Method for generating vegetation community classification map and related device

Abstract

The invention discloses a vegetation community classification map generation method and a related device, and relates to the technical field of electronic information. Comprising the following steps: acquiring a plurality of radar images and / or a plurality of spectral images of a target wetland as a plurality of images of the target wetland; extracting a feature map from each image of the target wetland to obtain a plurality of feature images; screening out a preset number of images with the maximum importance value from the plurality of feature images as a plurality of target feature images; and generating a vegetation community classification map of the target wetland according to the plurality of target feature images. The radar image is less influenced by weather, so that the structural characteristics of vegetation can be effectively reflected; the spectral image has higher spatial resolution and higher time resolution, and can accurately capture the color and texture details of the vegetation; therefore, the plurality of images of the target wetland obtained in the application can well reflect the characteristics of the vegetation community; by adopting the method provided by the invention, a relatively accurate wetland vegetation community classification image can be obtained.

Description

Technical Field

[0001] The present application relates to the field of electronic information technology, and in particular to a method for generating a vegetation community classification map and related devices. Background Art

[0002] Wetlands are ecological units that maintain the biodiversity of aquatic plants and terrestrial wildlife and are important ecosystems. Wetland ecosystems composed of grass flats, mud flats and water bodies provide habitats and foraging places for fish and winter migratory birds; different vegetation communities are crucial for wintering water birds to choose suitable foraging and nesting environments. Vegetation community classification maps are maps that show the geographical distribution of various vegetation communities and their distribution patterns. Drawing high-precision vegetation community classification maps of wetlands is of great significance for the protection, utilization and management of wetland resources. Summary of the invention

[0003] Based on the above problems, the present application provides a method for generating a vegetation community classification map, which is used to generate a vegetation community classification map of a wetland with higher accuracy by using at least one of radar images or spectral images.

[0004] The embodiments of the present application disclose the following technical solutions:

[0005] The first aspect of the present application discloses a method for generating a vegetation community classification map, comprising:

[0006] Acquire multiple images of the target wetland; the multiple images are multiple radar images and / or multiple spectral images;

[0007] Extracting a feature image from each image of the target wetland to obtain a plurality of feature images;

[0008] A vegetation community classification map of the target wetland is generated based on the multiple feature images.

[0009] In an optional implementation, generating a vegetation community classification map of the target wetland according to the multiple feature images includes:

[0010] Selecting a preset number of images with the largest importance values ​​from the plurality of feature images as a plurality of target feature images; the importance values ​​indicate the importance of the feature images to generating a vegetation community classification map;

[0011] A vegetation community classification map of the target wetland is generated based on the multiple target feature images.

[0012] In an optional implementation, if the multiple images are the multiple radar images, extracting a feature image from each image of the target wetland to obtain multiple feature images includes:

[0013] An image of a dual-polarization water body index is extracted from each of the radar images as the multiple feature images.

[0014] In an optional implementation, if the multiple images are the multiple radar images, extracting a feature image from each image of the target wetland to obtain multiple feature images includes:

[0015] An image of vertical polarization index and an image of cross polarization index are extracted from each radar image as the multiple feature images.

[0016] In an optional implementation, if the multiple images are the multiple spectral images, extracting a feature image from each image of the target wetland to obtain multiple feature images includes:

[0017] An image set of water body vegetation index is extracted from each of the spectral images as the multiple feature images; the image set of water body vegetation index includes: an image of normalized water body index, an image of normalized vegetation index, an image of optimized soil adjusted vegetation index, an image of enhanced vegetation index and an image of difference vegetation index.

[0018] In an optional implementation, the multiple images are the multiple radar images and the multiple spectral images, and the extracting a feature image from each image of the target wetland to obtain multiple feature images includes:

[0019] An image set of water body vegetation index is extracted from each of the spectral images, and an image of dual-polarization water body index is extracted from each of the radar images as the multiple feature images; the image set of water body vegetation index includes: an image of normalized water body index, an image of normalized vegetation index, an image of optimized soil adjustment vegetation index, an image of enhanced vegetation index and an image of difference vegetation index.

[0020] In an optional implementation, after obtaining the vegetation community classification map of the target wetland, the method further includes: calculating the classification accuracy of the vegetation community classification map of the target wetland.

