Salt marsh vegetation classification method, system, equipment, medium and product

Abstract

The invention discloses a salt marsh vegetation classification method, system and device, a medium and a product, and relates to the field of salt marsh vegetation classification, and the method comprises the steps: generating a vegetation index time sequence according to satellite remote sensing data; dividing the coastal zone latitude into a plurality of classification latitude zones as classification intervals, and calculating the monthly mean value and variance index of each vegetation index in each interval and the annual mean value index and variance index in the current year; based on the normalized difference vegetation index time sequence and the enhanced vegetation index time sequence, calculating growth period and decline period indexes of vegetation by using a double logistic function; calculating the J-M distance of each candidate classification index on different vegetation classification samples; on the basis of J-M distances of the indexes on different vegetation classification samples and preliminary classification contribution degree evaluation, the classification index which can be most indexed is screened out, a final random forest classification model is constructed, salt marsh vegetation in the coastal zone salt marsh wetland is classified, and the coastal plant classification task can be efficiently completed with low cost.

Description

Technical Field

[0001] The present application relates to the field of salt marsh vegetation classification, and in particular to a salt marsh vegetation classification method, system, equipment, medium and product. Background Art

[0002] Coastal salt marsh wetlands are located in the land-sea transition zone. They mainly refer to saltwater or brackish muddy flats covered with halophytes along the coast that are periodically or intermittently affected by ocean tides. Together with mangroves and seagrass beds, they are called "blue carbon" ecosystems and have strong carbon fixation capacity. Accurate identification of coastal salt marsh vegetation species and spatial distribution is an important prerequisite for accurately assessing coastal blue carbon. However, current methods for classifying coastal salt marsh plants often have multiple defects. Limited by the complex environment of coastal salt marsh wetlands, inefficient manual field surveys are often insufficient for large-scale salt marsh vegetation surveys on China's vast coastal belt. The lack of samples has led to the fact that most encapsulated general classification algorithms do not include the identification of specific types of coastal salt marsh vegetation. At present, most large-scale classification algorithms for salt marsh vegetation are often based on threshold classification methods constructed based on expert experience. This algorithm, which relies heavily on empirical judgment, not only has poor scalability in long-term classification, but also has the problem of insufficient accuracy in large-scale spatial classification. Although some studies have attempted to use high-resolution, hyperspectral and radar data for classification, the availability and processing difficulty of this type of data are far less than that of mainstream remote sensing satellite data such as Landsat and Sentinel, making it difficult to perform efficient and low-cost classification on a large scale. Summary of the invention

[0003] The purpose of this application is to provide a salt marsh vegetation classification method, system, equipment, medium and product to solve the problem of difficult to classify efficiently and at low cost on a large scale.

[0004] To achieve the above objectives, this application provides the following solutions:

[0005] In a first aspect, the present application provides a salt marsh vegetation classification method, comprising:

[0006] Generate vegetation index time series based on satellite remote sensing data; vegetation index time series includes normalized difference vegetation index time series, enhanced vegetation index time series, normalized difference water index time series and corrected red edge chlorophyll vegetation index time series;

[0007] The coastal zone latitude is divided into several classification latitude bands as classification intervals, and the monthly mean index and monthly variance index of each vegetation index in each classification interval are calculated, as well as the annual mean index and annual variance index of each vegetation index in the current year;

[0008] Based on the normalized difference vegetation index time series and the enhanced vegetation index time series, the growth period and decay period of vegetation are calculated using a bilogistic function as candidate classification indicators;

[0009] Combine different vegetation classification samples to calculate the JM distance of each candidate classification index on different vegetation classification samples; each candidate index includes monthly mean index, monthly variance index, annual mean index, annual variance index, growth period index and decay period index;

[0010] Based on the JM distance of each candidate indicator on different vegetation classification samples and the preliminary classification contribution evaluation, the most distinguishable classification indicator is selected to construct the final random forest classification model;

[0011] The salt marsh vegetation in the coastal salt marsh wetlands was classified according to the final random forest classification model.

[0012] Optionally, based on the JM distance and the above indicators, the most separable window vegetation index is selected to construct a random forest model, including:

[0013] The indicators ranked in the top 50% of the JM distance ranking were extracted as preliminary classification variables for the random forest model;

[0014] Calculate the contribution of each indicator to the random forest classification model and rank the contributions;

[0015] The top 10 indicators with the highest contribution were used as the most distinguishable vegetation index for the window period;

[0016] The most separable window vegetation index is used as the independent variable of the final random forest classification model to construct the final random forest classification model.

