Automatic remote sensing classification method, device and equipment for drawing vegetation type wetland coverage on national scale and medium

Abstract

The invention discloses an automatic remote sensing classification method, device and equipment for drawing vegetation type wetland coverage on the national scale and a medium. The method comprises the following steps: constructing climate-hydrological units for wetland classification, and dividing a country into different climate-hydrological units by using a random forest classifier based on climate and hydrological auxiliary data; appropriate parameters are selected, Savitzky-Golay filtering preprocessing is carried out on the data of the time sequence, and complete dense time sequence NDVI data are established; establishing a vegetation type wetland classification system, and constructing reference NDVI phenological curves of different wetland categories in each climate-hydrological unit based on a dense time sequence NDVI; and based on a feature matching method, classifying the vegetation-type wetland in the to-be-classified region in combination with the reference NDVI phenological curve. The vegetation type wetland classification method is established, and the requirement for large-scale wetland mapping is met.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image information processing, and is an automated remote sensing classification method, device, equipment and medium for mapping vegetation-type wetland coverage at a national scale. Background Art

[0002] In the existing wetland datasets, the classification accuracy of vegetated wetlands is often lower than that of non-vegetated wetlands, and they are easily confused with non-wetland vegetation. Due to the coverage of vegetation, vegetated wetlands have obvious seasonal changes, which is also an important reason why they are more difficult to accurately classify than other wetland types. In addition, the development of existing wetland mapping data products does not make sufficient use of wetland phenological information and relies heavily on a large number of samples.

[0003] In the land use / cover classification, wetlands have the lowest accuracy among all land cover categories. For example, in GlobeLand30, the mapping accuracy of forest land and farmland is 83.58% and 85.70% respectively, while the mapping accuracy of wetlands is only 74.87%; in GLC_FCS30, the mapping accuracy of farmland, forest land, bare land and water is higher than 80%, while the classification accuracy of wetlands is only 61.8%. In wetland classification products, the classification accuracy of vegetation wetlands is also low among all wetland categories. For example, in EA_Wetlands, the mapping accuracy of inland and coastal woody and herbaceous swamps is only about 70%, while the mapping accuracy of rivers and lakes is above 85%; in GWL_FCS30 products, the classification accuracy of woody, herbaceous swamps and salt marshes is less than 80%, while the classification accuracy of permanent water bodies, salt pans and tidal flats can reach 90%. The classification system of vegetation wetlands in most wetland mapping products is also insufficient, and often ignores the important artificial vegetation wetlands - rice.

[0004] In addition, existing large-scale wetland mapping work relies heavily on samples. The sample size of CAS_Wetlands is 16,496, the sample size of EA_Wetlands is 15,808, and the sample size of GWL_FCS30 even exceeds 20 million (globally). The production of large-scale wetland samples is a time-consuming and laborious task, and for highly dynamic wetlands, the inter-annual stability of samples cannot be guaranteed. Therefore, in the long-term mapping work of many years, samples need to be repeatedly selected, which greatly affects the efficiency of updating mapping products. Summary of the invention

[0005] In order to address the deficiencies of the above-mentioned prior art and the scientific research and management needs for rapid acquisition of large-scale wetland coverage information, the purpose of the present invention is to provide an automated remote sensing classification method, device, equipment and medium for mapping vegetation-type wetland coverage at a national scale.

[0006] The technical solution of the present invention is: an automated remote sensing classification method for mapping vegetation-type wetland cover at a national scale, comprising the following steps: Step 1: Construction of climate-hydrological units for wetland classification. Based on climate and hydrological auxiliary data, the country is divided into different climate-hydrological units using a random forest classifier. Step 2: Preprocess the time series data using Savitzky-Golay filtering to establish complete dense time series NDVI data; Step 3: Establish a vegetated wetland classification system, and construct reference phenological curves of different wetland categories in each climate-hydrological unit based on dense time series NDVI data; Step 4: Classify the vegetation-type wetlands in the classified area based on the feature matching method and the wetland category reference phenological curve.