[0021] The second aspect of the present application discloses a device for generating a vegetation community classification map, comprising:

[0022] A multiple image acquisition module, used to acquire multiple images of the target wetland; the multiple images are multiple radar images and / or multiple spectral images;

[0023] A feature image acquisition module is used to extract a feature image from each image of the target wetland to obtain multiple feature images;

[0024] The vegetation community classification map generation module is used to generate a vegetation community classification map of the target wetland based on the multiple feature images.

[0025] A third aspect of the present application discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method introduced in any implementation manner of the first aspect.

[0026] A fourth aspect of the present application discloses an electronic device, comprising:

[0027] a memory having a computer program stored thereon;

[0028] A processor is used to execute the computer program in the memory to implement the steps of the method introduced in any implementation manner of the first aspect.

[0029] Compared with the prior art, this application has the following beneficial effects:

[0030] The present application discloses a method for generating a vegetation community classification map, comprising: acquiring multiple radar images and / or multiple spectral images of a target wetland as multiple images of the target wetland; extracting a feature map from each image of the target wetland to obtain multiple feature images; screening a preset number of images with the largest importance values ​​from the multiple feature images as multiple target feature images; and generating a vegetation community classification map of the target wetland based on the multiple target feature images.

[0031] Since radar data are less affected by weather, they can effectively reflect the structural characteristics of vegetation, such as vegetation height, canopy structure, and underground biomass (the total amount of organic matter in plant roots and rhizomes accumulated below the vegetation surface per unit area), etc., and can provide reliable observation information even in cloud cover or at night. Optical images have high spatial resolution and high temporal resolution, and can accurately capture the color and texture details of vegetation and slight changes in surface cover. Therefore, multiple images of the target wetland can well reflect the characteristics of the vegetation community in the target wetland. Based on the analysis of these data, a more accurate classification image of the vegetation community of the target wetland can be obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0033] Figure 1 A flow chart of a method for generating a vegetation community classification map provided in an embodiment of the present application;

[0034] Figure 2 A schematic diagram of importance distribution of a characteristic image of a radar polarization data set provided in an embodiment of the present application;

[0035] Figure 3 A schematic diagram of the importance distribution of a characteristic image of a dual polarization index group provided in an embodiment of the present application;

[0036] Figure 4 A schematic diagram of the importance distribution of a characteristic image of a water body vegetation index group provided in an embodiment of the present application;

[0037] Figure 5 A schematic diagram of the importance distribution of a characteristic image of an enhanced water vegetation index group provided in an embodiment of the present application;

[0038] Figure 6 A schematic diagram of the annual changes in the SDWI values ​​of different vegetation communities in a wetland provided in an embodiment of the present application;

[0039] Figure 7 A wetland vegetation community classification diagram provided in an embodiment of the present application;

[0040] Figure 8 A schematic diagram of another method for generating a vegetation community classification map provided in an embodiment of the present application;

[0041] Fig. 9 A schematic diagram of the structure of a device for generating a vegetation community classification map provided in an embodiment of the present application. DETAILED DESCRIPTION

[0042] Wetlands are ecological units that maintain the biodiversity of aquatic plants and terrestrial wildlife and are important ecosystems. Wetland ecosystems, composed of grass flats, mud flats and water bodies, provide habitats and foraging places for fish and winter migratory birds; different vegetation communities are essential for wintering water birds to choose suitable foraging and nesting environments.

[0043] Vegetation community classification maps are maps showing the geographical distribution of various vegetation communities and their distribution patterns. However, due to the influence of factors such as lake microtopography, water conditions, groundwater depth, soil structure, and especially the influence of water level changes, it is difficult to draw wetland vegetation community classification maps.

[0044] Although satellite remote sensing data provides a large amount of data sources for wetland mapping, the classification accuracy of vegetation community classification maps drawn using satellite remote sensing data is low due to the coarse spatial resolution of satellite images and the long revisit period. Therefore, how to draw a high-precision vegetation community classification map for wetlands has become an important issue that needs to be solved urgently.

