Wetland vegetation spatial distribution extraction method and system based on remote sensing image

Abstract

The invention discloses a wetland vegetation spatial distribution extraction method and system based on a remote sensing image, and the method comprises the steps: obtaining remote sensing images of spartina alterniflora and other wetland vegetation, carrying out the sample label marking of spartina alterniflora in the remote sensing images, and building a sample data set; training a vector machine SVM by using the sample data set to obtain a trained SVM supervised learning model; collecting a plurality of remote sensing images of wetland vegetation in a wetland area in real time by using an unmanned aerial vehicle; splicing the plurality of remote sensing images to obtain a DEM (Digital Elevation Model), and generating a complete visible light orthoimage of the wetland area; inputting the complete visible light orthoimage of the wetland area into the trained SVM supervised learning model to obtain a wetland vegetation extraction graph and a spartina alterniflora extraction graph; and carrying out area calculation on the wetland vegetation extraction graph and the spartina alterniflora extraction graph to obtain a wetland vegetation area and a spartina alterniflora area. According to the invention, the remote sensing image shot by the unmanned aerial vehicle is utilized to accurately and automatically identify the spartina parviflora distribution area of the wetland area.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental protection, and in particular to a method and system for extracting the spatial distribution of wetland vegetation based on remote sensing images. Background Art

[0002] Spartina is an excellent pioneer plant on the seashore. It is tolerant to flooding, salt and silt. It can form dense communities on the beach to protect the beach and shore, improve soil, and promote siltation and land reclamation. However, some of its populations spread through human activities or with the help of natural forces, and are easy to breed and spread in coastal wetlands to form dominant populations. Spartina alterniflora has a large area and high density. It competes with local plants on coastal beaches for growth space, crowds out other plants, and changes the original ecosystem of the beach. Severe invasion will cause a decrease in biodiversity and ecosystem degradation, destroy the habitats of offshore organisms such as invertebrates and various seabirds; reduce the mass of planktonic and benthic biomass, and suffocate and die coastal shellfish, crabs, algae, fish and other organisms; affect the exchange capacity of seawater, easily induce red tides, and deteriorate the water quality environment; compete for nutrients, causing the production of kelp and laver to decline year by year; occupy waterways, silt up ports, affect the departure of various ships, and bring inconvenience to marine fisheries and transportation industries. With the expansion of its population, Spartina alterniflora has spread to almost all coastal areas of my country, changing the local natural environment and ecosystem, and seriously threatening the ecological, economic and social security of the introduced areas.

[0003] By monitoring the distribution of Spartina alterniflora, we can provide a reference for the control of Spartina alterniflora and the restoration of coastal wetland ecosystems, and provide accurate and timely information support for the restoration and management of coastal wetlands. Therefore, accurate monitoring of the distribution of Spartina alterniflora is of great significance for the protection of coastal wetland ecosystems. However, since Spartina alterniflora grows in coastal tidal flats with low traffic accessibility and affected by tides, traditional manual survey methods are not easy to implement.

[0004] Therefore, how to provide a method and system for extracting the spatial distribution of wetland vegetation for accurately monitoring the distribution of Spartina alterniflora based on remote sensing images is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the invention

[0005] In view of the above research status and existing problems, the present invention provides a method and system for extracting the spatial distribution of wetland vegetation based on remote sensing images, which uses remote sensing images taken by unmanned aerial vehicles to accurately and automatically identify the distribution area of ​​Spartina odorifera in wetland areas.

