A monitoring and early warning method for the distribution of Spartina alterniflora based on image recognition

Abstract

The present invention relates to the field of Spartina alterniflora image recognition technology, specifically a Spartina alterniflora distribution monitoring and early warning method based on image recognition, including: extracting vegetation features using Sentinel 1 radar data and Sentinel 2 optical remote sensing data, and realizing Spartina alterniflora regional classification in combination with TWDTW model, and obtaining the first Spartina alterniflora region; calculating the ecological factors and water conservancy factors of the first Spartina alterniflora region, and generating a grading region by weighted summation; regularly using an unmanned aerial vehicle to shoot the ground image of the first Spartina alterniflora region according to the priority order of the grading region, and using a target detection model to identify the distribution change of Spartina alterniflora, generating coverage area change rate, diffusion rate and coverage density change rate indicators, and forming a dynamic early warning report. The method improves the monitoring accuracy of the distribution of Spartina alterniflora by identifying the Spartina alterniflora region and analyzing the diffusion trend of Spartina alterniflora in real time, and provides a scientific decision-making basis for ecological environmental protection and Spartina alterniflora governance.

Description

Technical Field

[0001] The invention relates to the technical field of Spartina alterniflora image recognition, and in particular to a Spartina alterniflora distribution status monitoring and early warning method based on image recognition. Background Art

[0002] Spartina alterniflora is a typical wetland plant. Due to its strong adaptability and rapid spread, it has become a widely spread alien invasive species around the world in recent years, especially in aquatic environments such as coastal wetlands and intertidal zones. The spread of Spartina alterniflora not only changes the ecological environment, but also threatens local water quality and biodiversity. Therefore, how to accurately identify its growth area and effectively respond to the invasion of Spartina alterniflora has become an important issue that needs to be urgently addressed in the field of water conservancy management and ecological protection.

[0003] At present, remote sensing image data has been widely used in the identification and distribution analysis of Spartina alterniflora. For example, Sentinel-2 satellite images are widely used to identify vegetation such as Spartina alterniflora due to their high spatial resolution and multispectral characteristics. However, in complex wetland and intertidal environments, their recognition accuracy still has certain limitations. In addition, most existing studies are based on static image data and lack real-time monitoring of the growth dynamics of Spartina alterniflora. The growth of Spartina alterniflora is closely related to factors such as seasonal water levels and tides. Static images alone cannot accurately capture its growth pattern and diffusion trend, so the monitoring accuracy of the distribution of Spartina alterniflora needs to be further improved.

[0004] Therefore, a monitoring and early warning method for the distribution of Spartina alterniflora based on image recognition was proposed. Summary of the invention

[0005] The object of the present invention is to provide a kind of Spartina alterniflora distribution monitoring and early warning method based on image recognition, by using Sentinel-1 radar data and Sentinel-2 optical remote sensing data to extract vegetation features, and combine TWDTW model to realize Spartina alterniflora regional classification, obtain the first Spartina alterniflora region; calculate the ecological factors and water factors of the first Spartina alterniflora region, and generate classification areas by weighted summation; regularly use unmanned aerial vehicles to shoot ground images according to the priority order of classification areas, and identify the distribution change of Spartina alterniflora by target detection model, generate coverage area change rate, diffusion rate and coverage density change rate indicators, and form dynamic early warning reports. This method improves the monitoring accuracy of the distribution of Spartina alterniflora, and provides a scientific decision-making basis for ecological environmental protection and Spartina alterniflora governance.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for monitoring and early warning of the distribution status of Spartina alterniflora based on image recognition, comprising:

[0008] Acquire image data of the area to be measured, the image data including Sentinel-1 radar data and Sentinel-2 optical remote sensing data, and pre-process the image data to generate first data;

[0009] Extracting vegetation characteristic variables according to the first data;

[0010] Collecting vegetation sample points in the area to be tested, and marking vegetation labels on the vegetation sample points; the vegetation labels include Spartina alterniflora and Spartina non-alterniflora;

[0011] According to the vegetation sample points and the vegetation characteristic variables, the TWDTW model is used to classify the vegetation in the tested area to obtain a first Spartina alterniflora area;

[0012] An ecological factor is calculated based on the first Spartina alterniflora area and the first data; a water factor is calculated based on the water quality conditions of the first Spartina alterniflora area;

[0013] The ecological factor and the water factor are weighted and summed to generate an impact factor of Spartina alterniflora, and the first Spartina alterniflora area is divided according to the impact factor of Spartina alterniflora to obtain graded areas; the graded areas include high-risk areas, medium-risk areas and low-risk areas;

[0014] According to the priority order of the classified areas, using a drone to photograph ground images of the first Spartina alterniflora area at fixed time intervals;

[0015] Using a target detection model to identify the Spartina alterniflora in the ground image, to obtain a second Spartina alterniflora area;

[0016] The coverage area change rate, diffusion rate and coverage density change rate of the second Spartina alterniflora area are calculated, and early warning reports are generated respectively according to the classified areas.

[0017] Furthermore, the pretreatment process includes:

[0018] Downloading the Sentinel-1 radar data and the Sentinel-2 optical remote sensing data within a predetermined time period to generate second data and third data respectively;

[0019] Performing denoising processing on the second data to obtain fourth data;

[0020] Performing cloud cover filtering and atmospheric correction on the third data to obtain fifth data;

[0021] Aligning the fourth data and the fifth data to the same coordinate system, and clipping the fourth data and the fifth data according to the boundary of the area to be measured to obtain sixth data and seventh data;

[0022] The first data is obtained by extracting the band of the sixth data and the band of the seventh data, and fusing the band of the sixth data and the band of the seventh data.

[0023] Further, the band of the sixth data and the band of the seventh data include:

[0024] The band of the sixth data includes VV and VH polarization bands;

[0025] The bands of the seventh data include a B2 band, a B3 band, a B4 band, a B5 band, a B8 band, a B11 band, and a B12 band.

[0026] Furthermore, the process of obtaining the vegetation characteristic variables includes:

[0027] Calculate and extract vegetation index features, spectral feature features and radar remote sensing vegetation features according to the first data; the vegetation index features include NDVI index, EVI index, SAVI index, MSAVI index and NDWI index; the spectral feature features include band ratio and spectral smoothing value; the radar remote sensing vegetation features include backscattering coefficient, polarization index and vegetation texture features;

[0028] DEM data and meteorological data are obtained; terrain features are extracted according to the DEM data; the terrain features include the elevation, slope and slope direction of the area to be measured; meteorological features are extracted according to the meteorological data; the meteorological features include the precipitation and temperature of the area to be measured.

