Wetland monitoring method and device based on combination of multi-source remote sensing and unmanned aerial vehicle, and medium

Abstract

The invention discloses a multi-source remote sensing and unmanned aerial vehicle combined wetland monitoring method, equipment and a medium, relates to the technical field of wetland monitoring, and realizes efficient and accurate wetland monitoring through combination of satellite remote sensing and unmanned aerial vehicle technologies. Firstly, space-time continuity and noise suppression preprocessing is carried out on a satellite remote sensing image, and the image quality is improved; thirdly, the image is optimized through segmentation fusion processing, and the species distribution range and the community boundary are accurately extracted; the unmanned aerial vehicle monitoring path is planned based on the data, and the monitoring efficiency is improved. A high-resolution multispectral image obtained by the unmanned aerial vehicle is subjected to radiation correction and geometric correction and then is fused with a satellite remote sensing image to generate a high-precision wetland species distribution diagram. According to the scheme, the advantages of large-range coverage of satellite remote sensing and high resolution of the unmanned aerial vehicle are fully played, the defects of data quality, information extraction and monitoring efficiency in a traditional method are overcome, and powerful support is provided for wetland protection and management.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wetland monitoring, in particular to a wetland monitoring method, device and medium combining multi-source remote sensing and unmanned aerial vehicles. BACKGROUND

[0002] Wetlands, as important ecosystems with unique ecological functions on earth, their ecological environment monitoring is crucial for the protection, management and sustainable development of wetlands. Wetlands are the habitat of many species, including fish, amphibians, aquatic plants and various water birds. Wetlands are often affected by human activities around them, such as agricultural non-point source pollution, industrial wastewater discharge, etc. Therefore, it is necessary to monitor the vegetation growth of wetlands, and further monitor the water body according to the species distribution of wetlands, which can effectively take preventive measures to ensure the ecological environment of wetlands.

[0003] Traditional wetland monitoring methods rely heavily on field surveys. Although this method can obtain more accurate local information, it has significant shortcomings such as low efficiency, high labor cost, limited monitoring range, etc., and is difficult to meet the dynamic monitoring needs of large-area wetlands over a long time series.

[0004] The emergence of satellite remote sensing technology has brought new opportunities for wetland monitoring. It can periodically obtain large-scale wetland images and provide macro information such as wetland land cover type, vegetation distribution profile, and water area change. However, the spatial resolution of satellite remote sensing data is relatively limited, and it is difficult to accurately depict the complex ecological structure, vegetation community details and small-scale ecological processes within the wetland.

[0005] Unmanned aerial vehicle remote sensing technology, with its high spatial resolution and flexible maneuverability, can obtain more detailed wetland image information in local areas, effectively making up for the shortcomings of satellite remote sensing in micro-level monitoring. However, due to factors such as endurance and flight range, the use of unmanned aerial vehicles alone cannot fully cover the monitoring of large-area wetlands.

[0006] In addition, existing wetland monitoring often lacks deep integration and collaborative analysis of ground field monitoring data and remote sensing data, resulting in incomplete, in-depth and accurate monitoring of wetland ecosystems. The rise of smart plot technology provides a solution to this problem. By setting up smart plots in wetlands, we can obtain real-time in-situ monitoring data such as weather, soil moisture, water quality, and biodiversity, providing more abundant and accurate ground truth information for wetland monitoring, and effectively complementing multi-source remote sensing data. SUMMARY