[0006] The present invention provides a method for extracting the spatial distribution of wetland vegetation based on remote sensing images, comprising the following steps:

[0007] Model training steps:

[0008] Obtain remote sensing images of Spartina alterniflora and other wetland vegetation, label the samples of Spartina alterniflora in the remote sensing images, and establish a sample data set;

[0009] Using the sample data set to train a vector machine (SVM) to obtain a trained SVM supervised learning model;

[0010] Distribution extraction steps:

[0011] Use drones to collect multiple remote sensing images of wetland vegetation in wetland areas in real time;

[0012] The multiple remote sensing images are stitched together to obtain a DEM model and generate a complete orthophoto of the wetland area;

[0013] Inputting the complete orthophoto of the wetland area into the trained SVM supervised learning model to obtain a wetland vegetation extraction map and a Spartina alterniflora extraction map;

[0014] Area calculation is performed on the wetland vegetation extraction map and the Spartina alterniflora extraction map to obtain the wetland vegetation area and the Spartina alterniflora area.

[0015] Preferably, the real-time geographic location information of the UAV is obtained based on an RTK positioning algorithm, and multiple remote sensing images of wetland vegetation in the wetland area are collected in real time according to a predetermined sampling flight route.

[0016] Preferably, the steps of generating a complete orthophoto of the wetland area include:

[0017] Automatically extracting shooting routes and location information from the plurality of remote sensing images;

[0018] According to the shooting route and location information, the overlapping part of each remote sensing image is estimated and matching points are generated, the multiple remote sensing images are automatically aligned, and a sparse point cloud model is established;

[0019] A polygonal mesh model is generated based on the point cloud distribution in the sparse point cloud model, and the surface of the target is constructed between the points through the polygonal mesh model to obtain a DEM model of the area, and generate a complete orthophoto of the wetland area.

[0020] Preferably, the step of calculating the area of ​​the wetland vegetation extraction map and the Spartina alterniflora extraction map comprises:

[0021] Converting the RGB channels of the wetland vegetation extraction map and the Spartina alterniflora extraction map into PCT channels;

[0022] The pixel grid area calculation is performed on the converted wetland vegetation extraction map and the Spartina alterniflora extraction map in a single channel to obtain the wetland vegetation area and the Spartina alterniflora area.

[0023] Preferably, before extracting the wetland vegetation extraction map and the Spartina alterniflora extraction map, the following steps are also included:

[0024] According to the flight path of the UAV in the wetland area, the complete orthophoto of the wetland area is cut into a regular polygonal image that fits the flight path;

[0025] Performing coordinate system correction on the regular polygon image so that it satisfies a preset geographic coordinate system;

[0026] Perform atmospheric radiation correction on the regular polygonal image.

[0027] Preferably, the remote sensing image includes a visible light image and an infrared image.

[0028] Preferably, a normalized difference vegetation index is calculated for the wetland vegetation extraction image and the Spartina alterniflora extraction image extracted from the infrared image to obtain a vegetation coverage monitoring result.

[0029] The present invention also provides a wetland vegetation spatial distribution extraction system based on remote sensing images according to the wetland vegetation spatial distribution extraction method based on remote sensing images, comprising:

[0030] The sample database is used to store remote sensing images of Spartina alterniflora and other wetland vegetation, label the samples of Spartina alterniflora in the remote sensing images, and establish a sample data set;

[0031] A model training module is used to train a vector machine (SVM) using the sample data set to obtain a trained SVM supervised learning model;

[0032] UAVs are used to collect multiple remote sensing images of wetland vegetation in wetland areas in real time;

[0033] A stitching module is used to stitch the multiple remote sensing images to obtain a DEM model and generate a complete orthophoto of the wetland area;

[0034] A distribution extraction module, used for inputting the complete orthophoto of the wetland area into the trained SVM supervised learning model to obtain a wetland vegetation extraction map and a Spartina alterniflora extraction map;

[0035] The area calculation module is used to calculate the area of ​​the wetland vegetation extraction map and the Spartina alterniflora extraction map to obtain the wetland vegetation area and the Spartina alterniflora area.

[0036] Preferably, it also includes a positioning module for acquiring the real-time geographic location information of the drone based on an RTK positioning algorithm, and collecting multiple remote sensing images of wetland vegetation in the wetland area in real time according to a predetermined sampling flight route.