[0017] Optionally, generating a vegetation index time series according to satellite remote sensing data, the method further includes: the satellite remote sensing data includes Landsat-8 image data and Sentinel-2 image data;

[0018] Using the QA band of Landsat-8 to remove pixels covered by clouds and shadows in the Landsat-8 image data, to generate Landsat-8 image data after removal;

[0019] The sentinel-2 cloud detector library and the S2_CLOUD_PROBABILITY dataset are used to remove pixels covered by clouds and shadows in the Sentinel-2 image data to generate the removed Sentinel-2 image data;

[0020] Taking the removed Sentinel-2 image data as a benchmark, the difference between the removed Landsat-8 image data and the removed Sentinel-2 image data is corrected by using the least square method to determine the optimal satellite remote sensing data.

[0021] Optionally, the pixels covered by clouds and shadows in the Landsat-8 image data are removed by using the QA band of Landsat-8 to generate Landsat-8 image data after removal, specifically including:

[0022] The QA band of Landsat-8 is used to determine the pixel occlusion information covered by clouds and shadows in the Landsat-8 image data in a decimal manner;

[0023] Convert the pixel occlusion information into binary, and determine the pixel quality based on the number of bits corresponding to the binary pixel occlusion information;

[0024] Based on the pixel quality, a mask function is used to remove pixels covered by clouds and shadows in the Landsat-8 image data to generate Landsat-8 image data after removal.

[0025] Optionally, using the sentinel-2 cloud detector library and the S2_CLOUD_PROBABILITY dataset to remove pixels covered by clouds and shadows in the Sentinel-2 image data, and generating the removed Sentinel-2 image data, specifically includes:

[0026] Based on the sentinel-2 cloud detector library, determining the pixel coverage probability of being covered by clouds and shadows in the Sentinel-2 image data according to the S2_CLOUD_PROBABILITY dataset;

[0027] Based on the pixel coverage probability, a filter function is used to remove pixels covered by clouds and shadows in the Sentinel-2 image data to generate Sentinel-2 image data after removal.

[0028] Optionally, generate vegetation index time series based on satellite remote sensing data, including:

[0029] The Savitzky-Golay filtering algorithm is used to smooth the salt and pepper noise and extreme values ​​appearing in the satellite remote sensing data collected within a set period, and the Sentinel-2 image data in the satellite remote sensing data is resampled using cubic convolution, so that the spatial resolution of the resampled Sentinel-2 data is consistent with that of the Landsat-8 image data in the satellite remote sensing data, thereby generating a vegetation index time series; the vegetation index time series is displayed as an annual characteristic distribution map of the vegetation index.

[0030] In a second aspect, a salt marsh vegetation classification system includes:

[0031] Vegetation index time series generation module, used to generate vegetation index time series based on satellite remote sensing data; vegetation index time series includes normalized difference vegetation index time series, enhanced vegetation index time series, normalized difference water index time series and corrected red edge chlorophyll vegetation index time series;

[0032] The monthly index and annual index calculation module is used to divide the coastal zone latitude into several classification latitude bands as classification intervals, and calculate the monthly mean index and monthly variance index of each vegetation index in each classification interval, as well as the annual mean index and annual variance index of each vegetation index in the year;

[0033] A growth period candidate index and decay period candidate index calculation module, used to calculate the growth period and decay period of vegetation as candidate classification indicators using a bilogistic function based on the normalized difference vegetation index time series and the enhanced vegetation index time series;

[0034] The JM distance calculation module is used to combine different vegetation classification samples to calculate the JM distance of each candidate classification index on different vegetation classification samples; each candidate index includes a monthly mean index, a monthly variance index, an annual mean index, an annual variance index, a growth period index, and a decay period index;

[0035] Based on the JM distance of each candidate indicator on different vegetation classification samples and the preliminary classification contribution evaluation, the most distinguishable classification indicator is selected to construct the final random forest classification model;

[0036] A classification module is used to classify salt marsh vegetation in coastal salt marsh wetlands according to the final random forest classification model.

[0037] In a third aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of any of the above-described salt marsh vegetation classification methods.

[0038] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described salt marsh vegetation classification methods.