[0045] To solve the above problems, the present application discloses a method for generating a vegetation community classification map, comprising: acquiring multiple radar images and / or multiple spectral images of a target wetland as multiple images of the target wetland; extracting a feature map from each image of the target wetland to obtain multiple feature images; screening out a preset number of images with the largest importance values ​​from the multiple feature images as multiple target feature images; and generating a vegetation community classification map of the target wetland based on the multiple target feature images.

[0046] Since radar data is less affected by weather, it can effectively reflect the structural characteristics of vegetation, such as vegetation height, canopy structure, and underground biomass, and can provide reliable observation information even in cloud cover or at night. Optical images have high spatial resolution and high temporal resolution, and can accurately capture the color and texture details of vegetation and slight changes in surface cover. Therefore, multiple images of the target wetland can well reflect the characteristics of the vegetation community in the target wetland. Based on the analysis of these data, a more accurate classification image of the vegetation community of the target wetland can be obtained.

[0047] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0048] Figure 1 A flow chart of a method for generating a vegetation community classification map provided in an embodiment of the present application. Figure 1 As shown, the method for generating a vegetation community classification map disclosed in the present application includes:

[0049] S101, acquiring multiple images of a target wetland.

[0050] The wetlands in this application are natural wetlands. Natural wetlands are naturally formed water areas in nature, including swamps, peatlands, lakes, rivers, beaches, etc. They have unique ecological functions and biodiversity and are an important part of the earth's ecosystem. The target wetland is the natural wetland for which the vegetation community classification map is to be generated in this application.

[0051] The multiple images of the target wetland are multiple radar images and / or multiple spectral images; that is, the multiple images may be multiple radar images, multiple spectral images, or multiple radar images and multiple spectral images.

[0052] It should be noted that the acquisition date of each radar image of the target wetland is not limited in this application, nor is the acquisition date of each spectral image of the target wetland. However, in order to better study the distribution of vegetation in the target wetland, those skilled in the art can collect a radar image of the target wetland in each month of the year, and collect a spectral image of the target wetland in each month of the year, and use the above images together as multiple images of the target wetland in this application.

[0053] Exemplarily, the target wetland in this application is Poyang Lake. The multiple images of the target wetland are a total of 12 radar images collected from January to December (one image per month), and a total of 8 spectral images collected from 8 months from January to December except February, March, April and May (one image per month), so the multiple images of the target wetland total 20 images.

[0054] It should be noted that the radar image of the target wetland collected in the present application is a dual-polarization C-band radar image, the data recurrence period of the radar image is 6 days, and the spatial resolution is 5m×20m.

[0055] It should be noted that the spectral image of the target wetland collected in this application is a high-resolution, multispectral satellite image obtained using a multispectral imager. It has a width of 290 kilometers and includes 13 spectral bands. The ground spatial resolution is 10m, 20m and 60m respectively. The data revisit period of the spectral image is 5 days.

[0056] S102, extracting a feature image from each image of the target wetland to obtain a plurality of feature images.

[0057] In the first step, each image of the target wetland is preprocessed to obtain multiple preprocessed images.

[0058] For each radar image of the target wetland, radiation calibration, noise removal, geometric correction, terrain correction and resampling can be performed on the radar image to obtain an image with a spatial resolution of 10m; then the image corresponding to the target wetland can be cropped from the image with a spatial resolution of 10m as the preprocessed radar image.

[0059] For each spectral image of the target wetland, atmospheric correction, radiation calibration, terrain correction and resampling can be performed on the spectral image to obtain an image with a spatial resolution of 10m; then the image corresponding to the target wetland can be cropped from the image with a spatial resolution of 10m as the preprocessed spectral image.

[0060] The second step is to extract classification feature variables.

[0061] The four feature groups constructed in this application are: radar polarization data group, dual polarization index group, water body vegetation index group and enhanced water body vegetation index group.

[0062] There are two kinds of feature maps included in the radar polarization data group, namely, the image of the vertical polarization index (S1_VV, VV polarization image) and the image of the cross polarization index (S2_VH, VH polarization image); the data needed to obtain the feature map in the radar polarization data group are: the preprocessed radar images in multiple images of the target wetland; the formulas needed to obtain the feature map in the radar polarization data group are shown in Table 1.

[0063] The characteristic graph included in the dual polarization index group is: an image of a dual polarization water index (an image of SDWI); the data used to obtain the characteristic graph in the dual polarization index group is: a preprocessed radar image from multiple images of the target wetland; the formula required to obtain the characteristic graph included in the dual polarization index group is shown in Table 1.