[0037] Preferably, the splicing module includes:

[0038] A position information extraction unit, used to automatically extract shooting routes and position information from the plurality of remote sensing images;

[0039] A sparse point cloud model building unit, used to estimate the overlapping part of each remote sensing image and generate matching points according to the shooting route and location information, automatically align the multiple remote sensing images, and build a sparse point cloud model;

[0040] The complete orthophoto generation unit is used to generate a polygonal mesh model based on the point cloud distribution in the sparse point cloud model, construct the surface of the target between each point through the polygonal mesh model, obtain the DEM model of the area, and generate a complete orthophoto of the wetland area.

[0041] Preferably, the area calculation module includes:

[0042] A channel conversion unit, used for converting the RGB channels of the wetland vegetation extraction image and the Spartina alterniflora extraction image into PCT channels;

[0043] The pixel calculation unit is used to calculate the pixel grid area of ​​the converted wetland vegetation extraction map and the Spartina alterniflora extraction map in a single channel to obtain the wetland vegetation area and the Spartina alterniflora area.

[0044] Preferably, a complete orthophoto preprocessing module is also included, including:

[0045] A clipping unit, used for clipping the complete orthophoto of the wetland area into a regular polygonal image that fits the route path according to the route path of the drone in the wetland area;

[0046] A coordinate correction unit, used for performing coordinate system correction on the regular polygon image so as to make it satisfy a preset geographic coordinate system;

[0047] The atmospheric radiation correction unit is used to perform atmospheric radiation correction on the regular polygon image.

[0048] Compared with the prior art, it has the following beneficial effects:

[0049] Compared with traditional satellite images, the drone images used in the present invention can provide more detailed surface features, which helps to more accurately identify and classify Spartina alterniflora and other vegetation types.

[0050] The infrared reflection characteristics of Spartina alterniflora vary at different growth stages, and infrared band images can reveal the physiological and biochemical characteristics of plants. Using this feature, infrared imaging technology can more accurately identify the growth status and distribution range of Spartina alterniflora.

[0051] In the process of processing large-scale high-resolution satellite images to identify Spartina alterniflora, SVM deep learning technology was introduced. By combining infrared images with visible light images and performing feature fusion through machine learning algorithms, the spectral characteristics of Spartina alterniflora can be more fully understood, thereby improving the accuracy of identification.

[0052] Visible light and multispectral images acquired by low-altitude drones, combined with NDVI index models, can effectively monitor and analyze the vegetation coverage of Spartina alterniflora. This not only improves the accuracy of monitoring, but also reduces manpower and material costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on the provided drawings without creative work.

[0054] Figure 1 A flow chart of a method for extracting wetland vegetation spatial distribution based on remote sensing images provided in an embodiment of the present invention;

[0055] Figure 2 A schematic diagram of a UAV flight mission area provided by an embodiment of the present invention;

[0056] Figure 3 A remote sensing image stitching flow chart provided by an embodiment of the present invention;

[0057] Figure 4 The remote sensing image sample data from the Spartina alterniflora / Phragmites australis intergrowth area provided by the embodiment of the present invention;

[0058] Figure 5 The image target sample extraction and area calculation process provided by the embodiment of the present invention;

[0059] Figure 6 A remote sensing image collected by a drone provided in an embodiment of the present invention;

[0060] Figure 7 A point cloud image constructed by image stitching provided in an embodiment of the present invention;

[0061] Figure 8 A schematic diagram of the distribution of wetland vegetation and Spartina alterniflora provided in an embodiment of the present invention;

[0062] The corresponding figure (a) is an orthophoto image of a drone provided by an embodiment of the present invention;

[0063] The corresponding figure (b) is a wetland vegetation extraction image provided by an embodiment of the present invention;

[0064] The corresponding figure (c) is an extracted image of Spartina alterniflora provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0065] 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.