[0007] The technical problem to be solved by the present application is that the existing remote sensing image processing precision is low, the unmanned aerial vehicle flight monitoring path planning precision is low, which further leads to low collaborative analysis precision of the data of the two, and the purpose is to provide a kind of multi-source remote sensing combined with unmanned aerial vehicle wetland monitoring method, equipment and medium, through the combination of satellite remote sensing and unmanned aerial vehicle technology, the efficient and accurate wetland monitoring is realized. First, the satellite remote sensing image is preprocessed for spatio-temporal continuity and noise suppression, and the image quality is improved. Then, the image is optimized by segmentation fusion processing, and the species distribution range and community boundary are accurately extracted. Based on this data, the unmanned aerial vehicle monitoring path is planned to improve the monitoring efficiency. After the high-resolution multispectral image obtained by the unmanned aerial vehicle is radiometrically and geometrically corrected, it is fused with the satellite remote sensing image to generate a high-precision wetland species distribution map. The scheme fully utilizes the advantages of satellite remote sensing in wide coverage and unmanned aerial vehicle in high resolution, solves the problems of data quality, information extraction and monitoring efficiency in traditional methods, and provides strong support for wetland protection and management.

[0008] The present application is realized by the following technical solutions: The present application provides a kind of multi-source remote sensing combined with unmanned aerial vehicle wetland monitoring method in the first aspect, comprising the following specific steps: Based on satellite remote sensing, multi-band remote sensing image of target detection area is obtained; The multi-band remote sensing image of the target detection area is preprocessed to obtain the multi-band remote sensing image with spatio-temporal continuity and noise suppression; The multi-band remote sensing image with spatio-temporal continuity and noise suppression is subjected to segmentation fusion processing to obtain the optimized multi-band remote sensing image; According to the band information of the optimized multi-band remote sensing image, the spectral feature and spatial feature are extracted; Based on the spectral feature and spatial feature, the species distribution range and community boundary are extracted to obtain the community distribution data of the target detection area; Based on the community distribution data of the target detection area, the monitoring path of the unmanned aerial vehicle is planned; The target area is monitored using an unmanned aerial vehicle to obtain unmanned aerial vehicle multispectral image data, which contains multiple bands similar to the wavelength range of remote sensing image data; The unmanned aerial vehicle multispectral image data is subjected to radiometric correction and geometric correction to obtain unmanned aerial vehicle hyperspectral image data; The unmanned aerial vehicle hyperspectral image data is matched with the remote sensing image in terms of band, and the unmanned aerial vehicle hyperspectral image data and the remote sensing image are fused to identify the wetland distribution area and generate a wetland species distribution map.

[0009] Further, the pre-processing of the multi-band remote sensing image of the target detection area specifically includes: Interpolation operation and resampling processing are performed on the spatial features of the coarse resolution multi-band remote sensing image; The time series data of the coarse resolution image after interpolation operation and resampling processing are matched with the cloud-free high-resolution multi-band remote sensing image at the pixel level to obtain the corrected time series; The missing values in the original multi-band remote sensing image are filled by spatio-temporal modeling to restore the complete time series; The time series is smoothed by using a weighted seasonal-trend decomposition filter, and a multi-band remote sensing image with spatio-temporal continuity and noise suppression is output.

[0010] Further, the multi-band remote sensing image with spatio-temporal continuity and noise suppression is segmented and fused, specifically including: Randomly selecting a single pixel in the multi-band remote sensing image; In the target area, homogeneous and heterogeneous regions are segmented based on the pixel; The homogeneous regions are merged into a new object; The heterogeneous region is segmented, and the homogeneous regions in the heterogeneous region are merged into the new object until the segmentation of the heterogeneous region is completed, and the segmented multi-band remote sensing image is obtained; The segmented multi-band remote sensing image is integrated to obtain an optimized multi-band remote sensing image.

[0011] Further, according to the band information of the optimized multi-band remote sensing image, spectral features and spatial features are extracted, specifically including: The spectral reflectance curve of each pixel is extracted to obtain the spectral features reflecting the vegetation species; The spatial resolution of the multi-band remote sensing image is obtained, and the texture features are calculated to obtain the spatial features reflecting the vegetation community.

