the seventh step to obtain a series of segmentation units;

step nine: extracting characteristic parameters of a series of segmentation units segmented in the step eight; the characteristic parameters comprise the average value of each wave band, a normalized vegetation index, a normalized water body index, TM2+ TM3-TM4-TM5 and color tones;

step ten: determining the optimal classification wave band by using a JM distance method according to the characteristic parameters extracted in the step nine;

step eleven: according to the optimal classification wave band determined in the step ten, establishing a classification decision tree by referring to the land cover type survey sampling points; wherein, the reference soil coverage type survey sample points comprise peat swamp, herbaceous swamp, residential land, transportation land, farmland, forest land and water body soil coverage types;

step twelve: operating a classification decision tree, exporting a land cover type classification result, and producing a land cover type vector file; wherein, the land cover type vector file comprises land cover types of farmlands, forest lands, water bodies, residential traffic lands, herbaceous swamps and peat swamps;

step thirteen: making a peat marsh thematic map according to the land cover type vector file completed in the step twelve; thus completing a peat bog information extraction method based on ENVISAT ASAR, Landsat TM and DEM data.

2. The method for extracting peat bog information based on ENVISAT ASAR Landsat TM and DEM data as claimed in claim 1, wherein: the preprocessing process of the Landsat TM data in the first step is as follows:

(1) determining the track number of Landsat TM data of the peat bogs in the distribution range of the peat bogs, and downloading the Landsat TM data covering the distribution range of the peat bogs according to the track number;

(2) performing orthorectification on the Landsat TM data by using DEM data of an area corresponding to the Landsat TM data to obtain the orthorectified Landsat TM data;

(3) and selecting a ground control point in the ERDAS software by utilizing the topographic data, and performing geometric fine correction on the directly corrected Landsat TM data to obtain preprocessed Landsat TM data.

3. The method for extracting peat bog information based on ENVISAT ASAR Landsat TM and DEM data as claimed in claim 1, wherein: and in the second step, ENVISAT ASAR data are preprocessed:

(1) downloading ENVISAT ASAR fine image primary data within the coverage range of the Landsat TM data range;

(2) carrying out radiometric calibration on ENVISAT ASAR fine image primary data, namely converting DN value of ENVISAT ASAR fine image primary data into backscattering coefficient to obtain ENVISAT ASAR data of radiation correction; the radiometric calibration formula is as follows:

<math> <mrow> <msubsup> <mi>&sigma;</mi> <mi>ij</mi> <mn>0</mn> </msubsup> <mo>=</mo> <mn>10</mn> <mo>&CenterDot;</mo> <msub> <mi>log</mi> <mn>10</mn> </msub> <mo>[</mo> <mfrac> <msubsup> <mi>DN</mi> <mi>ij</mi> <mn>2</mn> </msubsup> <mi>K</mi> </mfrac> <mi>sin</mi> <mrow> <mo>(</mo> <msub> <mi>&theta;</mi> <mi>ij</mi> </msub> <mo>)</mo> </mrow> <mo>]</mo> </mrow> </math>

wherein,the backscattering coefficient of the ith row and the jth column of pixels; DN_ijThe original intensity value of the ith row and the jth column of pixels; theta_ijIs the incident angle of the radar wave of the ith row and the jth column of pixels; k is an absolute calibration coefficient;

(3) performing terrain correction on radiation correction ENVISAT ASAR data by utilizing DEM data of a region corresponding to ENVISAT ASAR data and adopting a range-Doppler imaging algorithm;

(4) and applying an Enhanced Lee filter to perform spatial filtering processing on ENVISAT ASAR data after the terrain correction is completed.

4. The method for extracting peat bog information based on ENVISAT ASAR Landsat TM and DEM data as claimed in claim 1, wherein: and the spatial registration error in the fourth step is controlled within 0.5 pixel.

5. The method for extracting peat bog information based on ENVISAT ASAR Landsat TM and DEM data as claimed in claim 1, wherein: the optimal polarization mode wave band of the radar image for extracting the peat bogs in the step seven is specifically as follows: the difference between the backscattering coefficients is obviously compared by counting the average value of the backscattering coefficients of different land cover types under the HV and HH polarization modes to be used as the optimal polarization mode wave band.

6. The method for extracting peat bog information based on ENVISAT ASAR Landsat TM and DEM data as claimed in claim 1, wherein: and step eight, each partition unit consists of spatially adjacent pixels with the homogeneity of 80-100%.

7. The method for extracting peat bog information based on ENVISAT ASAR Landsat TM and DEM data as claimed in claim 1, wherein: in the ninth step, the normalized vegetation index NDVI and the normalized water body index NDWI for extracting the characteristic parameters of the series of segmentation units obtained in the eighth step are as follows:

$\mathrm{NDVI}=\frac{\mathrm{TM}4-\mathrm{TM}3}{\mathrm{TM}4+\mathrm{TM}3},\mathrm{NDWI}=\frac{\mathrm{TM}2-\mathrm{TM}4}{\mathrm{TM}2+\mathrm{TM}4};$

wherein, TM2 is the 2 nd band of the Landsat TM sensor, TM3 is the 3 rd band of the Landsat TM sensor, and TM4 is the 4 th band of the Landsat TM sensor.

8. The method for extracting peat bog information based on ENVISAT ASAR Landsat TM and DEM data as claimed in claim 1, wherein: in the step ten, a calculation formula of the JM distance method is as follows:

${\mathrm{JM}}_{\mathrm{ij}}=\sqrt{2(1-e^{-b_{\mathrm{ij}}});}$

$b_{\mathrm{ij}}=1/8{(m_i-m_j)}^t\frac{c_i+c_j}{2}(m_i-m_j)+1/2\ln \frac{|(c_i-c_j)/2|}{{\left|c_i\right|}^{1/2}{\left|c_j\right|}^{1/2}};$

wherein i and j represent any two different classification types respectively; b_ijThe Bhattacharyya distance between the i-class classification type and the j-class classification type; m is_iMean vector, m, representing class i class classification type_jA mean vector representing class j classification types; c_iCovariance matrix representing class i class type, C_jA covariance matrix representing the class j classification type.