[0066] The present invention establishes a complete set of processes from image capture to Spartina alterniflora identification, including data preprocessing, feature extraction, classifier training and result verification. It forms a rapid identification capability for the distribution of Spartina alterniflora in other areas. It builds an early warning and response mechanism so that timely measures can be taken in the early stage of the spread of Spartina alterniflora.

[0067] like Figure 1 As shown, the first aspect of the embodiment of the present invention discloses a method for extracting the spatial distribution of wetland vegetation based on remote sensing images, comprising the following steps:

[0068] Model training steps:

[0069] Obtain remote sensing images of Spartina alterniflora and other wetland vegetation, label the samples of Spartina alterniflora in the remote sensing images, and establish a sample data set; use the sample data set to train a support vector machine (SVM) to obtain a trained SVM supervised learning model;

[0070] Distribution extraction steps:

[0071] Use drones to collect multiple remote sensing images of wetland vegetation in the wetland area in real time; stitch multiple remote sensing images to obtain a DEM model, and generate a complete visible light orthophoto of the wetland area; input the complete visible light orthophoto of the wetland area into the trained SVM supervised learning model to obtain a wetland vegetation extraction map and a Spartina alterniflora extraction map; calculate the area of ​​the wetland vegetation extraction map and the Spartina alterniflora extraction map to obtain the wetland vegetation area and the Spartina alterniflora area.

[0072] In this embodiment, the drone is equipped with an all-round visual system and an infrared sensor system, which can stably hover and fly indoors and outdoors, and has automatic return and obstacle sensing functions. The drone can detect the surrounding aircraft and support password protection to ensure the security of equipment and data. The drone is equipped with an infrared camera and a visible light camera, which can capture thermal images and visible light images at the same time. The infrared camera can record videos with a resolution of 640×512. The visible light camera can record 4K videos and take 48-megapixel photos, and supports 32x digital zoom.

[0073] In one embodiment, the real-time geographic location information of the drone is obtained based on the RTK positioning algorithm to achieve high-precision positioning function, and multiple remote sensing images of wetland vegetation in the wetland area are collected in real time according to the predetermined sampling flight route. The RTK module is equipped for execution.

[0074] The RTK (Real Time Kinematic) positioning algorithm is a real-time dynamic positioning technology based on carrier phase observations. It can provide real-time three-dimensional positioning results of the measuring station in a specified coordinate system, provide centimeter-level positioning accuracy for drones, and reduce aerial photography and modeling errors.

[0075] This embodiment integrates satellite images, aerial photography images, geographic information systems (GIS) and key geographic data to enable users to browse high-definition images of all parts of the world in a virtual three-dimensional space. A variety of operations are supported in the three-dimensional space, such as searching for specific locations, viewing terrain and 3D buildings, exploring historical images, and the ocean and starry sky.

[0076] The preliminary preparation work of this embodiment also includes planning the flight route of the drone sampling:

[0077] Users can create complex map views, perform spatial analysis, and manage geographic databases. In the study of planning drone sampling flight routes, the intuitive interface and rich images of virtual 3D space can be used to quickly identify key features of the study area, such as terrain undulations, building distribution, and potential obstacles. The precise location of sampling points can be determined in the virtual 3D map, and the distance and height differences between points can be evaluated.

[0078] Conduct more in-depth route analysis based on multiple factors such as terrain, land use, and environmental protection areas. Optimize flight paths to ensure that drones follow the most economical flight routes while meeting sampling needs and avoid no-fly zones and high-risk areas. This embodiment can efficiently plan safe and efficient drone sampling flight routes. The feasibility of the route can be further ensured through simulated flight and real-time adjustments. Share geographic locations and paths in KML (Keyhole Markup Language) or KMZ file formats to facilitate the acquisition of drone flight routes.