[0029] Furthermore, the calculation process of the ecological factor includes:

[0030] Dividing the first Spartina alterniflora area into N target areas;

[0031] Extracting the NDVI index and backscatter coefficient of the target area;

[0032] According to the NDVI index, the number of Spartina alterniflora pixels in the target area is obtained, and the number of Spartina alterniflora pixels is divided by the total number of pixels in the area to be measured to obtain a vegetation density factor;

[0033] According to the NDVI index, an average NDVI factor is calculated;

[0034] Establishing a first linear regression model according to the backscattering coefficient, wherein the first linear regression model is used to calculate soil moisture;

[0035] Calculating a salinity index according to the B11 band and the B12 band, and establishing a second linear regression model according to the salinity index, wherein the second linear regression model is used to calculate soil salinity;

[0036] The ecological factor is obtained by weighted summing the vegetation density factor, the average NDVI factor, the soil moisture and the soil salinity.

[0037] Furthermore, the calculation process of the water factor includes:

[0038] Extract the B3 band, B4 band, B8 band, B11 band, backscatter coefficient and NDWI index of the target area;

[0039] Calculating water flow velocity according to the backscatter coefficient of the target area;

[0040] A third linear regression model is established according to the B4 band and the B8 band, and the third linear regression model is used to calculate dissolved oxygen;

[0041] A fourth linear regression model is established according to the B3 band and the B11 band, and the fourth linear regression model is used to calculate total suspended particulate matter;

[0042] Extracting a water body area according to the NDWI index, and generating a water body mask according to the water body area;

[0043] Adding the water body masks at different phases to obtain the water body coverage times;

[0044] Divide the number of water body coverage times by the total observation time to obtain the tidal frequency;

[0045] The water factor is obtained by weighted summing the water velocity, the dissolved oxygen, the total suspended particulate matter and the tidal frequency.

[0046] Furthermore, the priority order includes:

[0047] The high-risk area has a higher priority than the medium-risk area and the low-risk area;

[0048] The medium-risk area has a higher priority than the low-risk area.

[0049] Furthermore, the process of the target detection model identifying the Spartina alterniflora includes:

[0050] Collecting historical Spartina alterniflora images and marking classification labels, wherein the historical Spartina alterniflora images are used to train the target detection model;

[0051] Inputting the ground image into the trained target detection model, the target detection model outputs the target frame position of the Spartina alterniflora and the classification label, to obtain a first image;

[0052] The first images are stitched together to generate the second Spartina alterniflora area.

[0053] Furthermore, the early warning report includes:

[0054] Calculating a risk index for the classified area;

[0055] Generate early warning reports for the classified areas respectively according to the risk index;

[0056] The warning reports include level one warning reports, level two warning reports and level three warning reports;

[0057] Among them, if the risk index is greater than the first threshold, a real-time alarm is triggered and the first-level warning report is generated; if the risk index is greater than the second threshold and less than or equal to the first threshold, the second-level warning report is generated and the monitoring frequency is increased; if the risk index is less than or equal to the second threshold, the third-level warning report is generated and the monitoring frequency is maintained.

[0058] Furthermore, the calculation process of the risk index includes:

[0059] Acquire the second Spartina alterniflora area at the fixed time interval to obtain the area, center coordinates and second NDVI index of Spartina alterniflora;

[0060] Calculate the change rate of the area of ​​the Spartina alterniflora at the previous moment and the area of ​​the Spartina alterniflora at the current moment to obtain the coverage area change rate;

[0061] Calculate the moving distance between the center coordinates at the previous moment and the center coordinates at the current moment to obtain the diffusion rate;

[0062] The risk index is calculated by taking a weighted sum of the coverage area change rate, the diffusion rate and the second NDVI index.

[0063] Compared with the prior art, the present invention has the following beneficial effects:

[0064] 1. By fusing Sentinel-1 radar data and Sentinel-2 optical remote sensing data, the characteristics of Spartina alterniflora were extracted, and the TWDTW model was used to classify all vegetation, which enabled the accurate identification of Spartina alterniflora areas. In addition, drones were used to regularly collect ground image data in the Spartina alterniflora area, and the distribution area of ​​Spartina alterniflora was further refined by combining the target detection model, realizing dynamic monitoring of the coverage of Spartina alterniflora, thereby improving the monitoring accuracy and efficiency of the distribution of Spartina alterniflora.

[0065] 2. By calculating ecological factors such as vegetation density, soil moisture and soil salinity, as well as hydraulic factors such as water flow velocity, dissolved oxygen and tidal frequency, and further calculating the impact factors of Spartina alterniflora based on ecological factors and hydraulic factors, we can comprehensively evaluate the growth of Spartina alterniflora in different hydrological environments. According to the impact factors of Spartina alterniflora, the Spartina alterniflora area is divided into high, medium and low risk areas, which can give priority to identifying and judging the risk situation of high-risk areas, thereby improving the monitoring accuracy of the distribution of Spartina alterniflora and optimizing the resource allocation for monitoring Spartina alterniflora.

[0066] 3. Combining remote sensing data with drone identification technology, we calculated the coverage area change rate, diffusion rate, and coverage density change rate of the Spartina alterniflora area, and generated risk indicators, realizing accurate monitoring of the dynamic changes in the distribution of Spartina alterniflora. At the same time, we automatically generated early warning reports based on risk indicators, which can give priority to high-risk areas, and dynamically optimize monitoring frequency and response strategies based on real-time data, thereby further improving the monitoring accuracy of the distribution of Spartina alterniflora and providing a solid technical guarantee for long-term ecological and environmental management. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 A schematic diagram of a flow chart of a method for monitoring and early warning of the distribution status of Spartina alterniflora provided in an embodiment of the present invention;

[0068] Figure 2 A schematic diagram of a process for obtaining a first Spartina alterniflora region provided in an embodiment of the present invention;

[0069] Figure 3 A schematic diagram of the calculation structure of the risk index provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] See also Figures 1 to 3 The present invention provides a method for monitoring and early warning of the distribution of Spartina alterniflora based on image recognition, and the technical solution is as follows:

[0072] Embodiment 1

[0073] Area A is a typical coastal salt marsh wetland with rich and unique ecological resources. However, in recent years, the invasion of Spartina alterniflora has shown an accelerated spread, posing a serious threat to the local ecosystem. Since Spartina alterniflora has a strong ability to grow and spread, it has rapidly expanded in Area A with the help of factors such as tides and wind. Native halophytes, such as sea sedge and reed, have also been gradually replaced by Spartina alterniflora, and the ecological balance of the wetland has also suffered certain damage. The spatial scale of Area A is relatively small, and the spatial distribution of aquatic plants changes dramatically. Existing technologies such as satellite remote sensing technology are limited by multiple factors such as resolution and climatic conditions when monitoring Spartina alterniflora in coastal wetlands, and there are certain limitations in detection accuracy. In addition, existing technologies have also failed to provide real-time dynamics of the spread of Spartina alterniflora, and cannot provide support for rapid response. In order to solve these problems, a new monitoring and early warning method is needed.