[0064] The characteristic images included in the water body vegetation index group are: images of normalized water body index (NDWI images), images of normalized vegetation index (NDVI images), images of optimized soil adjusted vegetation index (OSAVI images), images of enhanced vegetation index (EVI images) and images of difference vegetation index (DVI images); the data required to obtain the characteristic images included in the water body vegetation index group are: preprocessed spectral images from multiple images of the target wetland; the formulas required to obtain the characteristic images included in the water body vegetation index group are shown in Table 1.

[0065] The characteristic images included in the enhanced water vegetation index group are: images of normalized water index (NDWI images), images of normalized vegetation index (NDVI images), images of optimized soil adjusted vegetation index (OSAVI images), images of enhanced vegetation index (EVI images), images of difference vegetation index (DVI images) and images of dual polarization water index (SDWI images); the data required to obtain the characteristic images included in the water vegetation index group are: preprocessed spectral images and preprocessed radar images in multiple images of the target wetland; the formulas required to obtain the characteristic images included in the enhanced water vegetation index group are described in Table 1.

[0066] It should be noted that B3, B4, and B8 in Table 1 refer to optical remote sensing data in the green, red, and near-infrared bands, respectively; K SDWI It is the numerical value of the polarization water index, also known as the SDWI value. The VV and VH in the formula for calculating the SDWI value refer to the backscattering coefficient values ​​of the VV band and the VH band respectively.

[0067] Table 1 Schematic diagram of characteristic variable grouping

[0068]

[0069] A person skilled in the art may refer to the contents in Table 1 and the multiple images of the preprocessed target wetland obtained in the aforementioned embodiments to obtain multiple feature maps included in each feature grouping, that is, multiple feature maps included in the radar polarization data group, multiple feature maps included in the dual polarization index group, multiple feature maps included in the water body vegetation index group, and multiple feature maps included in the enhanced water body vegetation index group.

[0070] It can be understood that each preprocessed image is combined with a feature extraction formula, i.e., the calculation formula in Table 1, to obtain a feature map. As mentioned in the above embodiment, 12 radar images and 8 spectral images of the target wetland were obtained in this application. Therefore, the number of feature maps in the radar polarization data group is 24; the number of feature maps in the dual polarization index group is 12; the number of feature maps in the water body-vegetation index group is 40; the number of feature maps in the enhanced water body-vegetation index group is 48.

[0071] It is understandable that those skilled in the art may first extract multiple feature images from each preprocessed image of the target wetland, and then group all the extracted feature images to obtain four feature groups.

[0072] S103: Generate a vegetation community classification map of the target wetland based on the multiple feature images.

[0073] After obtaining multiple feature images included in each of the four feature groups, the following steps can be taken to generate a vegetation community classification map corresponding to each feature group, specifically:

[0074] In the first step, for each feature group, a preset number of images with the largest importance values ​​are selected from the multiple feature images of the feature group as multiple target feature images, wherein the importance value indicates the importance of the feature image to generating the vegetation community classification map.

[0075] Specifically, the random forest relative importance algorithm can be used to perform importance analysis on the feature images in the radar polarization data group, dual polarization index group, water vegetation index group and enhanced water vegetation index group in turn, that is, to calculate the importance value of each feature image in each feature group and display it in the form of a chart.

[0076] Figure 2 A schematic diagram of the importance distribution of a characteristic image of a radar polarization data set provided in an embodiment of the present application. Figure 2The importance values ​​of each feature image in the radar polarization data set calculated based on the random forest relative importance algorithm are arranged in descending order. The horizontal axis represents the name of each feature image (abbreviated as index-month), and the vertical axis represents the importance value of each feature image (abbreviated as importance value). Figure 2 As shown in Figure 2, the VV polarimetric image with the highest importance value is the one in May, namely VV_5.

[0077] Figure 3 A schematic diagram of the importance distribution of a feature image of a dual-polarization index group provided in an embodiment of the present application. Figure 3 The importance value of each feature image in the dual polarization index group calculated based on the random forest relative importance algorithm is arranged in descending order. The horizontal axis represents the name of each feature image (abbreviated as index-month), and the vertical axis represents the importance value of each feature image (abbreviated as importance value). Figure 3 As shown in Figure 2, the image with the highest importance value is the SDWI image in March, namely SDWI_3.