[0079] like Figure 2As shown, the total area of ​​the study area of ​​this embodiment is 5210 hectares, and a total of 8 sampling areas with an area of ​​1023 hectares were flown in the study area. The image acquisition time is 8-12 am, sunny to cloudy weather, and the temperature is 20°C. The shooting route planning is set, the average route length is 22000m, the flight altitude is 120m, the flight speed is 8m / s, the photo mode is equidistant 20m, the GSD (Ground Sampling Distance) resolution is 3.67cm / pixel, the lateral overlap rate is 65%, and the heading overlap rate is 75%. After the setting is completed, the drone will automatically perform aerial photography according to the planned route. Each block takes about 40 minutes, and all eight areas have a total of 5435 corresponding images of visible light images and thermal infrared images.

[0080] like Figure 6 As shown, a total of 11,080 images were collected, half of which were visible light images and half were thermal imaging images.

[0081] In one embodiment, the steps of generating a complete visible light orthophoto of a wetland area include:

[0082] Automatically extract shooting routes and location information from multiple remote sensing images;

[0083] Estimate the overlapping parts of each remote sensing image and generate matching points based on the shooting route and location information, automatically align multiple remote sensing images, and establish a sparse point cloud model;

[0084] A polygonal mesh model is generated based on the point cloud distribution in the sparse point cloud model. The surface of the target is constructed between the points through the polygonal mesh model to obtain the DEM model of the area and generate a complete visible light orthophoto of the wetland area.

[0085] In this embodiment, Figure 3 As shown, the specific execution process is as follows:

[0086] Import photos: Import filtered remote sensing images, automatically extract location information from the images, and display the camera position and shooting route when the image was taken.

[0087] Photo alignment: After the remote sensing image is imported, "Align photos" is performed. According to the research needs, the image stitching accuracy and pairing mode can be automatically aligned, each image can be estimated, the overlapping parts in the image can be analyzed and matching points can be generated, and a sparse point cloud model can be established. Figure 7 shown.

[0088] Mesh generation: By calculating the dense point cloud, a polygonal mesh model is generated based on the dense point cloud distribution, and the surface of the target is constructed between the points through the polygonal mesh model. The mesh is automatically generated based on the source data, surface type, and number of generated faces of the mesh. At this point, the image stitching process is basically completed, and the model can be used to generate stitched orthophotos outside the visible light.

[0089] Generation and export of regional orthophotos and DEM: Set the projection type, coordinate system, resolution, regional boundaries and other parameters for generating orthophotos. After the settings are completed, the orthophotos after visible light image stitching can be generated. The image stitching results can be exported according to the required file format. The same steps are used to obtain the regional DEM image.

[0090] In one embodiment, the step of calculating the area of ​​the wetland vegetation extraction map and the Spartina alterniflora extraction map includes:

[0091] Convert the RGB channels of the wetland vegetation extraction map and the Spartina alterniflora extraction map into PCT channels;

[0092] The pixel grid area of ​​the converted wetland vegetation extraction map and Spartina alterniflora extraction map was calculated in a single channel to obtain the wetland vegetation area and Spartina alterniflora area.

[0093] In one embodiment, the remote sensing image includes a visible light image and an infrared image;

[0094] In one embodiment, a normalized difference vegetation index (NDVI index model) is calculated for a wetland vegetation extraction image and a Spartina alterniflora extraction image extracted from an infrared image to obtain a vegetation coverage monitoring result.

[0095] It should be noted that the calculation principle of the NDVI index model is: the difference between the reflectance value of the near-infrared band and the reflectance value of the red light band is ratioed to the sum of the two, that is:

[0096] NDVI = (NIR-R) / (NIR+R);

[0097] Among them, NIR is the reflection value of the near infrared band, and R is the reflection value of the red light band.

[0098] In one embodiment, before extracting the wetland vegetation extraction map and the Spartina alterniflora extraction map, the following steps are also included:

[0099] According to the flight path of the UAV in the wetland area, the complete visible light orthophoto of the wetland area is cut into a regular polygonal image that fits the flight path;

[0100] Correct the coordinate system of the regular polygon image to make it meet the preset geographic coordinate system;

[0101] Perform atmospheric radiation correction on regular polygon images.