[0074] Figure 1 A schematic flow chart of a method for monitoring and early warning of the distribution status of Spartina alterniflora provided in an embodiment of the present invention.

[0075] like Figure 1 As shown, a method for monitoring and early warning of the distribution status of Spartina alterniflora based on image recognition comprises:

[0076] Step S1: Acquire image data of the area to be measured, the image data including Sentinel-1 radar data and Sentinel-2 optical remote sensing data, and pre-process the image data to generate first data.

[0077] The area to be tested may be an ecologically sensitive area such as an urban area, a rural area, a coastal wetland, etc. In this embodiment, the area to be tested is area A.

[0078] Furthermore, the pretreatment process includes:

[0079] Downloading the Sentinel-1 radar data and the Sentinel-2 optical remote sensing data within a predetermined time period to generate second data and third data respectively;

[0080] In this embodiment, the predetermined time period is 1 year.

[0081] Specifically, region A is a coastal wetland, and optical satellites may not be able to image it due to cloud cover. By adding radar data, continuous observation images can be provided. In this embodiment, Sentinel-1 radar images and Sentinel-2 optical images are downloaded from the European Space Agency platform, and satellite images of region A this year are screened, as shown in Table 1, and finally 19 images are obtained.

[0082] Table 1 Satellite data collection example

[0083]

[0084] Performing denoising processing on the second data to remove speckle noise in the radar image to obtain fourth data;

[0085] Performing cloud coverage filtering and atmospheric correction on the third data, filtering pixels with higher cloud coverage, and obtaining fifth data;

[0086] Aligning the fourth data and the fifth data to the same coordinate system, and clipping the fourth data and the fifth data according to the boundary of the area to be measured, removing irrelevant areas outside the A area, to obtain the sixth data and the seventh data;

[0087] The first data is obtained by extracting the band of the sixth data and the band of the seventh data, and fusing the band of the sixth data and the band of the seventh data.

[0088] Specifically, on the Google Earth Engine (GEE) platform, the Sentinel-1 radar data and Sentinel-2 optical remote sensing data were preprocessed. Through denoising and atmospheric correction, the interference factors in the original data can be effectively removed, providing accurate data support for subsequent dynamic monitoring. In addition, the use of two types of satellite image data, Sentinel-1 radar data and Sentinel-2 optical remote sensing data, made up for the shortcomings of a single data source, thereby improving the monitoring accuracy of the distribution of Spartina alterniflora.

[0089] Further, the band of the sixth data and the band of the seventh data include:

[0090] The bands of the sixth data include:

[0091] VV polarization band: vertical polarization band, used for vegetation research and soil moisture estimation;

[0092] VH polarization band: vertical polarization transmission and horizontal polarization reception band, used to distinguish different vegetation types;

[0093] The bands of the seventh data include:

[0094] B2 band: blue light band, used for water body detection, Spartina alterniflora health assessment and bare land distribution analysis;

[0095] B3 band: green light band, used for monitoring the health status of Spartina alterniflora;

[0096] B4 band: red light band, used for Spartina alterniflora cover analysis and land use classification;

[0097] B5 band: cyan band, used for analyzing water and atmospheric composition near Spartina alterniflora;

[0098] Band B8: Near infrared band, used to assess the density of Spartina alterniflora;

[0099] B11 band: short-wave infrared band, used to distinguish soil and Spartina alterniflora;

[0100] B12 band: shortwave infrared band, used for surface classification and wetland hydrological research, etc.

[0101] Specifically, electromagnetic waves in different bands have different responses to vegetation. The radar band is more sensitive to surface roughness and humidity, while the optical band is more sensitive to vegetation canopy characteristics and moisture content. Since the distribution of Spartina alterniflora is closely related to water humidity and soil conditions, by extracting the bands of the sixth data (processed Sentinel-1 radar data) and the seventh data (processed Sentinel-2 optical remote sensing data), not only the amount of data is reduced, but also these complementary information can be integrated to carefully depict the surface roughness and other information of Spartina alterniflora in the image, thereby enhancing the comprehensive understanding of the distribution of Spartina alterniflora and the monitoring accuracy.

[0102] Step S2: extracting vegetation characteristic variables according to the first data.

[0103] Furthermore, the process of obtaining the vegetation characteristic variables includes:

[0104] Calculate and extract vegetation index features, spectral feature features and radar remote sensing vegetation features according to the first data; wherein the vegetation index features include NDVI index, EVI index, SAVI index, MSAVI index and NDWI index; the spectral feature features include band ratio and spectral smoothing value; the radar remote sensing vegetation features include backscattering coefficient, polarization index and vegetation texture features;

[0105] Obtain DEM data and meteorological data;

[0106] Extracting terrain features according to the DEM data;

[0107] Among them, the terrain characteristics include the elevation, slope and slope direction of the area to be measured; the meteorological characteristics are extracted according to the meteorological data; the meteorological characteristics include the precipitation and temperature of the area to be measured.

[0108] Specifically, in this embodiment, the calculation and acquisition process of the vegetation index class feature variables is shown in Table 2, the calculation and acquisition process of the spectral feature class features is shown in Table 3, and the calculation and acquisition process of the radar remote sensing vegetation class features is shown in Table 4.

[0109] Vegetation index features are standard features in remote sensing analysis, which can directly reflect the health status, growth status and coverage of Spartina alterniflora; spectral feature features can capture the differences in spectral responses of different vegetation types, help distinguish different vegetation, and thus more easily extract Spartina alterniflora areas; radar remote sensing vegetation features are not affected by weather and light, and can effectively supplement the deficiencies of optical remote sensing data; for areas with complex terrain, geographical features have a constraining effect on vegetation identification, and extracting terrain features can provide more comprehensive environmental information; meteorological features are key factors in vegetation growth and can reflect the phenological characteristics of vegetation. Combining these features can reflect the growth and distribution of Spartina alterniflora and other vegetation from a multi-dimensional perspective, help improve the accuracy of classification, and thus improve the monitoring accuracy of the distribution of Spartina alterniflora.