[0039] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of any of the above-mentioned salt marsh vegetation classification methods.

[0040] According to the specific embodiments provided by the present application, the present application discloses the following technical effects: the present application uses satellite remote sensing data to generate a vegetation index time series, and secondly, synthesizes and smoothes the monthly indicators and annual indicators of multiple vegetation indices, and uses a dual logistic function to calculate the vegetation growth period indicator and decay period indicator, and then, combines different vegetation classification samples, selects the most separable window period vegetation index according to the JM distance, and constructs a random forest model, and finally, uses the constructed random forest model to classify the salt marsh vegetation in the coastal salt marsh wetlands, thereby completing the coastal plant classification task at low cost and high efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] 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 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 work.

[0042] Figure 1 This is a flow chart of a salt marsh vegetation classification method in one embodiment of the present application;

[0043] Figure 2 An annual characteristic distribution map of key plant indexes of different vegetation types provided in an embodiment of the present application; Figure 2 (a) is the NDVI sequence diagram. Figure 2 (b) in the figure is the EVI sequence diagram. Figure 2 (c) in the figure is the NDWI sequence diagram; Figure 2 (d) in the figure is the MRECI sequence diagram;

[0044] Figure 3 A technical diagram of a salt marsh vegetation classification method in an embodiment of the present application;

[0045] Figure 4 A map of the distribution and succession characteristics of Chinese salt marsh wetland vegetation provided in an embodiment of the present application; Figure 4 (a) is the latitudinal distribution map of the average change in salt marsh land cover in the past six years. Figure 4(b) is the latitudinal distribution map of the change in the surface cover of salt marsh wetlands in the past six years. Figure 4 (c) is a schematic diagram of the Liaohe River Estuary. Figure 4 (d) is a schematic diagram of the Yellow River Estuary. Figure 4 (e) is a schematic diagram of the Yangtze River Estuary. Figure 4 (f) in the figure is a schematic diagram of the Pearl River Estuary. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions 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.

[0047] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0048] In an exemplary embodiment, Figure 1 As shown, a salt marsh vegetation classification method is provided, comprising:

[0049] Step 101: Generate a vegetation index time series based on satellite remote sensing data; the vegetation index time series includes a normalized difference vegetation index time series, an enhanced vegetation index time series, a normalized difference water index time series, and a corrected red-edge chlorophyll vegetation index time series.

[0050] Step 102: Divide the coastal zone latitude into several classification latitude bands as classification intervals, and calculate the monthly mean index and monthly variance index of each vegetation index in each classification interval, as well as the annual mean index and annual variance index of each vegetation index in the current year.

[0051] Step 103: Based on the normalized difference vegetation index time series and the enhanced vegetation index time series, the growth period and decay period of vegetation are calculated using a bilogistic function as candidate classification indicators.

[0052] Step 104: combining different vegetation classification samples, and calculating the JM distance of each candidate classification index on different vegetation classification samples.

[0053] Step 105: Based on the JM distance of each candidate indicator on different vegetation classification samples and the preliminary classification contribution evaluation, the most distinguishable classification indicator is selected to construct the final random forest classification model.

[0054] Step 106: classify the salt marsh vegetation in the coastal salt marsh wetland according to the final random forest classification model.

[0055] In another exemplary embodiment of the present application, in order to accurately construct a random forest model to achieve the effect of classifying salt marsh vegetation in different coastal salt marsh wetlands, the above step 105 is replaced by the following steps 201 to 204:

[0056] The step 105 specifically includes:

[0057] Step 201: extract the indicators whose JM distance ranking is in the top 50% as the preliminary classification variables of the random forest model.

[0058] Step 202: Calculate the contribution of each indicator to the random forest classification model and rank the contributions.

[0059] Step 203: The top 10 contribution indexes are used as the most distinguishable window vegetation indexes.

[0060] Step 204: taking the most separable window vegetation index as an independent variable of a final random forest classification model to construct a final random forest classification model.

[0061] In another exemplary embodiment of the present application, in order to ensure the consistency of the input satellite remote sensing data, the input satellite remote sensing data may be preliminarily processed. Before the above step 101, the method may further include the following steps 301 to 303. Among them:

[0062] Before step 101, the method further includes: the satellite remote sensing data includes Landsat-8 image data and Sentinel-2 image data.