[0007] The present invention also proposes an automated remote sensing classification device for mapping vegetation-type wetland coverage at a national scale, comprising: The climate-hydrological unit construction module is used to construct climate-hydrological units for wetland classification, including: using random forest classifiers to divide countries into different climate-hydrological units based on climate and hydrological auxiliary data; The filter preprocessing module is used to perform Savitzky-Golay filter preprocessing on the time series data to establish a complete dense time series NDVI data; Vegetation-type wetland classification system establishment module, used to establish a vegetation-type wetland classification system, including: based on dense time series NDVI data, constructing wetland category reference phenological curves of different wetland categories in each climate-hydrological unit; The vegetation type wetland classification module classifies the vegetation type wetlands in the classified area based on the feature matching method and the wetland category reference phenological curve.

[0008] The present invention also proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the above-mentioned automated remote sensing classification method for mapping vegetation-type wetland coverage at a national scale are implemented.

[0009] A non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned automated remote sensing classification method for mapping vegetation-type wetland cover at a national scale.

[0010] The present invention has the following beneficial effects: Based on Sentinel-2 remote sensing data, as well as climate and hydrological data, the present invention has established a vegetation type wetland classification method to meet the needs of large-scale wetland mapping. Based on the differences and similarities in climate and hydrological conditions, the country is divided into primary and secondary climate-hydrological units to ensure that the climate and hydrological conditions in each unit are relatively uniform. The NDVI time series data is processed using Savitzky-Golay filtering to reduce the disturbance caused by external factors, and the reference phenological curve of the vegetation type wetland category in each climate-hydrological unit is obtained. Through the feature matching algorithm, rapid classification and mapping of vegetation type wetlands can be achieved. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] Figure 1 is a flow chart of the method of the present invention; Figure 2 This is an example of the classification result of the present invention. DETAILED DESCRIPTION

[0012] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the protection scope of the present invention.

[0013] The present invention provides an automated remote sensing classification method for mapping vegetation wetland coverage at a national scale. This classification method makes full use of the time dimension information of dense time series remote sensing data and can realize automated mapping of large-scale vegetation wetlands.

[0014] like Figure 1 As shown in FIG. 1 , a flow chart of an automated remote sensing classification method for mapping vegetation-type wetland coverage at a national scale according to the present invention is shown, which includes the following steps: Step 1: Construction of climate-hydrological units for wetland classification: Based on auxiliary data such as climate and hydrology, the country is divided into different climate-hydrological units using a random forest classifier. Specifically, it includes: Step 1.1, collect sample data for the division of climate-hydrological units. Use the national hydrological zoning map of a certain year as the basis, which includes the primary and secondary hydrological zoning. Establish a 50km wide buffer zone inward at the boundary of each secondary hydrological zoning, randomly generate sample points in the area outside the buffer zone, and assign each sample point the primary and secondary hydrological zoning numbers as category labels. Randomly select 30% of all samples as validation samples and 70% as training samples.

[0015] Step 1.2: Process auxiliary data. Collect publicly available national-level climate and hydrological data such as monthly precipitation, monthly average temperature, and monthly potential evapotranspiration, and obtain multiple derived data such as multi-year average temperature and multi-year monthly average temperature through methods such as mean synthesis and maximum synthesis.

[0016] Step 1.3: First-level climate-hydrological unit division: Using the random forest classifier, the auxiliary data are used as classification data and combined with the training samples to divide the world into 11 first-level climate-hydrological units.

[0017] Step 1.4: Secondary climate-hydrological unit division: In each primary climate-hydrological unit, a random forest classifier was used, with the auxiliary data as classification data and combined with the training samples to divide the country into 55 primary climate-hydrological units.

[0018] Step 2: Preprocess the time series data using Savitzky-Golay filtering: Step 2.1, obtain Sentinel-2 images from December 1st of the year before the target year to January 31st of the year after the target year, perform cloud removal and cropping preprocessing operations on each original image, and calculate the original NDVI time series using the formula NDVI=(B8-B4) / (B8+B4), where B4 and B8 are the red and near-infrared bands of Sentinel-2; Step 2.2: Perform linear time interpolation on the original NDVI time series to fill in the image gaps caused by cloud removal and obtain complete NDVI time series data that is continuous in time and space. The formula for linear time interpolation is: ; Among them, y is the missing value that needs to be interpolated, y1 is the data before the missing value, y2 is the data after the missing value, and t, t1, and t2 are the corresponding times.