[0078] Figure 4 A schematic diagram of the importance distribution of a feature image of a water body vegetation index group provided in an embodiment of the present application. Figure 4 The importance value of each feature image in the water vegetation index group calculated based on the random forest relative importance algorithm is arranged in descending order. The horizontal axis represents the name of each feature image (abbreviated as index-month), and the vertical axis represents the importance value of each feature image (abbreviated as importance value). Figure 4 As shown in FIG. 1 , the image with the highest importance value is the DVI image of October, namely DVI_10.

[0079] Figure 5 A schematic diagram of the importance distribution of a feature image of an enhanced water vegetation index group provided in an embodiment of the present application. Figure 5 The importance value of each feature image in the enhanced water vegetation index group calculated based on the random forest relative importance algorithm is arranged in descending order. The horizontal axis represents the name of each feature image (abbreviated as index-month), and the vertical axis represents the importance of each feature image (abbreviated as importance value). Figure 5 As shown in FIG. 1 , the image with the highest importance value is the DVI image of October, namely DVI_10.

[0080] After obtaining the feature images of each feature group, a preset number of images with the largest importance values ​​can be screened out as multiple target feature images. The specific value of the preset number is not limited in this application, and those skilled in the art can set the preset number according to actual needs.

[0081] It can be understood that the purpose of selecting multiple target feature images is to generate a vegetation community classification map with the highest possible accuracy based on the least possible data.

[0082] Figure 6 A schematic diagram of the annual change of the SDWI values ​​of different vegetation communities in a wetland provided in an embodiment of the present application. Using the backscatter coefficient values ​​of the VV band and the VH band of the radar image of the target wetland preprocessed in the present application, and the formula K in Table 2 SDWI =ln(10×VV×VH)-8, the SDWI value of each vegetation community in the target wetland can be calculated.

[0083] Combination Figure 6 As shown in the figure, the SDWI values ​​of the two types of vegetation in the water and mudflats in December are between -18 and -20; the SDWI values ​​of Zizania latifolia, Carex phalaenopsis and Polygonum aviculare are slightly higher, and the SDWI value of Phragmites australis is the highest; Zizania latifolia can be further separated using the data from March. The differences between water and mudflats, as well as Carex phalaenopsis and Polygonum aviculare, become more obvious in July. Figure 6 The data from March, July and December provide valuable information for distinguishing different wetland vegetation types.

[0084] In an optional implementation, to simplify the calculation process, the random forest relative importance algorithm can be used as the model to be trained, and the trained model is trained with the training data to obtain a trained target model. In subsequent work, all feature images of each feature group are input into the target model, and the importance value of each feature image is directly input.

[0085] The training data for the target model can be obtained by the following methods. Specifically, based on field sampling, reference samples of different vegetation community types are selected; based on time series multispectral data (i.e., spectral images collected in different months), the annual NDVI index sequence variation characteristics of different vegetation communities are obtained; according to different NDVI time series characteristics, 300 reference samples are selected using the known NDVI index data set; 70% of them are used as training samples, and 30% of them are used as verification samples. The sample selection of different types of land objects is shown in Table 2.

[0086] Table 2

[0087] Feature Type Training samples Verification Sample Water 30 10 Mudflats 35 15 Phragmites australis 35 15 mushroom 35 15 Carex phalaenopsis 40 20 Polygonum and others 35 15 overall 210 90

[0088] It should be noted that this application only schematically provides a possible implementation method for obtaining training data. Those skilled in the art can use other data as training data to train the training model to obtain the target model.

[0089] In the second step, for each feature grouping, a vegetation community classification map of the target wetland is generated based on multiple target feature images of the feature grouping.

[0090] Specifically, methods such as support vector machines, random forest algorithms, convolutional neural networks, and deep neural networks can be used to process multiple target feature images in each feature group to directly generate a vegetation community classification map of the target wetland.

[0091] It is understandable that if there are several feature groupings, several vegetation community classification maps will be generated.

[0092] Exemplarily, in the present application, a random forest algorithm is used to process multiple target feature images to obtain a vegetation community classification map of the target wetland.