[0063] Step 301: using the QA band of Landsat-8 to remove pixels covered by clouds and shadows in the Landsat-8 image data, and generating Landsat-8 image data after removal.

[0064] Step 302: Using the sentinel-2 cloud detector library and the S2_CLOUD_PROBABILITY dataset, pixels covered by clouds and shadows in the Sentinel-2 image data are removed to generate Sentinel-2 image data after removal.

[0065] Step 303: Taking the removed Sentinel-2 image data as a reference, the difference between the removed Landsat-8 image data and the removed Sentinel-2 image data is corrected by using the least square method to determine the optimal satellite remote sensing data.

[0066] Among them, step 301 specifically includes: using the QA band of Landsat-8 to determine the pixel occlusion information covered by clouds and shadows in the Landsat-8 image data in a decimal manner; converting the pixel occlusion information into binary, and determining the pixel quality based on the number of bits corresponding to the binary pixel occlusion information; based on the pixel quality, using a mask function to remove the pixels covered by clouds and shadows in the Landsat-8 image data to generate the removed Landsat-8 image data.

[0067] Step 302 specifically includes: determining, based on the sentinel-2 cloud detector library and according to the S2_CLOUD_PROBABILITY data set, the pixel coverage probability of being covered by clouds and shadows in the Sentinel-2 image data; based on the pixel coverage probability, using a filter function to remove pixels covered by clouds and shadows in the Sentinel-2 image data to generate the removed Sentinel-2 image data.

[0068] In another exemplary embodiment of the present application, for salt marsh plants, after literature research and local pre-classification results, the following four indicators are selected as key parameters that can reflect the characteristics of surface vegetation. The above step 101 includes the following steps.

[0069] Step 101 specifically includes: using a Savitzky-Golay filtering algorithm to smooth the salt and pepper noise and extreme values ​​appearing in the satellite remote sensing data collected within a set period, and using cubic convolution to resample the Sentinel-2 image data in the satellite remote sensing data so that the spatial resolution of the resampled Sentinel-2 data is consistent with that of the Landsat-8 image data in the satellite remote sensing data, and generating a vegetation index time series; the vegetation index time series is displayed as an annual characteristic distribution map of the vegetation index.

[0070] The present application also provides an application scenario, which applies the above-mentioned salt marsh vegetation classification method. Specifically: Figure 3As shown in the figure, all Landsat-8 Collection2Level-2 OLI (TOA) data and Sentinel-2 Level-2A-MSI (TOA) data stored in the remote sensing cloud computing platform are taken as an example. Among them, the Landsat-8 series data includes all available Landsat 8 Operational Land Imager (OLI) images from the United States Geological Survey, with a spatial resolution of 30m and a ground revisit time of 8 days; Sentinel-2 data includes all available Sentinel-2A and Sentinel-2B multispectral imager (MSI) images from the European Space Agency, with a spatial resolution of 10m and a ground revisit time of 5 days.

[0071] Secondly, the satellite remote sensing data are preprocessed: Since the remote sensing datasets in the remote sensing cloud computing platform have been atmospherically corrected / terrain corrected, the QA band of Landsat-8, the sentinel2 cloud detector library and the S2_CLOUD_PROBABILITY dataset are used to remove pixels covered by clouds and shadows in the Landsat-8 image data and Sentinel-2 image data.

[0072] Among them, the Landsat-QA band provides information in decimal form whether the pixel is blocked by clouds or fog. After converting it to binary, the pixel quality is determined according to the corresponding number of bits, and the corresponding removal is performed using the mask function provided by the remote sensing cloud computing platform. S2_CLOUD_PROBABILITY directly provides the possibility of whether the image pixels in Sentinel-2 are covered by clouds, so that the filter function can be used with a threshold of 80% to remove pixels that are very likely to be contaminated by clouds. At the same time, in order to eliminate the differences in sensor bands between different satellites, the Sentinel-2 satellite with a large number of images is used as a benchmark, and the differences between different sensor data are corrected by the least squares method, and finally the optimal remote sensing data image collection (i.e., the optimal satellite remote sensing data) is obtained.

[0073] Furthermore, the vegetation index time series was generated based on satellite remote sensing data: for salt marsh plants, after literature research and local pre-classification results, the following four indicators were selected as key parameters that can reflect the characteristics of surface vegetation, including:

[0074] 1) Normalized Differential Vegetation Index (NDVI): It is calculated by the difference in vegetation absorption in the near-infrared and infrared bands, and can reflect the health and growth of vegetation.