[0110] Table 2 Vegetation index characteristic variables

[0111]

[0112] Table 3 Spectral feature variables

[0113]

[0114] Table 4 Characteristics of vegetation types using radar remote sensing

[0115]

[0116] Step S3: collecting vegetation sample points in the area to be tested, and marking vegetation labels on the vegetation sample points; the vegetation labels include Spartina alterniflora and Spartina non-alterniflora;

[0117] Specifically, in this implementation, vegetation sample points are mainly obtained through field sampling and remote sensing image interpretation. According to the downloaded remote sensing image data, representative areas in region A are selected for sampling, and the precise coordinates of each sample point are recorded, and a total of 65 sample points are selected. Then, through the Collect Earth platform, the image data of the sampling area is selected, and a total of 153 sample points are selected. All sample points cover different time periods (spring, summer, autumn and winter). Combining the sample points of field sampling and the sample points of image data, the vegetation sample points are classified to obtain a total of 7 vegetation categories, specifically Spartina alterniflora, Reed, Moss, Zizania latifolia, water bodies and mudflats.

[0118] Step S4: According to the vegetation sample points and the vegetation characteristic variables, the TWDTW model is used to classify the vegetation in the area to be tested, so as to obtain a first Spartina alterniflora area;

[0119] Among them, the TWDTW (Time-Weighted Dynamic Time Warping) model is a classification model based on time series, which can be used to classify different vegetation and extract the Spartina alterniflora area.

[0120] Figure 2 A schematic diagram of the process of obtaining the first Spartina alterniflora area provided in an embodiment of the present invention.

[0121] Specifically, in this embodiment, Figure 2 As shown in the figure, the vegetation characteristic variables of each sample point are extracted and arranged into a time series in chronological order. The time series of each vegetation type is normalized and its mean is calculated to form a template curve. For example, the NDVI curve of Spartina alterniflora is: 0.45 (spring), 0.78 (summer), 0.62 (autumn) and 0.32 (winter). The weight function is then introduced to assign high weights to the growth period of Spartina alterniflora, expressed as:

[0122] ;

[0123] in, is the weight factor at time t, is the width of the weight distribution, set to 1.5, It is the critical time point, i.e. the growth period, which is set to 6.5.

[0124] Then, the matching error between the time series of each sample point and the template curve is calculated according to the weight factor, which is expressed as:

[0125] ;

[0126] in, is the matching error, is the time series value of the sample point at time t, is the value of the template curve at time t, and the vegetation type corresponding to the template with the smallest error is the classification result of the sample point. In order to prove the beneficial effect of the TWDTW model, this embodiment provides three comparative models. As shown in Table 5, the TWDTW model has the highest classification accuracy and can be well used for the extraction of Spartina alterniflora in area A.

[0127] Table 5 Comparison of classification results

[0128]

[0129] Step S5: calculating an ecological factor according to the first Spartina alterniflora area and the first data; calculating a water factor according to the water quality conditions of the first Spartina alterniflora area;

[0130] The ecological factor and the water factor are weighted and summed to generate an impact factor of Spartina alterniflora, and the first Spartina alterniflora area is divided according to the impact factor of Spartina alterniflora to obtain graded areas; the graded areas include high-risk areas, medium-risk areas and low-risk areas;

[0131] Furthermore, the calculation process of the ecological factor includes:

[0132] Dividing the first Spartina alterniflora area into N target areas;

[0133] Specifically, in this embodiment, the boundaries of water bodies and the distribution areas of Spartina alterniflora in region A are extracted to generate preliminary partitions, which are then divided into regular 50 m×50 m grids to obtain 12 target areas.

[0134] Extracting the NDVI index and backscatter coefficient of the target area;

[0135] According to the NDVI index, the number of Spartina alterniflora pixels in the target area is obtained, and the number of Spartina alterniflora pixels is divided by the total number of pixels in the area to be measured to obtain a vegetation density factor, which is expressed as:

[0136] ;

[0137] in, is the vegetation density factor, is the number of pixels of Spartina alterniflora, is the total number of pixels in the area to be tested;

[0138] Among them, when the NDVI index is greater than 2, the Spartina alterniflora pixels are extracted.

[0139] The average NDVI factor is calculated based on the NDVI index and is expressed as:

[0140] ;

[0141] in, is the average NDVI factor, is the total number of NDVI factors, is the NDVI factor number, For the NDVI factors;

[0142] A first linear regression model is established according to the backscattering coefficient. The first linear regression model is used to calculate soil moisture and is expressed as:

[0143] ;

[0144] in, is soil moisture, and is the parameter obtained by regression fitting based on the ground measured data, is the backscattering coefficient;

[0145] Among them, in this embodiment, and Can be preset to and 25.6, and further adjustments will be made based on measured data.

[0146] The salinity index is calculated according to the B11 band and the B12 band, and a second linear regression model is established according to the salinity index. The second linear regression model is used to calculate soil salinity, which is expressed as:

[0147] ;

[0148] in, is the soil salinity, and The parameters are obtained by regression fitting based on the ground-based measured data and are pre-set to 5.2 and 0.8.

[0149] The vegetation density factor, the average NDVI factor, the soil moisture and the soil salinity are weighted and summed to obtain the ecological factor, which is expressed as:

[0150] ;

[0151] in, As ecological factors, , , and are weight factors, which are set to 0.3, 0.2, 0.25, and 0.25 respectively.

[0152] The vegetation density factor is the coverage degree of Spartina alterniflora in the target area. Calculating the vegetation density factor can reflect its expansion intensity and distribution. NDVI is an important indicator for evaluating the biomass of Spartina alterniflora. Calculating the average NDVI factor can reflect the overall health of the target area. Soil moisture and soil salinity affect the growth of Spartina alterniflora. Calculating soil moisture and soil salinity can help identify the growth conditions of Spartina alterniflora. The comprehensive calculation of ecological factors can achieve accurate quantification of the ecological factors of Spartina alterniflora, avoid the one-sidedness that may be caused by a single indicator, and thus accurately evaluate the growth status of Spartina alterniflora in each target area and improve the monitoring accuracy of the distribution status of Spartina alterniflora.

[0153] Furthermore, the calculation process of the water factor includes:

[0154] Extract the B3 band, B4 band, B8 band, B11 band, backscatter coefficient and NDWI index of the target area;

[0155] The water velocity is calculated according to the backscatter coefficient of the target area, which is expressed as:

[0156] ;

[0157] in, is the water flow velocity, is the proportionality constant, is the backscatter coefficient, h is the assumed water depth, and n is the exponent;

[0158] In this embodiment, the preset is 0.05, h is 2, and n is 1.5, which can be adjusted later according to the actual deviation.