[0093] Figure 7 A wetland vegetation community classification diagram provided in an embodiment of the present application. Figure 7 The image in the upper left corner is a vegetation community classification map generated based on multiple target feature images selected from the radar polarimetric data set; Figure 7 The image in the upper right corner is a vegetation community classification map generated based on multiple target feature images selected from the dual polarization index group; Figure 7 The image in the lower left corner is a vegetation community classification map generated based on multiple target feature images selected from the water body vegetation index group; Figure 7 The image in the lower right corner is a vegetation community classification map generated based on multiple target feature images selected from the enhanced water vegetation index group.

[0094] In an optional implementation, after obtaining the vegetation community classification map corresponding to each feature grouping, the overall classification accuracy and Kappa coefficient of each vegetation community classification map can be calculated, and then the classification accuracy of the vegetation community classification map obtained using different feature grouping data can be evaluated; thereby selecting the scheme with the highest classification accuracy from the four schemes for generating vegetation community classification maps.

[0095] Among them, the overall classification accuracy (OA) refers to the ratio of the total number of correctly classified pixels to the total number of all pixels to be classified. The calculation formula of the overall classification accuracy is as follows:

[0096]

[0097] In this formula, r represents the number of categories, X jj represents the value on the diagonal of the confusion matrix (i.e., the number of samples correctly classified into the jth class), and N represents the total number of test samples.

[0098] Among them, the Kappa coefficient is an indicator for consistency testing and can also be used to measure the effect of classification. Because for classification problems, the so-called consistency refers to whether the model prediction results are consistent with the actual classification results. The calculation of the Kappa coefficient is based on the confusion matrix, and the value is between -1 and 1, but the Kappa coefficient is usually greater than 0. The calculation formula of the Kappa coefficient based on the confusion matrix is ​​as follows:

[0099]

[0100] The overall classification accuracy and Kappa coefficient Figure 7 The classification accuracy of the four images is evaluated and the results in Table 3 are obtained.

[0101] Table 3 Accuracy comparison results

[0102]

[0103] Combined with Table 3, the following conclusions can be drawn: the classification accuracy of the vegetation community classification map obtained by using multiple target feature images of the enhanced water vegetation index group is the highest; the classification accuracy of the vegetation community classification map obtained by using multiple target feature images of the water vegetation index group is the second highest; the classification accuracy of the vegetation community classification map obtained by using multiple target feature images of the dual-polarization water index group is the third highest; the classification accuracy of the vegetation community classification map obtained by using multiple target feature images of the radar polarization data group is the lowest.

[0104] It can be seen that using a mixed image of the spectral image and radar image of the target wetland can well classify various types of vegetation communities in the target wetland, and can generate a vegetation community classification map with high classification accuracy.

[0105] Furthermore, the present application also provides a confusion matrix of the enhanced water vegetation index classification results, as shown in Table 4.

[0106] Table 4 Confusion matrix of enhanced water vegetation index classification results

[0107]

[0108]

[0109] Table 4 shows the accuracy evaluation results of the enhanced water vegetation index classification results. Among them, water and mudflats have the highest classification accuracy, while southern reed and sedge have the lowest classification accuracy. Among them, the values ​​of the normalized vegetation index of Zizania latifolia, southern reed, Polygonum aviculare and other species are relatively similar, and the spectral characteristics of the values ​​of the normalized vegetation index are all M-shaped, making Zizania latifolia, Polygonum aviculare and other species more difficult to distinguish in vegetation identification. Therefore, Zizania latifolia and Polygonum aviculare and other species have lower classification accuracy.

[0110] In summary, the present application discloses a method for generating a vegetation community classification map. Figure 8 As shown, the method for generating a vegetation community classification map disclosed in the present application is to obtain multiple radar images and multiple spectral images of a target wetland; perform preprocessing operations on the obtained multiple images, for example, perform preprocessing operations such as noise removal, radiation calibration, geometric correction, terrain correction and cropping on each radar image, and perform preprocessing operations such as atmospheric correction, resampling, band synthesis and cropping on each spectral image to obtain multiple images of the preprocessed target wetland; extract multiple features from the multiple images of the preprocessed target wetland; divide the multiple features into four groups, namely, radar polarization data group, polarization water index group, water vegetation index group and enhanced water vegetation index group; optimize and extract the feature map of each feature group, namely, select a preset number of target feature images with the largest importance value from the multiple feature maps of each feature group (namely, feature optimization); use a random forest algorithm to process the multiple target feature maps of each feature group to produce a community classification map corresponding to each feature group.