[0075] 2) Enhanced Vegetation Index (EVI): The NDVI index corrects the atmospheric and soil noise and can well reflect the dense plant growth conditions.

[0076] 3) Normalized Difference Water Index (NDWI): It can reflect the surface soil moisture and vegetation coverage through the difference in water absorption in the green light and near-infrared bands.

[0077] 4) Modified Red Edge Chlorophyll Index (MRECI): The chlorophyll content in leaves nourished by nitrogen responds and can indicate the photosynthetic activity of the canopy.

[0078] The calculation of the above indices is shown in equations (1), (2), (3) and (4) respectively. The results are obtained by resampling various vegetation indices through cubic convolution to make the resolution of each image uniform. In this way, four indices are obtained for each pixel in a single image.

[0079]

[0080] Among them, NIR is the near infrared band (845-885nm); RED is the red band (630-688nm); BLUE is the blue band (450-515nm); GREEN is the green band (530-590nm).

[0081] Due to the cloud masking, there are often pixel missing problems in a single remote sensing image, accompanied by natural salt and pepper noise. To address this, with a 15-day period, all remote sensing images within the range are first aggregated to fill the pixels lost due to the cloud, and then the Savitzky-Golay filter algorithm that can smooth the data and maintain the original trend of change is used to smooth the salt and pepper noise and extreme values, and the Sentinel-2 data with higher resolution is resampled using cubic convolution to ensure that it is consistent with the spatial resolution of the Landsat remote sensing image. Finally, a high-quality annual vegetation index annual characteristic distribution map with a temporal resolution of 15 days and a spatial resolution of 30m is obtained. The annual characteristic distribution map of the vegetation index is the vegetation index time series.

[0082] In 2022, detailed field surveys and corresponding classified sample collection were completed in multiple areas with widespread salt marsh vegetation along the coast of China. Taking the vegetation at the Yangtze River Estuary as an example, by combining the classified sample data of field surveys, this application draws the annual characteristic distribution map of key plant indexes in different regions and different vegetation, such as Figure 2 shown.

[0083] Due to the large latitudinal span of China's coastal zone, based on a large number of ground samples obtained from long-term coastal zone surveys, several classified latitude zones are divided at 3°-5° (with an elastic range to ensure continuity in some areas of the coastal zone).

[0084] For each classified latitude band, the monthly mean index and monthly variance index of each calculated vegetation index were first synthesized month by month, and the annual mean index and annual variance index of each vegetation index were calculated.

[0085] Based on the annual vegetation index time series, the NDVI series and EVI series were further fitted with the double logistic function to obtain the start of growing season (SOS) and end of growing season (EOS) phenological indices of different vegetation.

[0086] So far, the monthly / annual means and variances of the four vegetation indices calculated above, as well as the SOS and EOS phenological indices generated based on the NDVI and EVI sequences were calculated for each pixel, totaling 106 candidate classification indices.

[0087] Using different vegetation classification samples obtained based on field surveys, the Jeffries-Matusita distance (JM distance) of each vegetation index on different vegetation classification samples is calculated to represent the index separability. The JM distance is a transformation of the Bhattacharyya distance and is a widely used measure of the spectral separability distance between two class density functions, and is usually used as a class separability measure.

[0088] First, the indicators ranked in the top 50% of the JM distance ranking were extracted as the preliminary classification variables of the random forest (smileRandomForest) model, and then the contribution of each indicator to vegetation classification was ranked. After sorting, the top 10 indicators with the highest contribution were used as the most separable window vegetation index to construct the final random forest model for final classification.

[0089] The vegetation classification samples were divided into 8:2, and the 10-fold test method was used. After iterative training with the real vegetation classification samples obtained from field surveys, the random forest model with optimal parameters (the accuracy of each classification indicator was above 0.9) was used as the independent variables. The above 10 most distinguishable window period vegetation indices were used as independent variables, and the random forest model (machine learning model) was deployed in salt marsh wetlands across the country to complete the classification of salt marsh vegetation across the country.

[0090] For accuracy verification, indexes such as User's Accuracy (UA), Producer's Accuracy (PA), Overall Accuracy (OA), F-score and Kappa coefficient were used for verification.