[0012] Further, based on the spectral features and spatial features, the species distribution range and community boundary are extracted to obtain the community distribution data of the target detection area, specifically including: According to the band information of the optimized multi-band remote sensing image, the vegetation index is calculated; Based on the spectral features and spatial features, the boundary information of each vegetation community unit is extracted by morphological method to obtain the species distribution range and community boundary; According to the vegetation index, species distribution range and community boundary, a threshold is set to separate the vegetation area from the non-vegetation area; The vegetation area is divided into multiple community units by using a region growing algorithm to obtain the community distribution data of the target detection area.

[0013] Further, based on the community distribution data of the target detection area, the monitoring path of the unmanned aerial vehicle is planned, specifically including: Based on the community distribution data of the target detection area, the vegetation community types that need to be monitored are determined according to the demand of the target detection area, and priorities are assigned to different communities; Obtain the performance parameters of the unmanned aerial vehicle, and divide the monitoring area in combination with the community priorities; Based on the terrain data of the monitoring area, the heading overlap and the lateral overlap are calculated; Based on the overlap requirement, an initial flight point distribution is automatically generated; According to the initial flight point distribution, the data of the range area is collected, and the overlap and the splicing quality of the data are analyzed; If the overlap and the splicing quality do not meet the set threshold, the flight point density is adjusted; According to the adjusted flight point density, the monitoring path of the unmanned aerial vehicle is generated.

[0014] Further, the multi-spectral image data of the unmanned aerial vehicle is radiometrically and geometrically corrected: The radiometric correction includes: The pixel brightness value data of the original image is converted into atmospheric outer surface reflectivity data, the radiation calibration parameters monitored by the sensor are extracted, and the radiation calibration data is obtained; Taking the radiation value of the calibration tarpaulin as the standard value, the reflection correction is performed on the radiation calibrated data, and the radiation value data is converted into reflectivity data; The geometric correction includes: A semiconductor photosensitive element is used for imaging to obtain spectral information on a space line; The entire space image and spectral data acquisition is completed by mechanical push scanning; The image data of the entire area obtained according to the monitoring path of the unmanned aerial vehicle is spliced in multiple rows, and the unmanned aerial vehicle hyperspectral image data is obtained.

[0015] Further, the unmanned aerial vehicle hyperspectral image data is matched with the remote sensing image, the unmanned aerial vehicle hyperspectral image data is fused with the remote sensing image, the wetland distribution area is identified, and the wetland species distribution map is generated, which specifically includes: Based on the band information of the multi-band remote sensing image, the band range corresponding to the hyperspectral data is determined; The spectral reflectivity data corresponding to the bands of the multi-band remote sensing image is extracted from the hyperspectral data, and the extracted spectral reflectivity data is consistent with the band information of the remote sensing image in wavelength range and resolution; According to the extracted spectral reflectivity information of the hyperspectral data, the same vegetation index as the remote sensing vegetation index is calculated; Obtain the vegetation index calculated according to the band reflectivity value of the multi-band remote sensing image; Fuse the calculated vegetation index of the unmanned aerial vehicle image with the vegetation index of the remote sensing data; Extract the fusion data features based on the fusion data; Build a classification model, classify the fusion data features, identify the wetland distribution area, and generate a wetland species distribution map.

[0016] The second aspect of the application provides an electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to realize a wetland monitoring method combining multi-source remote sensing and unmanned aerial vehicles.

[0017] The third aspect of the application provides a computer-readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to realize a wetland monitoring method combining multi-source remote sensing and unmanned aerial vehicles.