[0111] It is understandable that when only multiple radar images of the target wetland are obtained or only multiple spectral images of the target wetland are obtained, the method disclosed in the above embodiment can also be used to generate a vegetation classification map of the target wetland.

[0112] Since most of the world's wetlands are located in cloudy and rainy areas, this invention proposes a vegetation classification method that combines feature optimization and machine learning, fully leveraging the advantages of radar data and coordinating with optical data to perform high-precision classification of wetland vegetation, which has great application value.

[0113] Since different plant communities in the wetland area have differences in vegetation structure and growth cycle, the spectral images and radar images of the target wetland collected in different months of a year are selected in this application as multiple images of the target wetland. The advantage of this is that it can reflect the characteristics of the vegetation community at different times.

[0114] Based on the radar data of the target wetland collected in different months of a year, the present application constructs the dual-polarization water index time series data of the target wetland, that is, constructs the dual-polarization index of the target wetland including multiple feature images obtained at different collection times; at the same time, based on the spectral data of the target wetland collected in different months of a year, the spectral vegetation index time series data of the target wetland is constructed, that is, constructs the enhanced water vegetation index group of the target wetland including multiple feature images obtained at different collection times.

[0115] Compared with the vegetation community classification map generated by directly using the characteristic image in the radar polarization data set, the classification accuracy of the vegetation community classification map generated by combining the dual-polarization water index with the water vegetation index obtained by multispectral remote sensing data has been greatly improved. The accuracy of the vegetation community classification map generated by combining the dual-polarization water index with the water vegetation index obtained by multispectral remote sensing data can reach more than 80%, and the result is more accurate, which is an effective mapping scheme for the classification map of wetland vegetation communities.

[0116] Based on the method for generating a vegetation community classification map disclosed in the aforementioned embodiment, the present application also discloses a device for generating a vegetation community classification map. Fig. 9 A schematic diagram of the structure of a device for generating a vegetation community classification map provided in an embodiment of the present application. Fig. 9 As shown, the device 900 for generating a vegetation community classification map disclosed in the present application includes:

[0117] A multiple image acquisition module 901 is used to acquire multiple images of a target wetland; the multiple images are multiple radar images and / or multiple spectral images;

[0118] A feature image acquisition module 902 is used to extract a feature image from each image of the target wetland to obtain multiple feature images;

[0119] The vegetation community classification map generating module 903 is used to generate a vegetation community classification map of the target wetland according to the multiple feature images.

[0120] In an optional implementation, the vegetation community classification map generation module 903 includes:

[0121] A target feature image screening unit is used to screen out a preset number of images with the largest importance values ​​from the multiple feature images as multiple target feature images; the importance values ​​indicate the importance of the feature images to generating a vegetation community classification map;

[0122] The vegetation community classification map generating unit is used to generate a vegetation community classification map of the target wetland according to the multiple target feature images.

[0123] In an optional implementation, the feature image acquisition module 902 includes: a first feature image acquisition unit, configured to extract an image of a dual-polarization water body index from each of the radar images as the plurality of feature images.

[0124] In an optional implementation, the feature image acquisition module 902 includes: a second feature image acquisition unit, configured to extract an image of a vertical polarization index and an image of a cross polarization index from each of the radar images as the multiple feature images.

[0125] In an optional implementation, the feature image acquisition module 902 includes: a third feature image acquisition unit, used to extract an image set of water body vegetation index from each of the spectral images as the multiple feature images; the image set of water body vegetation index includes: an image of normalized water body index, an image of normalized vegetation index, an image of optimized soil adjustment vegetation index, an image of enhanced vegetation index and an image of difference vegetation index.

[0126] In an optional implementation, the feature image acquisition module 902 includes:

[0127] The fourth feature image acquisition unit is used to extract an image set of water body vegetation index from each of the spectral images, and to extract an image of dual-polarization water body index from each of the radar images as the multiple feature images; the image set of water body vegetation index includes: an image of normalized water body index, an image of normalized vegetation index, an image of optimized soil adjustment vegetation index, an image of enhanced vegetation index and an image of difference vegetation index.