[0102] In this embodiment, the photos with coordinates are combined into a whole RGB image through stitching. Each image is cut into regular polygons according to the route mission to facilitate the subsequent calculation and display.

[0103] The images are spatially corrected and all output images meet the WGS1984 geographic coordinate system.

[0104] Since the images collected by drones are low-altitude images, they are greatly affected by atmospheric effects and the solar altitude angle during shooting periods with bad weather conditions, so atmospheric radiation correction is performed in sequence. During a concentrated period of time in clear weather, the effects of atmospheric effects and the solar altitude angle are less and can be ignored, so no atmospheric radiation correction is required.

[0105] The specific process of extracting coastal wetland vegetation and Spartina alterniflora from the stitched images is as follows:

[0106] A high-definition camera was used to shoot RGB images. With the help of high-resolution and high-definition images, attempts were made to extract features of individual plants to establish a database. And a support vector machine (SVM) algorithm was used for distinction and extraction. Using the high-precision distribution data of Spartina alterniflora extracted from drone images, a more accurate training set can be constructed. These training sets will be used to optimize machine learning and deep learning algorithms to improve the accuracy of extracting Spartina alterniflora from large-scale, high-resolution satellite images. Through continuous iteration and optimization, the recognition ability of the algorithm can be gradually improved, and more efficient monitoring and governance can be achieved. Figure 5 As shown, it is mainly divided into the following steps:

[0107] (1) Feature extraction

[0108] according to Figure 4 The dataset was established by labeling the samples of Spartina alterniflora and Phragmites australis in the high-definition image. 20 samples were collected for each vegetation to increase the accuracy.

[0109] (2) Feature extraction

[0110] The above data set was used to train SVM. The maximum sample size parameter of each classification was set to 150, and the coastal wetland vegetation extraction map and the coastal wetland Spartina alterniflora extraction map were extracted by continuous classification, such as Figure 8 shown.

[0111] (3) Area calculation

[0112] The RGB channel of the remote sensing image is converted into a PCT channel so that the image area can be calculated on a single channel. Then the pixel level area calculation is performed to obtain the final coastal wetland vegetation and Spartina alterniflora area, as shown in Table 1.

[0113] Table 1 Area classification and calculation table of sampling area

[0114]

[0115] This embodiment successfully spliced ​​it into 8 larger images. These images provide a solid foundation for subsequent analysis. The spliced ​​images were further cut to obtain a sample area of ​​773 hectares in total. Based on the SVM supervised learning model, the vegetation distribution and Spartina alterniflora distribution of the coastal wetland were successfully extracted from the original images. The cross validation rate reached 0.9845 and 0.9650 respectively, showing a higher recognition accuracy.

[0116] The second aspect of the embodiment of the present invention provides a wetland vegetation spatial distribution extraction system based on remote sensing images according to the first aspect of the embodiment, which is used to execute all the steps in the first aspect of the embodiment. It includes: a sample database, which is used to store remote sensing images of Spartina alterniflora and other wetland vegetation, and to mark the sample labels of Spartina alterniflora in the remote sensing images to establish a sample data set; a model training module, which is used to train the vector machine SVM using the sample data set to obtain the trained SVM supervised learning model; a drone, which is used to collect multiple remote sensing images of wetland vegetation in the wetland area in real time; a splicing module, which is used to splice multiple remote sensing images to obtain a DEM model and generate a complete visible light orthophoto of the wetland area; a distribution extraction module, which is used to input the complete visible light orthophoto of the wetland area into the trained SVM supervised learning model to obtain a wetland vegetation extraction map and a Spartina alterniflora extraction map; an area calculation module, which is used to calculate the area of ​​the wetland vegetation extraction map and the Spartina alterniflora extraction map to obtain the wetland vegetation area and the Spartina alterniflora area.