[0159] A third linear regression model is established based on the B4 band and the B8 band. The third linear regression model is used to calculate dissolved oxygen and is expressed as:

[0160] ;

[0161] in, For dissolved oxygen, and are the parameters obtained by regression fitting based on the ground measured data;

[0162] In this embodiment, the preset is 0.8, is -0.5.

[0163] A fourth linear regression model is established based on the B3 band and the B11 band. The fourth linear regression model is used to calculate the total suspended particulate matter, which is expressed as:

[0164] ;

[0165] in, is the total suspended particulate matter, and are the parameters obtained by regression fitting based on the ground measured data;

[0166] In this embodiment, the preset is 1.2, is 0.7.

[0167] Extracting a water body area according to the NDWI index, and generating a water body mask according to the water body area;

[0168] Among them, the NDWI index is greater than 0.3.

[0169] Adding the water body masks at different phases to obtain the water body coverage times;

[0170] Divide the number of water body coverage times by the total observation time to obtain the tidal frequency;

[0171] The water factor is obtained by weighted summing of the water velocity, the dissolved oxygen, the total suspended particulate matter and the tidal frequency, and is expressed as:

[0172] ;

[0173] in, is the water factor, , , and are weight factors, which are set to 0.3, 0.2, 0.2, and 0.3 respectively.

[0174] Specifically, Spartina alterniflora is usually distributed in areas with low water flow rates. The growth of Spartina alterniflora affects water flow velocity and tidal frequency. Calculating water flow velocity and tidal frequency can help preliminarily determine the diffusion capacity of Spartina alterniflora. The dissolved oxygen level may affect the competitive relationship between other wetland plants and Spartina alterniflora. Calculating dissolved oxygen can assess whether Spartina alterniflora puts pressure on the original ecosystem. The root system of Spartina alterniflora changes the properties of soil and sediments, thereby affecting the concentration of total suspended particulate matter. Calculating total suspended particulate matter can assist in assessing the impact range of Spartina alterniflora. By combining these factors, the hydraulic factors that affect the growth and spread of Spartina alterniflora are quantified, which improves the monitoring accuracy of the distribution of Spartina alterniflora.

[0175] Further, dividing the first Spartina alterniflora area according to the Spartina alterniflora impact factor to obtain graded areas;

[0176] The impact factor of Spartina alterniflora is expressed as:

[0177] ;

[0178] in, is the impact factor of Spartina alterniflora, and are weight factors, all set to 0.5. As ecological factors, are water conservancy factors, and are all normalized results.

[0179] The classification areas include:

[0180] High-risk areas (such as ): The vegetation density factor is high, and the ecological factors and water factors are seriously disturbed.

[0181] Medium risk areas (such as ): The distribution of Spartina alterniflora is locally concentrated, and the ecological factors and water factors are disturbed to a certain extent.

[0182] Low risk areas (such as ); Spartina alterniflora is sparsely distributed, and water conservancy and ecology are less affected by it.

[0183] Step S6: using a drone to photograph ground images of the first Spartina alterniflora area at fixed time intervals according to the priority order of the grading areas;

[0184] Using a target detection model to identify the Spartina alterniflora in the ground image, to obtain a second Spartina alterniflora area;

[0185] Furthermore, the priority order includes:

[0186] The high-risk area has a higher priority than the medium-risk area and the low-risk area;

[0187] The medium-risk area has a higher priority than the low-risk area.

[0188] The spread and destructiveness of Spartina alterniflora are high. Prioritizing monitoring of high-risk areas can help identify problems and issue early warnings. At the same time, drone monitoring and image processing consume time and resources. Prioritizing monitoring of high-risk areas can help concentrate resources and reduce resource waste. In addition, if a medium-risk area enters a high-risk state due to changes, the priority of this target area can be adjusted at any time, thereby improving the monitoring accuracy of the distribution of Spartina alterniflora.

[0189] Furthermore, the process of the target detection model identifying the Spartina alterniflora includes:

[0190] Collecting historical Spartina alterniflora images and marking classification labels, wherein the historical Spartina alterniflora images are used to train the target detection model;

[0191] Inputting the ground image into the trained target detection model, the target detection model outputs the target frame position of the Spartina alterniflora and the classification label, to obtain a first image;

[0192] Among them, in this embodiment, the target detection model selects the Faster R-CNN model, and the test results of the trained target detection model are shown in Table 6. It can be observed that Faster R-CNN is particularly suitable for the detection of Spartina alterniflora in complex backgrounds, with the highest accuracy.

[0193] Table 6 Comparison of recognition results

[0194]

[0195] All the first images were geometrically corrected, registered and stitched to generate a complete Spartina alterniflora distribution image covering all target areas, which was recorded as the second Spartina alterniflora area.

[0196] Through the stitching of the second Spartina alterniflora area images, the specific distribution of Spartina alterniflora in area A can be displayed more accurately and comprehensively. In addition, based on the real-time images taken by drones, the Spartina alterniflora distribution map can be updated regularly, and the expansion trend of Spartina alterniflora can be dynamically tracked, thereby improving the monitoring accuracy of the distribution status of Spartina alterniflora.

[0197] Step S7: calculating the coverage area change rate, diffusion rate and coverage density change rate of the second Spartina alterniflora area, and generating early warning reports respectively according to the classified areas.

[0198] Furthermore, the early warning report includes:

[0199] Calculating a risk index for the classified area;

[0200] Generate early warning reports for the classified areas respectively according to the risk index;

[0201] The warning reports include level one warning reports, level two warning reports and level three warning reports;

[0202] Among them, if the risk index is greater than the first threshold, a real-time alarm is triggered and the first-level warning report is generated; if the risk index is greater than the second threshold and less than or equal to the first threshold, the second-level warning report is generated and the monitoring frequency is increased; if the risk index is less than or equal to the second threshold, the third-level warning report is generated and the monitoring frequency is maintained.

[0203] Specifically, in this embodiment, the first threshold is set to 0.7, and the second threshold is set to 0.4.

[0204] If the risk index of a high-risk area is 0.8, a real-time alarm will be automatically triggered to remind management personnel to intervene quickly and generate a first-level warning report with the following content:

[0205] Risk index: 0.8;

[0206] Change trend: The coverage area change rate is 12%;

[0207] Diffusion direction: northeast, diffusion rate is 3m / half month.