[0018] Compared with the prior art, the application has the following advantages and beneficial effects: 1. By combining satellite remote sensing and unmanned aerial vehicles, more abundant and accurate data can be obtained, thereby improving the accuracy of wetland monitoring. Satellite remote sensing can provide macro information on a large scale, while unmanned aerial vehicles can obtain high-resolution local information. The combination of the two can make up for the shortcomings of a single technology and achieve complementary advantages. 2. Unmanned aerial vehicles are fast and flexible, and can obtain multispectral image data of the target area in a short time. At the same time, reasonable unmanned aerial vehicle monitoring path planning can further improve the monitoring efficiency and reduce repeated monitoring and invalid monitoring. 3. Preprocessing and segmentation fusion processing of satellite remote sensing images can effectively restore missing values, smooth noise, and improve the spatiotemporal continuity of the data. 4. Extracting spectral features and spatial features can provide more comprehensive information for wetland monitoring. Spectral features can reflect the types and health status of vegetation, and spatial features can reveal the distribution pattern and structural characteristics of vegetation communities. 5. Through data fusion and classification model construction, a high-precision wetland species distribution map can be generated. The distribution map can clearly show the distribution range and boundary of different species in the wetland, providing strong support for the protection and management of wetland biodiversity. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the example embodiments of the application, the following will briefly introduce the drawings needed in the examples. It should be understood that the following drawings only show some embodiments of the application, and therefore should not be considered as limiting the scope. For those skilled in the art, other related drawings can also be obtained without creative labor. In the drawings: Figure 1 A monitoring flowchart provided for the embodiment of the present application; Figure 2 A land cover type distribution map provided for the embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to make the objectives, technical solutions and advantages of the present application clearer, further detailed description will be made to the present application in combination with embodiments and drawings, and the illustrative embodiments of the present application and the description thereof are only used to explain the present application, but not to limit the present application.

[0021] As a possible implementation, as shown in Figure 1 The embodiment provides a wetland monitoring method combining multi-source remote sensing and unmanned aerial vehicles, and the method comprises the following specific steps: acquiring multi-band remote sensing images of a target detection area based on satellite remote sensing; performing pretreatment on the multi-band remote sensing images of the target detection area to obtain multi-band remote sensing images that are continuous in time and space and have noise suppressed; performing segmentation fusion processing on the multi-band remote sensing images that are continuous in time and space and have noise suppressed to obtain optimized multi-band remote sensing images; extracting spectral features and spatial features according to band information of the optimized multi-band remote sensing images; extracting species distribution ranges and community boundaries based on the spectral features and the spatial features to obtain community distribution data of the target detection area; planning a monitoring path of an unmanned aerial vehicle based on the community distribution data of the target detection area; monitoring the target area using the unmanned aerial vehicle to acquire unmanned aerial vehicle multi-spectral image data, wherein the unmanned aerial vehicle multi-spectral image data comprises multiple bands similar to a wavelength range of the remote sensing image data; performing radiation correction and geometric correction on the unmanned aerial vehicle multi-spectral image data to obtain unmanned aerial vehicle hyperspectral image data; performing band matching on the unmanned aerial vehicle hyperspectral image data and the remote sensing images, fusing the unmanned aerial vehicle hyperspectral image data and the remote sensing images, identifying a wetland distribution area, and generating a wetland species distribution map.

[0022] The embodiment realizes efficient and accurate wetland monitoring by combining satellite remote sensing and unmanned aerial vehicle technology. Firstly, the satellite remote sensing images are pretreated for time and space continuity and noise suppression to improve image quality. Then, the images are optimized by segmentation fusion processing to accurately extract species distribution ranges and community boundaries. Based on this data, the monitoring path of the unmanned aerial vehicle is planned to improve the monitoring efficiency. The high-resolution multi-spectral images acquired by the unmanned aerial vehicle are corrected for radiation and geometry, and then fused with the satellite remote sensing images to generate a high-precision wetland species distribution map. This scheme fully utilizes the advantages of wide coverage of satellite remote sensing and high resolution of unmanned aerial vehicles, and solves the problems of data quality, information extraction and monitoring efficiency in traditional methods, thereby providing strong support for wetland protection and management.