[0117] In one embodiment, a positioning module is also included, which is used to obtain the real-time geographic location information of the drone based on the RTK positioning algorithm, and to collect multiple remote sensing images of wetland vegetation in the wetland area in real time according to a predetermined sampling flight route.

[0118] In one embodiment, the splicing module includes:

[0119] The position information extraction unit is used to automatically extract the shooting routes and position information from multiple remote sensing images; the sparse point cloud model construction unit is used to estimate the overlapping parts of each remote sensing image and generate matching points according to the shooting routes and position information, automatically align multiple remote sensing images, and establish a sparse point cloud model; the complete visible light orthophoto generation unit is used to generate a polygonal mesh model based on the point cloud distribution in the sparse point cloud model, construct the surface of the target between each point through the polygonal mesh model, obtain the DEM model of the area, and generate a complete visible light orthophoto of the wetland area.

[0120] In one embodiment, the area calculation module includes:

[0121] The channel conversion unit is used to convert the RGB channels of the wetland vegetation extraction map and the Spartina alterniflora extraction map into PCT channels; the pixel calculation unit is used to calculate the pixel point grid area of ​​the converted wetland vegetation extraction map and the Spartina alterniflora extraction map in a single channel to obtain the wetland vegetation area and the Spartina alterniflora area.

[0122] In one embodiment, a complete visible light orthophoto preprocessing module is also included, including:

[0123] The cropping unit is used to crop the complete visible light orthophoto of the wetland area into a regular polygonal image that fits the route of the drone in the wetland area; the coordinate correction unit is used to correct the coordinate system of the regular polygonal image so that it meets the preset geographic coordinate system; the atmospheric radiation correction unit is used to correct the atmospheric radiation of the regular polygonal image.

[0124] The above is a detailed introduction to a method and system for extracting the spatial distribution of wetland vegetation based on remote sensing images provided by the present invention. This article uses specific examples to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

[0125] In this article, relational terms such as first and second, etc. are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

Claims

1. A method and system for extracting the spatial distribution of wetland vegetation based on remote sensing images, characterized in that: The steps include: Model training steps: Obtain remote sensing images of Spartina alterniflora and other wetland vegetation, label the samples of Spartina alterniflora in the remote sensing images, and establish a sample data set; Using the sample data set to train a vector machine (SVM) to obtain a trained SVM supervised learning model; Distribution extraction steps: Use drones to collect multiple remote sensing images of wetland vegetation in wetland areas in real time; The multiple remote sensing images are stitched together to obtain a DEM model and generate a complete orthophoto of the wetland area; Inputting the complete orthophoto of the wetland area into the trained SVM supervised learning model to obtain a wetland vegetation extraction map and a Spartina alterniflora extraction map; Area calculation is performed on the wetland vegetation extraction map and the Spartina alterniflora extraction map to obtain the wetland vegetation area and the Spartina alterniflora area.

2. The method for extracting the spatial distribution of wetland vegetation based on remote sensing images according to claim 1, characterized in that: The real-time geographic location information of the UAV is obtained based on the RTK positioning algorithm, and multiple remote sensing images of wetland vegetation in the wetland area are collected in real time according to a predetermined sampling flight route.

3. The method for extracting the spatial distribution of wetland vegetation based on remote sensing images according to claim 1, characterized in that: The steps to generate a complete orthophoto of the wetland area include: Automatically extracting shooting routes and location information from the plurality of remote sensing images; According to the shooting route and location information, the overlapping part of each remote sensing image is estimated and matching points are generated, the multiple remote sensing images are automatically aligned, and a sparse point cloud model is established; A polygonal mesh model is generated based on the point cloud distribution in the sparse point cloud model, and the surface of the target is constructed between the points through the polygonal mesh model to obtain a DEM model of the area, and generate a complete orthophoto of the wetland area.