[0208] By generating early warning reports based on risk index classification, targeted prevention and control is achieved, with priority given to high-risk areas, thereby optimizing resource allocation. In addition, the monitoring frequency can be flexibly adjusted according to changes in diffusion behavior, thereby improving the monitoring accuracy of the distribution of Spartina alterniflora.

[0209] Figure 3 A schematic diagram of the structure of calculating the risk index provided by an embodiment of the present invention.

[0210] Furthermore, if Figure 3 As shown, the calculation process of the risk index includes:

[0211] Acquire the second Spartina alterniflora area at the fixed time interval to obtain the area, center coordinates and second NDVI index of Spartina alterniflora;

[0212] In this embodiment, image data of the second Spartina alterniflora area is acquired every half month, and the actual area of ​​the second Spartina alterniflora area is calculated by combining the pixel area of ​​the detected Spartina alterniflora target frame with the actual area corresponding to each pixel; the center position coordinates are found according to the Spartina alterniflora target frame; and the NDVI index is calculated according to the multispectral segment of the image data of the second Spartina alterniflora area, which is recorded as the second NDVI index.

[0213] The change rate of the area of ​​the Spartina alterniflora at the previous moment and the area of ​​the Spartina alterniflora at the current moment is calculated to obtain the coverage area change rate, which is expressed as:

[0214] ;

[0215] in, is the coverage area change rate, is the area of ​​Spartina alterniflora at the current moment, is the area of ​​Spartina alterniflora at the previous moment;

[0216] The moving distance between the center coordinates at the previous moment and the center coordinates at the current moment is calculated to obtain the diffusion rate, which is expressed as:

[0217] ;

[0218] in, is the diffusion rate, is the center coordinate at the current moment, is the center coordinate at the previous moment.

[0219] The risk index is calculated by weighted summing the coverage area change rate, the diffusion rate and the second NDVI index, and is expressed as:

[0220] ;

[0221] in, is the risk index, is the total number of NDVI factors, is the NDVI factor number, For the NDVI factors, , and are weight factors, which are set to 0.4, 0.4, and 0.2 respectively.

[0222] By calculating the rate of change of the area of ​​Spartina alterniflora, the diffusion speed and trend can be accurately reflected; the diffusion rate can capture the spatial migration characteristics during the diffusion process, thereby judging the degree of impact on important areas; the second NDVI index can judge the intensity of Spartina alterniflora's invasion of the surrounding environment. The risk index calculated based on these factors can show the growth status of Spartina alterniflora and highlight the diffusion dynamics of Spartina alterniflora, thereby improving the monitoring accuracy of the distribution of Spartina alterniflora, helping to formulate targeted intervention plans and improve governance efficiency.

[0223] In order to prove the validity and accuracy of the method, the embodiment of the present invention is compared with the existing Spartina alterniflora distribution monitoring method in A region, and a pre-divided region in an A region is randomly selected for detection. As shown in Table 7, the accuracy and dynamic monitoring capability of the present invention are the strongest. Although the ground sample monitoring error is relatively low, it is limited to the sample range, the entire region cannot be covered, and it takes a very long time. Satellite remote sensing monitoring takes the shortest time, but is limited to orbital period and weather conditions, and has a high error for specific Spartina alterniflora regional identification.

[0224] Table 7 Recognition results of a region in region A

[0225]

[0226] The present invention combines Sentinel-1 radar data and Sentinel-2 optical data, and can effectively overcome the limitation of lighting conditions through the comprehensive utilization of multi-source remote sensing information, thereby improving the accuracy and comprehensive coverage of Spartina alterniflora detection; adopts the TWDTW model for vegetation classification, and can accurately identify the difference between Spartina alterniflora and other vegetation according to time series characteristics; when dividing the first Spartina alterniflora area, comprehensively considers ecological factors and water factors, and generates Spartina alterniflora impact factors, which can accurately evaluate the growth state and diffusion potential of Spartina alterniflora; divides the detection area into high-risk area, medium-risk area and low-risk area, and assigns different priority monitoring orders, thereby realizing efficient allocation of resources; uses unmanned aerial vehicles to regularly shoot the classified areas, and realizes real-time dynamic monitoring of Spartina alterniflora in combination with the target detection model. The risk index is calculated by comprehensively using key indicators such as diffusion rate, coverage area change rate and coverage density change rate, which comprehensively reflects the actual diffusion trend of Spartina alterniflora and generates graded early warning reports on the diffusion trend. It not only achieves accurate monitoring of the distribution status of Spartina alterniflora, but also helps staff quickly identify the risk level of Spartina alterniflora in the area, providing a basis for the restoration of the ecosystem.

[0227] Embodiment 2

[0228] Area B is located in an inland freshwater wetland. The wetland environment in Area B is mainly composed of rivers, lakes and rice fields, with complex hydrological conditions and geomorphological features. In addition, Area B is foggy and has frequent rainfall, which interferes with the acquisition and quality of traditional remote sensing images, making the dynamic monitoring and spread assessment of Spartina alterniflora more difficult. Therefore, a new method for monitoring the distribution of Spartina alterniflora is needed.

[0229] A method for monitoring and early warning of the distribution status of Spartina alterniflora based on image recognition, comprising:

[0230] Acquire image data of the area to be measured, the image data including Sentinel-1 radar data and Sentinel-2 optical remote sensing data, and pre-process the image data to generate first data;

[0231] Extracting vegetation characteristic variables according to the first data;

[0232] Collecting vegetation sample points in the area to be tested, and marking vegetation labels on the vegetation sample points; the vegetation labels include Spartina alterniflora and Spartina non-alterniflora;

[0233] According to the vegetation sample points and the vegetation characteristic variables, the TWDTW model is used to classify the vegetation in the tested area to obtain a first Spartina alterniflora area;

[0234] Calculate an ecological factor based on the first Spartina alterniflora area and the first data; calculate a water factor based on the water quality conditions of the first Spartina alterniflora area;

[0235] The ecological factor and the water factor are weighted and summed to generate an impact factor of Spartina alterniflora, and the first Spartina alterniflora area is divided according to the impact factor of Spartina alterniflora to obtain graded areas; the graded areas include high-risk areas, medium-risk areas and low-risk areas;

[0236] According to the priority order of the classified areas, using a drone to photograph ground images of the first Spartina alterniflora area at fixed time intervals;

[0237] Specifically, the drone is equipped with a multispectral camera, and the image resolution can reach the centimeter level. The grading area is divided into regular grids, and a shooting point is set for each grid center point. Each shooting point is traversed according to the optimal path, and the image data is directly transmitted to the ground station after being collected.