[0019] Step 2.3: Use Savitzky-Golay filtering on the interpolated NDVI time series data, where the filter window size is set to 25 and the fitting function order is set to 3.

[0020] Step 3: Establish a vegetated wetland classification system, and construct reference phenological curves of different wetland categories in each climate-hydrological unit based on dense time series NDVI; Step 3.1: Construction of classification system. Vegetated wetlands are divided into herbaceous swamps, woody swamps, salt marshes, mangroves, single-season rice and double-season rice, and the wetland categories included in each secondary climate-hydrological zone are determined based on the existing wetland products; Step 3.2: Sample point selection: Use multiple existing wetland mapping products to obtain wetland samples, and use the Mahalanobis distance method to screen samples; Step 3.3: Construct a wetland reference phenological curve. Obtain the NDVI time series curve at the sample point location, and synthesize the mean of the curves of the sample points of the same category to form a wetland category reference phenological curve.

[0021] Step 4: Classify the vegetation-type wetlands in the classified area based on the feature matching method and the wetland category reference phenological curve.

[0022] Step 4.1, calculate the Euclidean distance between the NDVI time series curve of each pixel to be classified and the reference phenological curve of each wetland category, and output a raster image with the pixel value as the Euclidean distance value; Step 4.2, calculate the histogram of the raster image, and set the Euclidean distance value at the first inflection point on the histogram as the threshold; Step 4.3: Use the threshold to segment the image output in 4.1, and retain the part smaller than the threshold, which is the wetland category to be extracted.

[0023] According to one embodiment of the present invention, the vegetation type wetland classification system in a certain area is woody swamp, herbaceous swamp, single-season rice, and double-season rice.

[0024] According to an embodiment of the present invention, the classification result is as follows: Figure 2 shown.

[0025] Although the above describes the illustrative specific embodiments of the present invention to facilitate those skilled in the art to understand the present invention, and it should be clear that the present invention is not limited to the scope of the specific embodiments, for those of ordinary skill in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the attached claims, these changes are obvious, and all inventions and creations using the concept of the present invention are protected.

Claims

1. An automated remote sensing classification method for mapping vegetated wetland cover at a national scale, characterized in that: The steps include: Step 1: Construction of climate-hydrological units for wetland classification. Based on climate and hydrological auxiliary data, the country is divided into different climate-hydrological units using a random forest classifier. Step 2: Preprocess the time series data using Savitzky-Golay filtering to establish complete dense time series NDVI data; Step 3: Establish a vegetated wetland classification system, and construct reference phenological curves of different wetland categories in each climate-hydrological unit based on dense time series NDVI data; Step 4: Classify the vegetation-type wetlands in the classified area based on the feature matching method and the wetland category reference phenological curve.

2. The automated remote sensing classification method for mapping vegetation-type wetland cover at a national scale according to claim 1, characterized in that: The step 1 specifically includes: Step 1.1, collect sample data for dividing climate-hydrological units; use the national hydrological zoning map as the basis, which includes the primary and secondary hydrological zoning, establish a width buffer inward at the boundary of each secondary hydrological zoning, randomly generate sample points in the area outside the buffer zone, and assign each sample point the primary and secondary hydrological zoning numbers as category labels, randomly select 30% of all samples as validation samples, and 70% as training samples; Step 1.2, process auxiliary data; collect public national-level data on monthly precipitation, monthly average temperature, and monthly potential evapotranspiration, and obtain multiple derived data including multi-year average temperature and multi-year monthly average temperature through mean synthesis and maximum synthesis methods; Step 1.3: First-level climate-hydrological unit division: Use the random forest classifier in Google Earth Engine cloud platform or ENVI, take the auxiliary data as classification data, and combine with the training samples to divide the country into n first-level climate-hydrological units; Step 1.4, secondary climate-hydrological unit division: In each primary climate-hydrological unit, use the random forest classifier, take the auxiliary data as classification data, and combine with the training samples to divide the country into m secondary climate-hydrological units, where m>n.