4. The method for extracting the spatial distribution of wetland vegetation based on remote sensing images according to claim 1, characterized in that: The step of calculating the area of ​​the wetland vegetation extraction map and the Spartina alterniflora extraction map comprises: Converting the RGB channels of the wetland vegetation extraction map and the Spartina alterniflora extraction map into PCT channels; The pixel grid area calculation is performed on the converted wetland vegetation extraction map and the Spartina alterniflora extraction map in a single channel to obtain the wetland vegetation area and the Spartina alterniflora area.

5. The method for extracting the spatial distribution of wetland vegetation based on remote sensing images according to claim 1, characterized in that: Before extracting the wetland vegetation extraction map and the Spartina alterniflora extraction map, the following steps are also included: According to the flight path of the UAV in the wetland area, the complete orthophoto of the wetland area is cut into a regular polygonal image that fits the flight path; Performing coordinate system correction on the regular polygon image so that it satisfies a preset geographic coordinate system; Perform atmospheric radiation correction on the regular polygonal image.

6. A wetland vegetation spatial distribution extraction system based on remote sensing images according to any one of claims 1 to 5, characterized in that: include: The sample database is used to store remote sensing images of Spartina alterniflora and other wetland vegetation, label the samples of Spartina alterniflora in the remote sensing images, and establish a sample data set; A model training module is used to train a vector machine (SVM) using the sample data set to obtain a trained SVM supervised learning model; UAVs are used to collect multiple remote sensing images of wetland vegetation in wetland areas in real time; A stitching module is used to stitch the multiple remote sensing images to obtain a DEM model and generate a complete orthophoto of the wetland area; A distribution extraction module, used for inputting the complete orthophoto of the wetland area into the trained SVM supervised learning model to obtain a wetland vegetation extraction map and a Spartina alterniflora extraction map; The area calculation module is used to calculate the area of ​​the wetland vegetation extraction map and the Spartina alterniflora extraction map to obtain the wetland vegetation area and the Spartina alterniflora area.

7. The wetland vegetation spatial distribution extraction system based on remote sensing images according to claim 6 is characterized in that: It also includes a positioning module, which is used to obtain the real-time geographic location information of the drone based on the RTK positioning algorithm, and to collect multiple remote sensing images of wetland vegetation in the wetland area in real time according to a predetermined sampling flight route.

8. The wetland vegetation spatial distribution extraction system based on remote sensing images according to claim 6 is characterized in that: The splicing module comprises: A position information extraction unit, used to automatically extract shooting routes and position information from the plurality of remote sensing images; A sparse point cloud model building unit, used to estimate the overlapping part of each remote sensing image and generate matching points according to the shooting route and location information, automatically align the multiple remote sensing images, and build a sparse point cloud model; The complete orthophoto generation unit is used to generate a polygonal mesh model based on the point cloud distribution in the sparse point cloud model, construct the surface of the target between each point through the polygonal mesh model, obtain the DEM model of the area, and generate a complete orthophoto of the wetland area.

9. The wetland vegetation spatial distribution extraction system based on remote sensing images according to claim 6, characterized in that: The area calculation module comprises: A channel conversion unit, used for converting the RGB channels of the wetland vegetation extraction image and the Spartina alterniflora extraction image into PCT channels; The pixel calculation unit is used to calculate the pixel grid area of ​​the converted wetland vegetation extraction map and the Spartina alterniflora extraction map in a single channel to obtain the wetland vegetation area and the Spartina alterniflora area.

10. The wetland vegetation spatial distribution extraction system based on remote sensing images according to claim 6, characterized in that: Also included is a complete orthophoto pre-processing module, including: A cropping unit, for cropping the complete orthophoto of the wetland area into a regular polygonal image that fits the route path according to the route path of the drone in the wetland area; A coordinate correction unit, used for performing coordinate system correction on the regular polygon image so as to make it satisfy a preset geographic coordinate system; The atmospheric radiation correction unit is used to perform atmospheric radiation correction on the regular polygon image.

Images