[0091]

[0092] Among them, TP is the positive sample correctly classified by the model, TN is the negative sample correctly classified by the model, FP is the positive sample incorrectly classified by the model, FN is the negative sample incorrectly classified by the model, and N is the total number of pixels. i is the total number of actual observations of a certain sample type, b i The total number of model classification values ​​for a certain sample type.

[0093] The technical solution of this application is verified using the entire eastern coastal zone of China. The Landsat 8TOA and Sentinel-2TOA data from 2017 to 2022 are obtained in Google Earth Engine, and the necessary basic image preprocessing such as geometric correction, cloud mask, and vegetation index calculation is completed to synthesize the vegetation index time series.

[0094] Using a large number of field vegetation classification samples obtained from long-term coastal field surveys and ultra-high-resolution satellite remote sensing images from 2017 to 2022, key classification windows and classification indicators with high separability in multiple vegetation indices were screened to establish a random forest model and perform parameter tuning. The final constructed random forest model was used to classify salt marsh vegetation in China's coastal zone from 2017 to 2022. The classification results and accuracy are shown in Figure 2. Figure 4 And as shown in Table 1.

[0095] Table 1 Accuracy evaluation of model classification results (in the past five years)

[0096]

[0097] Based on the same inventive concept, the embodiment of the present application also provides a salt marsh vegetation classification system, including:

[0098] The vegetation index time series generation module is used to generate vegetation index time series based on satellite remote sensing data; the vegetation index time series includes normalized difference vegetation index time series, enhanced vegetation index time series, normalized difference water index time series and corrected red edge chlorophyll vegetation index time series.

[0099] The monthly index and annual index calculation module is used to divide the coastal zone latitude into several classified latitude bands as classification intervals, and calculate the monthly mean index and monthly variance index of each vegetation index in each classification interval, as well as the annual mean index and annual variance index of each vegetation index in the year.

[0100] The growth period candidate index and decay period candidate index calculation module is used to calculate the growth period and decay period of vegetation as candidate classification indicators based on the normalized difference vegetation index time series and the enhanced vegetation index time series using a bilogistic function.

[0101] The JM distance calculation module is used to combine different vegetation classification samples and calculate the JM distance of each candidate classification index on different vegetation classification samples.

[0102] Based on the JM distance of each candidate indicator on different vegetation classification samples and the preliminary classification contribution evaluation, the most distinguishable classification indicator was screened out and the final random forest classification model was constructed.

[0103] A classification module is used to classify salt marsh vegetation in coastal salt marsh wetlands according to the final random forest classification model.

[0104] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (I / O for short) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a salt marsh vegetation classification method is implemented.

[0105] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0106] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0107] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0108] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0109] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0110] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0111] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0112] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for classifying salt marsh vegetation, characterized in that: The salt marsh vegetation classification method comprises: Generate vegetation index time series based on satellite remote sensing data; vegetation index time series includes normalized difference vegetation index time series, enhanced vegetation index time series, normalized difference water index time series and corrected red edge chlorophyll vegetation index time series; The coastal zone latitude is divided into several classification latitude bands as classification intervals, and the monthly mean index and monthly variance index of each vegetation index in each classification interval are calculated, as well as the annual mean index and annual variance index of each vegetation index in the current year; Based on the normalized difference vegetation index time series and the enhanced vegetation index time series, the growth period and decay period of vegetation are calculated using a bilogistic function as candidate classification indicators; Combine different vegetation classification samples to calculate the JM distance of each candidate classification index on different vegetation classification samples; each candidate index includes monthly mean index, monthly variance index, annual mean index, annual variance index, growth period index and decay period index; Based on the JM distance of each candidate indicator on different vegetation classification samples and the preliminary classification contribution evaluation, the most distinguishable classification indicator is selected to construct the final random forest classification model; The salt marsh vegetation in the coastal salt marsh wetlands was classified according to the final random forest classification model.

2. The salt marsh vegetation classification method according to claim 1, characterized in that: Based on the JM distance and preliminary classification contribution evaluation, the most separable window vegetation index was selected and a random forest model was constructed, including: The indicators ranked in the top 50% of the JM distance ranking were extracted as preliminary classification variables for the random forest model; Calculate the contribution of each indicator to the random forest classification model and rank the contributions; The top 10 indicators with the highest contribution were used as the most distinguishable vegetation index for the window period; The most separable window vegetation index is used as the independent variable of the final random forest classification model to construct the final random forest classification model.