[0238] The optimal path may be a pre-set path or may be generated according to the distance of each shooting point using a greedy algorithm.

[0239] Using a target detection model to identify the Spartina alterniflora in the ground image, to obtain a second Spartina alterniflora area;

[0240] The coverage area change rate, diffusion rate and coverage density change rate of the second Spartina alterniflora area are calculated, and early warning reports are generated respectively according to the classified areas.

[0241] Furthermore, the pretreatment process includes:

[0242] Downloading the Sentinel-1 radar data and the Sentinel-2 optical remote sensing data within a predetermined time period to generate second data and third data respectively;

[0243] Performing denoising processing on the second data to obtain fourth data;

[0244] Performing cloud cover filtering and atmospheric correction on the third data to obtain fifth data;

[0245] Aligning the fourth data and the fifth data to the same coordinate system, and clipping the fourth data and the fifth data according to the boundary of the area to be measured to obtain sixth data and seventh data;

[0246] The first data is obtained by extracting the band of the sixth data and the band of the seventh data, and fusing the band of the sixth data and the band of the seventh data.

[0247] Further, the band of the sixth data and the band of the seventh data include:

[0248] The band of the sixth data includes VV and VH polarization bands;

[0249] The bands of the seventh data include a B2 band, a B3 band, a B4 band, a B5 band, a B8 band, a B11 band, and a B12 band.

[0250] Furthermore, the process of obtaining the vegetation characteristic variables includes:

[0251] Calculate and extract vegetation index features, spectral feature features and radar remote sensing vegetation features according to the first data; the vegetation index features include NDVI index, EVI index, SAVI index, MSAVI index and NDWI index; the spectral feature features include band ratio and spectral smoothing value; the radar remote sensing vegetation features include backscattering coefficient, polarization index and vegetation texture features;

[0252] DEM data and meteorological data are obtained; terrain features are extracted according to the DEM data; the terrain features include the elevation, slope and slope direction of the area to be measured; meteorological features are extracted according to the meteorological data; the meteorological features include the precipitation and temperature of the area to be measured.

[0253] Furthermore, the calculation process of the ecological factor includes:

[0254] Dividing the first Spartina alterniflora area into N target areas;

[0255] Extracting the NDVI index and backscatter coefficient of the target area;

[0256] The number of Spartina alterniflora pixels in the target area is obtained according to the threshold value of the NDVI index, and the number of Spartina alterniflora pixels is divided by the total number of pixels in the area to be measured to obtain a vegetation density factor;

[0257] Calculate the average NDVI factor according to the NDVI index;

[0258] Establishing a first linear regression model according to the backscattering coefficient, wherein the first linear regression model is used to calculate soil moisture;

[0259] Calculating a salinity index according to the B11 band and the B12 band, and establishing a second linear regression model according to the salinity index, wherein the second linear regression model is used to calculate soil salinity;

[0260] The ecological factor is obtained by weighted summing the vegetation density factor, the average NDVI factor, the soil moisture and the soil salinity.

[0261] Furthermore, the calculation process of the water factor includes:

[0262] Extract the B3 band, B4 band, B8 band, B11 band, backscatter coefficient and NDWI index of the target area;

[0263] Calculating water flow velocity according to the backscatter coefficient of the target area;

[0264] A third linear regression model is established according to the B4 band and the B8 band, and the third linear regression model is used to calculate dissolved oxygen;

[0265] A fourth linear regression model is established according to the B3 band and the B11 band, and the fourth linear regression model is used to calculate total suspended particulate matter;

[0266] Extracting a water body area according to the NDWI index, and generating a water body mask according to the water body area;

[0267] Adding the water body masks at different phases to obtain the water body coverage times;

[0268] Divide the number of water body coverage times by the total observation time to obtain the tidal frequency;

[0269] The water factor is obtained by weighted summing the water velocity, the dissolved oxygen, the total suspended particulate matter and the tidal frequency.

[0270] Furthermore, the priority order includes:

[0271] The high-risk area has a higher priority than the medium-risk area and the low-risk area;

[0272] The medium-risk area has a higher priority than the low-risk area.

[0273] Furthermore, the process of the target detection model identifying the Spartina alterniflora includes:

[0274] Collecting historical Spartina alterniflora images and marking classification labels, wherein the historical Spartina alterniflora images are used to train the target detection model;

[0275] Inputting the ground image into the trained target detection model, the target detection model outputs the target frame position of the Spartina alterniflora and the classification label, to obtain a first image;

[0276] The first images are stitched together to generate the second Spartina alterniflora area.

[0277] Furthermore, the early warning report includes:

[0278] Calculating a risk index for the classified area;

[0279] Generate early warning reports for the classified areas respectively according to the risk index;

[0280] The warning reports include level one warning reports, level two warning reports and level three warning reports;

[0281] Among them, if the risk index is greater than the first threshold, a real-time alarm is triggered and the first-level warning report is generated; if the risk index is greater than the second threshold and less than or equal to the first threshold, the second-level warning report is generated and the monitoring frequency is increased; if the risk index is less than or equal to the second threshold, the third-level warning report is generated and the monitoring frequency is maintained.

[0282] Furthermore, the calculation process of the risk index includes:

[0283] Acquire the second Spartina alterniflora area at the fixed time interval to obtain the area, center coordinates and second NDVI index of Spartina alterniflora;

[0284] Calculate the change rate of the area of ​​the Spartina alterniflora at the previous moment and the area of ​​the Spartina alterniflora at the current moment to obtain the coverage area change rate;

[0285] Calculate the moving distance between the center coordinates at the previous moment and the center coordinates at the current moment to obtain the diffusion rate;

[0286] The risk index is calculated by taking a weighted sum of the coverage area change rate, the diffusion rate and the second NDVI index.

[0287] Table 8 Identification results of region B

[0288]

[0289] The original scheme in area B used satellite data to monitor Spartina alterniflora. To further demonstrate the effect of the present invention, the original scheme in area B was compared with the present invention. As shown in Table 8, the omission rate of the present invention was reduced to 7.4%, which improved the detection capability of small areas and hidden areas. In addition, the monitoring frequency was customized according to the conditions in area B, which improved the early warning capability and achieved the improvement of the monitoring accuracy of the distribution of Spartina alterniflora.