[0023] In this embodiment, satellite remote sensing images will be disturbed by various noises during acquisition, such as sensor noise, atmospheric noise, etc. These noises will reduce the quality of the image and affect subsequent analysis and application. Through the preprocessing of spatio-temporal continuity and noise suppression, these noises can be effectively reduced and the quality of the image can be improved. At the same time, the multi-spectral image data of the unmanned aerial vehicle will also be affected by radiation and geometric distortion during acquisition. Through radiation correction and geometric correction, these distortions can be eliminated and the accuracy of the image can be improved. Traditional remote sensing image analysis methods may be limited by image resolution and noise when extracting species distribution range and community boundary. Through segmentation fusion processing and feature extraction, vegetation distribution and community boundary can be more accurately identified, and the accuracy of the analysis can be improved. When using satellite remote sensing images or unmanned aerial vehicle images alone to identify wetland distribution areas, the identification accuracy may be affected due to insufficient data resolution or spectral information. By fusing unmanned aerial vehicle hyperspectral image data with satellite remote sensing images, the advantages of both can be fully utilized to improve the accuracy of wetland distribution area identification. Satellite remote sensing images and unmanned aerial vehicle images have different advantages: satellite images have a wide coverage but relatively low resolution; unmanned aerial vehicle images have high resolution but limited coverage. By combining the two, large-scale monitoring and high-resolution monitoring can be combined to improve monitoring efficiency. The planning of unmanned aerial vehicle monitoring paths needs to be based on accurate community distribution data. By extracting community distribution data from satellite remote sensing images, scientific basis can be provided for the planning of unmanned aerial vehicle monitoring paths, and the targeting and efficiency of monitoring can be improved. Therefore, through the preprocessing of spatio-temporal continuity and noise suppression, the multi-band remote sensing images obtained have consistency in time and space, and the noise level is significantly reduced, providing a high-quality data basis for subsequent analysis. Moreover, after radiation correction and geometric correction of the multi-spectral image data of the unmanned aerial vehicle, the radiation accuracy and geometric accuracy are significantly improved, which can more accurately reflect the spectral and spatial characteristics of ground objects. Through segmentation fusion processing and feature extraction, species distribution range and community boundary can be more accurately extracted, providing more accurate data support for ecological monitoring and protection. By fusing unmanned aerial vehicle hyperspectral image data with satellite remote sensing images, wetland distribution areas can be more accurately identified, and high-precision wetland species distribution maps can be generated. By planning unmanned aerial vehicle monitoring paths based on community distribution data extracted from satellite remote sensing images, the efficiency and targeting of unmanned aerial vehicle monitoring can be improved, and unnecessary monitoring areas can be reduced. By matching and fusing satellite remote sensing images and unmanned aerial vehicle images, large-scale monitoring and high-resolution monitoring are combined, and the overall efficiency and data integration capability of monitoring are improved.

[0024] In some possible implementation manners, a multi-band remote sensing image of a target detection area is acquired based on satellite remote sensing. The multi-band remote sensing image can provide rich spectral information, which is helpful to distinguish different types of vegetation and ground objects. The multi-band remote sensing image of the target detection area is preprocessed, specifically including the following steps. Interpolation operation and resampling processing are performed on the spatial features of the coarse-resolution multi-band remote sensing image. The time series data of the coarse-resolution image after the interpolation operation and resampling processing are matched with the cloud-free high-resolution multi-band remote sensing image at the pixel level to form a corrected time series. The corrected time series is formed based on similarity weighted fusion. The missing values in the original multi-band remote sensing image are filled by spatiotemporal modeling to restore a complete time series. According to the selected modeling method, a spatiotemporal model is constructed by using known spatiotemporal data. The missing values in the original multi-band remote sensing image are estimated by using the constructed spatiotemporal model. For each missing pixel point, the spectral value of the missing pixel point is calculated by using the spatiotemporal information of the surrounding known pixel points, so that a complete time series image is restored.