3. The salt marsh vegetation classification method according to claim 1, characterized in that: Generating a vegetation index time series according to satellite remote sensing data, which also includes: the satellite remote sensing data includes pre-processed Landsat-8 image data and Sentinel-2 image data; Using the QA band of Landsat-8 to remove pixels covered by clouds and shadows in the Landsat-8 image data, to generate Landsat-8 image data after removal; The sentinel-2 cloud detector library and the S2_CLOUD_PROBABILITY dataset are used to remove pixels covered by clouds and shadows in the Sentinel-2 image data to generate the removed Sentinel-2 image data; Taking the removed Sentinel-2 image data as a benchmark, the difference between the removed Landsat-8 image data and the removed Sentinel-2 image data is corrected by using the least square method to determine the optimal satellite remote sensing data.

4. The salt marsh vegetation classification method according to claim 3, characterized in that: The pixels covered by clouds and shadows in the Landsat-8 image data are removed by using the QA band of Landsat-8, so as to generate the Landsat-8 image data after removal, which specifically includes: The QA band of Landsat-8 is used to determine the pixel occlusion information covered by clouds and shadows in the Landsat-8 image data in a decimal manner; Convert the pixel occlusion information into binary, and determine the pixel quality based on the number of bits corresponding to the binary pixel occlusion information; Based on the pixel quality, a mask function is used to remove pixels covered by clouds and shadows in the Landsat-8 image data to generate Landsat-8 image data after removal.

5. The salt marsh vegetation classification method according to claim 3, characterized in that: The sentinel-2 cloud detector library and the S2_CLOUD_PROBABILITY dataset are used to remove pixels covered by clouds and shadows in the Sentinel-2 image data to generate the removed Sentinel-2 image data, specifically including: Based on the sentinel-2 cloud detector library, determining the pixel coverage probability of being covered by clouds and shadows in the Sentinel-2 image data according to the S2_CLOUD_PROBABILITY dataset; Based on the pixel coverage probability, a filter function is used to remove pixels covered by clouds and shadows in the Sentinel-2 image data to generate Sentinel-2 image data after removal.

6. The salt marsh vegetation classification method according to claim 1, characterized in that: Generate vegetation index time series based on satellite remote sensing data, including: The Savitzky-Golay filtering algorithm is used to smooth the salt and pepper noise and extreme values ​​appearing in the satellite remote sensing data collected within a set period, and the Sentinel-2 image data in the satellite remote sensing data is resampled using cubic convolution, so that the spatial resolution of the resampled Sentinel-2 data is consistent with that of the Landsat-8 image data in the satellite remote sensing data, thereby generating a vegetation index time series; the vegetation index time series is displayed as an annual characteristic distribution map of the vegetation index.

7. A salt marsh vegetation classification system, characterized in that: include: Vegetation index time series generation module, used to generate vegetation index time series based on satellite remote sensing data; vegetation index time series includes normalized difference vegetation index time series, enhanced vegetation index time series, normalized difference water index time series and corrected red edge chlorophyll vegetation index time series; The monthly index and annual index calculation module is used to divide the coastal zone latitude into several classification latitude bands as classification intervals, and calculate the monthly mean index and monthly variance index of each vegetation index in each classification interval, as well as the annual mean index and annual variance index of each vegetation index in the year; A growth period candidate index and decay period candidate index calculation module, used to calculate the growth period and decay period of vegetation as candidate classification indicators using a bilogistic function based on the normalized difference vegetation index time series and the enhanced vegetation index time series; The JM distance calculation module is used to combine different vegetation classification samples to calculate the JM distance of each candidate classification index on different vegetation classification samples; each candidate index includes a monthly mean index, a monthly variance index, an annual mean index, an annual variance index, a growth period index, and a decay period index; Based on the JM distance of each candidate indicator on different vegetation classification samples and the preliminary classification contribution evaluation, the most distinguishable classification indicator is selected to construct the final random forest classification model; A classification module is used to classify salt marsh vegetation in coastal salt marsh wetlands according to the final random forest classification model.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the salt marsh vegetation classification method according to any one of claims 1 to 6.

9. A 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 salt marsh vegetation classification method described in any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the salt marsh vegetation classification method described in any one of claims 1 to 6 are implemented.

Images