[0128] In an optional implementation, the vegetation community classification map generation device 900 further includes:

[0129] The accuracy determination module is used to calculate the classification accuracy of the vegetation community classification map of the target wetland.

[0130] Based on the method and device for generating a vegetation community classification map provided in the aforementioned embodiments, the present application also provides a computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements some or all of the steps in the method for generating a vegetation community classification map mentioned above.

[0131] Based on the method and device for generating a vegetation community classification map provided in the above-mentioned embodiment, the present application further provides an electronic device, including:

[0132] A memory having a computer program stored thereon; a processor configured to execute the computer program in the memory to implement part or all of the steps in the method for generating a vegetation community classification map provided in the aforementioned embodiment.

[0133] It should be noted that each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device embodiment described above is merely schematic, in which the unit described as a separate component may or may not be physically separated, and the component prompted as a unit may or may not be a physical unit, that is, it may be located in one place, or it may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative work.

[0134] The above is only a specific implementation of the present application. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed in the present application should be included in the protection scope of the present application.

Claims

1. A method for generating a vegetation community classification map, characterized in that: The method comprises: Acquire multiple images of the target wetland; the multiple images are multiple radar images and / or multiple spectral images; Extracting a feature image from each image of the target wetland to obtain a plurality of feature images; A vegetation community classification map of the target wetland is generated based on the multiple feature images.

2. The method according to claim 1, characterized in that Generating a vegetation community classification map of the target wetland according to the plurality of feature images includes: Selecting a preset number of images with the largest importance values ​​from the plurality of feature images as a plurality of target feature images; the importance values ​​indicate the importance of the feature images to generating a vegetation community classification map; A vegetation community classification map of the target wetland is generated based on the multiple target feature images.

3. The method according to claim 1, characterized in that If the multiple images are the multiple radar images, extracting a feature image from each image of the target wetland to obtain multiple feature images includes: An image of a dual-polarization water body index is extracted from each of the radar images as the multiple feature images.

4. The method according to claim 1, characterized in that: If the multiple images are the multiple radar images, extracting a feature image from each image of the target wetland to obtain multiple feature images includes: An image of vertical polarization index and an image of cross polarization index are extracted from each radar image as the multiple feature images.

5. The method according to claim 1, characterized in that If the multiple images are the multiple spectral images, extracting a feature image from each image of the target wetland to obtain multiple feature images includes: An image set of water body vegetation index is extracted from each of the spectral images as the multiple feature images; the image set of water body vegetation index includes: an image of normalized water body index, an image of normalized vegetation index, an image of optimized soil adjusted vegetation index, an image of enhanced vegetation index and an image of difference vegetation index.

6. The method according to claim 1, characterized in that The multiple images are the multiple radar images and the multiple spectral images, and the extracting a feature image from each image of the target wetland to obtain multiple feature images includes: An image set of water body vegetation index is extracted from each of the spectral images, and an image of dual-polarization water body index is extracted from each of the radar images as the multiple feature images; the image set of water body vegetation index includes: an image of normalized water body index, an image of normalized vegetation index, an image of optimized soil adjustment vegetation index, an image of enhanced vegetation index and an image of difference vegetation index.

7. The method according to any one of claims 1 to 6, characterized in that After obtaining the vegetation community classification map of the target wetland, the method further includes: The classification accuracy of the vegetation community classification map of the target wetland is calculated.

8. A device for generating a vegetation community classification map, characterized in that: The device comprises: A multiple image acquisition module, used to acquire multiple images of the target wetland; the multiple images are multiple radar images and / or multiple spectral images; A feature image acquisition module is used to extract a feature image from each image of the target wetland to obtain multiple feature images; The vegetation community classification map generation module is used to generate a vegetation community classification map of the target wetland based on the multiple feature images.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method described in any one of claims 1 to 7 are implemented.

10. An electronic device, characterized in that: include: a memory having a computer program stored thereon; A processor, configured to execute the computer program in the memory to implement the steps of the method according to any one of claims 1 to 7.

Images