[0025] A weighted seasonal-trend decomposition filter is a filter used for time series analysis, which decomposes a time series into trend, seasonality, and random components. This decomposition helps to understand the structure of the time series and can be used for tasks such as prediction and anomaly detection. Therefore, the weighted seasonal-trend decomposition filter is used in this embodiment to smooth the time series, output a multi-band remote sensing image with continuous spatiotemporal and noise suppression, and determine the weight distribution principle of the weighted filter according to the characteristics of the time series data and the analysis requirements. The weighted filter is designed according to the determined weight distribution principle. The designed weighted filter is applied to the decomposed trend component, seasonal component, and residual. The trend decomposition is first performed on the time series data of each multi-band remote sensing image. The trend decomposition decomposes the time series into trend component, seasonal component, and residual component. The trend component reflects the long-term trend of the time series, the seasonal component represents the periodic variation law of the time series, and the residual component is the random fluctuation part after removing the trend and seasonality. Through this decomposition, different change characteristics in the time series can be clearly separated.

[0026] In some possible implementations, the time series is smoothed using a weighted seasonal-trend decomposition filter, specifically including: first, a trend component of the time series is extracted using a moving average or other smoothing technique. The moving average can be a simple moving average (SMA) or a weighted moving average (WMA), in which weights can be used to emphasize data at certain time points. The trend component is subtracted from the original time series to obtain a detrended time series. Then, a seasonal component is extracted by calculating the average value of the detrended time series within each seasonal period. For example, for monthly data, the average value of each month can be calculated. The seasonal component is subtracted from the detrended time series to obtain a random component. The random component is the part of the time series that cannot be explained by the trend and seasonality. In extracting the trend and seasonal components, a weighted method can be used to emphasize data at certain time points. For example, a weighted moving average can be used to extract the trend component, in which weights can be used to emphasize recent data points.

[0027] In some possible implementations, the time series data of the coarse resolution image after interpolation operation and resampling processing is pixel-level matched with the cloud-free high-resolution multi-band remote sensing image to form a corrected time series, specifically including: taking the cloud-free high-resolution multi-band remote sensing image as a registration reference image, because it has a high spatial resolution and clear texture features, it can provide accurate spatial reference for the coarse resolution image; extracting feature points in the reference image and the coarse resolution image after interpolation operation and resampling processing; according to the matching result of the feature points, calculating the geometric transformation parameters of the coarse resolution image relative to the reference image. Then, the coarse resolution image is geometrically transformed and resampled again to align it with the reference image in space, realizing pixel-level matching. In the pixel-level matching process, for each pixel point, the similarity of the pixel point in the coarse resolution image and the high-resolution image is calculated, the higher the similarity, the closer the features of the pixel point in the two images, and the greater the corresponding weight. According to the calculated similarity weight, the pixel values of the coarse resolution image and the high-resolution image are weighted and fused. For each pixel point, the fused value is the weighted sum of the values of the corresponding pixel points of the coarse resolution image and the high-resolution image multiplied by their similarity weights. The fused image can integrate the advantages of the two images, retaining the spatial details of the high-resolution image and utilizing the spectral information of the coarse resolution image to form the corrected time series image.

[0028] In some possible embodiments, the multi-band remote sensing image after space-time continuity and noise suppression is subjected to segmentation fusion processing to obtain an optimized multi-band remote sensing image. The segmentation fusion processing process is as follows: a single pixel in the multi-band remote sensing image is randomly selected from the multi-band remote sensing image, the position (row and column number) and spectral characteristics (reflectivity value of each band) of the pixel are recorded, and the pixel in the neighborhood (usually 4-neighborhood or 8-neighborhood) around the randomly selected starting pixel is searched in the target area. The spectral difference between each pixel in the neighborhood and the starting pixel is calculated. For each band, the difference between the reflectivity of the two pixels is calculated, and then the difference values of all bands are weighted and summed. If the spectral difference between a pixel in the neighborhood and the starting pixel is less than a set threshold, the pixel is divided into a homogeneous region; otherwise, it is divided into a heterogeneous region. The homogeneous regions are merged into a new object, the heterogeneous regions are segmented, and the homogeneous regions in the heterogeneous regions are merged into the new object until the segmentation of the heterogeneous regions is completed, and the segmented multi-band remote sensing image is obtained. Finally, the segmented multi-band remote sensing image is integrated to obtain an optimized multi-band remote sensing image. According to the above segmentation and merging operations, a new image is generated, in which the value of each pixel corresponds to the object number to which it belongs. In this way, the image is divided into multiple object regions with similar spectral characteristics. Through the above segmentation fusion processing, the homogeneous regions in the image are effectively merged, and the heterogeneous regions are reasonably segmented, thereby improving the readability and analysis value of the image while retaining important information. This method is particularly suitable for remote sensing applications that require feature classification, change detection or landscape analysis.