3. The automated remote sensing classification method for mapping vegetation-type wetland cover at a national scale according to claim 2, characterized in that: The step 2 specifically includes: Step 2.1, obtain Sentinel-2 images from December 1st of the year before the target year to January 31st of the year after the target year, perform cloud removal and cropping preprocessing operations on each original image, and calculate the original NDVI time series using the formula NDVI=(B8-B4) / (B8+B4), where B4 and B8 are the red and near-infrared bands of Sentinel-2; Step 2.2, perform linear time interpolation on the original NDVI time series to fill in the image gaps caused by cloud removal and obtain complete NDVI time series data that is continuous in time and space; Step 2.3: Apply Savitzky-Golay filtering to the interpolated NDVI time series data.

4. The automated remote sensing classification method for mapping vegetation-type wetland cover at a national scale according to claim 3, characterized in that: The step 3 specifically includes: Step 3.1: Construction of classification system; categorize vegetation-type wetlands into herbaceous swamps, woody swamps, salt marshes, mangroves, single-season rice, and double-season rice. Based on the existing wetland products, determine the wetland categories included in each secondary climate-hydrological zoning division. Step 3.2, sample point selection: use multiple existing wetland mapping products to obtain wetland samples, calculate the Mahalanobis distance and P-value between samples, and remove samples with a P-value lower than 0.001; Step 3.3, construct the reference phenological curve of wetland categories; obtain the NDVI time series curve at the sample point location, and synthesize the mean of the curves of the sample points of the same category to obtain the reference phenological curve of the wetland category.

5. The automated remote sensing classification method for mapping vegetation-type wetland cover at a national scale according to claim 4, characterized in that: The step 4 specifically includes: Step 4.1, calculate the Euclidean distance between the NDVI time series curve of each pixel to be classified and the reference phenological curve of each wetland category, and output a raster image with the pixel value as the Euclidean distance value; Step 4.2, calculate the histogram of the raster image, and set the Euclidean distance value at the first inflection point on the histogram as the threshold; Step 4.3: Use the threshold to segment the output raster image and retain the part smaller than the threshold, which is the wetland category to be extracted.

6. The automated remote sensing classification method for mapping vegetation-type wetland cover at a national scale according to claim 2, characterized in that: The mean synthesis and maximum synthesis methods are to statistically analyze the data of multiple images at each pixel position, take the mean and maximum value to generate a new image.

7. The automated remote sensing classification method for mapping vegetation-type wetland cover at a national scale according to claim 2, characterized in that: The formula for linear time interpolation is: ； Among them, y is the missing value that needs to be interpolated, y1 is the data before the missing value, y2 is the data after the missing value, and t, t1, and t2 are the corresponding times.

8. The automated remote sensing classification method for mapping vegetation wetland cover at a national scale according to claim 3, characterized in that: In the Savitzky-Golay filter, the filter window size is set to 25 and the fitting function order is set to 3.

9. An automated remote sensing classification device for mapping vegetation wetland cover at a national scale, characterized in that: include: The climate-hydrological unit construction module is used to construct climate-hydrological units for wetland classification, including: using random forest classifiers to divide countries into different climate-hydrological units based on climate and hydrological auxiliary data; The filter preprocessing module is used to perform Savitzky-Golay filter preprocessing on the time series data to establish a complete dense time series NDVI data; Vegetation-type wetland classification system establishment module, used to establish a vegetation-type wetland classification system, including: based on dense time series NDVI data, constructing wetland category reference phenological curves of different wetland categories in each climate-hydrological unit; The vegetation type wetland classification module classifies the vegetation type wetlands in the classified area based on the feature matching method and the wetland category reference phenological curve.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of the automated remote sensing classification method for mapping vegetated wetland coverage at a national scale as described in any one of claims 1 to 8 are implemented.

11. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the automated remote sensing classification method for mapping vegetated wetland cover at a national scale as described in any one of claims 1 to 8 are implemented.

Images