[0290] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring and early warning the distribution of Spartina alterniflora based on image recognition, characterized in that: include: Acquire image data of the area to be measured, wherein the image data includes Sentinel-1 radar data and Sentinel-2 optical remote sensing data; Preprocessing the image data to generate first data; Extracting vegetation characteristic variables according to the first data; Collecting vegetation sample points in the area to be tested, and marking vegetation labels on the vegetation sample points; the vegetation labels include Spartina alterniflora and Spartina non-alterniflora; According to the vegetation sample points and the vegetation characteristic variables, the TWDTW model is used to classify the vegetation in the tested area to obtain a first Spartina alterniflora area; Calculating ecological factors and water factors according to the first Spartina alterniflora area and the first data; The calculation process of the ecological factor includes: Dividing the first Spartina alterniflora area into N target areas; Extracting the NDVI index and backscatter coefficient of the target area; According to the NDVI index, the number of Spartina alterniflora pixels in the target area is obtained, and the number of Spartina alterniflora pixels is divided by the total number of pixels in the area to be measured to obtain a vegetation density factor; According to the NDVI index, an average NDVI factor is calculated; Establishing a first linear regression model according to the backscattering coefficient, wherein the first linear regression model is used to calculate soil moisture; A salinity index is calculated based on the B11 band and the B12 band; a second linear regression model is established based on the salinity index, and the second linear regression model is used to calculate soil salinity; The ecological factor is obtained by weighted summing the vegetation density factor, the average NDVI factor, the soil moisture and the soil salinity; The calculation process of the water factor includes: Extracting the B3 band, the B4 band, the B8 band, the B11 band, the backscatter coefficient and the NDWI index of the target area; Calculating the water flow velocity according to the backscatter coefficient of the target area; A third linear regression model is established according to the B4 band and the B8 band, and the third linear regression model is used to calculate dissolved oxygen; A fourth linear regression model is established according to the B3 band and the B11 band, and the fourth linear regression model is used to calculate total suspended particulate matter; Extracting a water body area according to the NDWI index, and generating a water body mask according to the water body area; Adding the water body masks at different phases to obtain the water body coverage times; Divide the number of water body coverage times by the total observation time to obtain the tidal frequency; The water factor is obtained by weighted summing the water velocity, the dissolved oxygen, the total suspended particulate matter and the tidal frequency; The ecological factor and the water factor are weighted and summed to generate an impact factor of Spartina alterniflora, and the first Spartina alterniflora area is divided according to the impact factor of Spartina alterniflora to obtain graded areas; the graded areas include high-risk areas, medium-risk areas and low-risk areas; According to the priority order of the classified areas, using a drone to photograph ground images of the first Spartina alterniflora area at fixed time intervals; Using a target detection model to identify the Spartina alterniflora in the ground image, to obtain a second Spartina alterniflora area; Calculate the coverage area change rate, diffusion rate and second NDVI index of the second Spartina alterniflora area, take weighted sum of the coverage area change rate, the diffusion rate and the second NDVI index, calculate the risk index of the classified area, and generate early warning reports for the classified areas according to the risk index.

2. The method for monitoring and early warning of the distribution of Spartina alterniflora based on image recognition according to claim 1, characterized in that: The pretreatment process includes: Downloading the Sentinel-1 radar data and the Sentinel-2 optical remote sensing data within a predetermined time period to generate second data and third data respectively; Performing denoising processing on the second data to obtain fourth data; Performing cloud cover filtering and atmospheric correction on the third data to obtain fifth data; Aligning the fourth data and the fifth data to the same coordinate system, and clipping the fourth data and the fifth data according to the boundary of the area to be measured to obtain sixth data and seventh data; The first data is obtained by fusing the band of the sixth data and the band of the seventh data.

3. A method for monitoring and early warning the distribution of Spartina alterniflora based on image recognition according to claim 2, characterized in that: The band of the sixth data and the band of the seventh data include: The bands of the sixth data include VV and VH polarization bands; The bands of the seventh data include the B2 band, the B3 band, the B4 band, the B5 band, the B8 band, the B11 band, and the B12 band.

4. The method for monitoring and early warning of the distribution of Spartina alterniflora based on image recognition according to claim 1, characterized in that: The process of obtaining the vegetation characteristic variables includes: Calculate and extract vegetation index features, spectral feature features and radar remote sensing vegetation features according to the first data; the vegetation index features include the NDVI index, EVI index, SAVI index, MSAVI index and NDWI index; the spectral feature features include band ratio and spectral smoothing value; the radar remote sensing vegetation features include backscattering coefficient, polarization index and vegetation texture features; DEM data and meteorological data are obtained; terrain features are extracted according to the DEM data; the terrain features include the elevation, slope and slope direction of the area to be measured; meteorological features are extracted according to the meteorological data; the meteorological features include the precipitation and temperature of the area to be measured.

5. The method for monitoring and early warning of the distribution of Spartina alterniflora based on image recognition according to claim 1, characterized in that: The order of priority includes: The high-risk area has a higher priority than the medium-risk area and the low-risk area; The medium-risk area has a higher priority than the low-risk area.

6. The method for monitoring and early warning of the distribution of Spartina alterniflora based on image recognition according to claim 1, characterized in that: The process of the target detection model identifying the Spartina alterniflora includes: Collecting historical Spartina alterniflora images and marking classification labels, wherein the historical Spartina alterniflora images are used to train the target detection model; Inputting the ground image into the trained target detection model, the target detection model outputs the target frame position of the Spartina alterniflora and the classification label, to obtain a first image; The first images are stitched together to generate the second Spartina alterniflora area.

7. The method for monitoring and early warning of the distribution of Spartina alterniflora based on image recognition according to claim 1, characterized in that: The early warning report includes: Calculating the risk index of the classified areas; generating early warning reports for the classified areas respectively according to the risk index; The warning reports include level one warning reports, level two warning reports and level three warning reports; Among them, if the risk index is greater than the first threshold, a real-time alarm is triggered and the first-level warning report is generated; if the risk index is greater than the second threshold and less than or equal to the first threshold, the second-level warning report is generated and the monitoring frequency is increased; if the risk index is less than or equal to the second threshold, the third-level warning report is generated and the monitoring frequency is maintained.

8. The method for monitoring and early warning of the distribution of Spartina alterniflora based on image recognition according to claim 1, characterized in that: The calculation process of the risk index includes: Acquire the second Spartina alterniflora area at the fixed time interval to obtain the area, center coordinates and the second NDVI index of Spartina alterniflora; According to the area of ​​the Spartina alterniflora at the previous moment and the area of ​​the Spartina alterniflora at the current moment, obtaining the coverage area change rate; Calculating the moving distance between the center coordinates at the previous moment and the center coordinates at the current moment to obtain the diffusion rate; The risk index is calculated by taking a weighted sum of the coverage area change rate, the diffusion rate and the second NDVI index.

Images