[0029] wherein the homogeneous region refers to a region with similarity in spectral characteristics (such as reflectivity of each band) and spatial position. The similarity between pixels can be determined by setting a threshold (such as a spectral difference threshold). If the reflectivity difference between two pixels in each band is less than the threshold, they are considered to belong to a homogeneous region. The heterogeneous region refers to a region with large differences in spectral characteristics or spatial position, i.e., a region that does not meet the homogeneity condition.

[0030] In some possible embodiments, spectral features and spatial features are extracted according to the band information of the optimized multi-band remote sensing image, specifically including: The information of all bands is obtained from the multi-band remote sensing image, including the center wavelength of the band, the band range, etc. For example, a common multi-band image may include a blue band (0.45-0.52 microns), a green band (0.52-0.60 microns), a red band (0.60-0.70 microns) and a near-infrared band (0.70-0.90 microns); The spectral reflectance curve of each pixel is extracted to obtain the spectral characteristics of the vegetation species. The spectral reflectance curve is a set of reflectance values of each pixel at different wavebands, which can reflect the spectral characteristics of the vegetation species. The spatial resolution of the multi-band remote sensing image is obtained, and the texture features are calculated to obtain the spatial characteristics of the vegetation community. The spatial resolution refers to the actual ground area represented by each pixel in the remote sensing image, which determines the minimum ground detail that the image can distinguish. The texture features reflect the spatial distribution characteristics of the vegetation community.

[0031] In some possible embodiments, based on the spectral characteristics and the spatial characteristics, the species distribution range and the community boundary are extracted to obtain the community distribution data of the target detection area, specifically including: According to the band information of the optimized multi-band remote sensing image, the vegetation index is calculated. Based on the spectral characteristics and the spatial characteristics, the boundary information of each vegetation community unit is extracted using morphological operations to obtain the species distribution range and the community boundary. The morphological dilation operation is used to fill small holes in the vegetation area to enhance the connectivity of the vegetation area. The morphological erosion operation is used to refine the boundary of the vegetation area to remove noise on the boundary. The combination of dilation and erosion is used to further optimize the boundary of the vegetation area. The edge detection algorithm is used to extract the boundary of the vegetation area to obtain the boundary information of the vegetation community unit. According to the vegetation index, the species distribution range and the community boundary, the threshold is set to separate the vegetation area from the non-vegetation area. The vegetation area is divided into multiple community units using the region growing algorithm to obtain the community distribution data of the target detection area. Specifically, multiple seed points are randomly selected in the vegetation area, and the growth conditions such as spectral similarity (the difference in vegetation index is less than a certain threshold) are set. Starting from the seed point, the pixels in the neighborhood of the pixel are checked pixel by pixel to see if they meet the growth conditions. If they meet, the pixel is merged into the current growing region, and the pixel is used as a new seed point for further growth. Repeat the above process until no new pixel can be merged into the current region. The generated multiple regions are merged and optimized to remove small regions or merge adjacent similar regions. The divided vegetation community units are labeled as different categories to generate a community distribution map.

[0032] In some possible embodiments, based on the community distribution data of the target detection area, the monitoring path of the unmanned aerial vehicle is planned, specifically including: Based on the community distribution data of the target detection area, the vegetation community types that need to be monitored are determined according to the requirements of the target detection area, and different communities are assigned priorities. The performance parameters of the UAV are obtained, including flight speed, endurance time, camera resolution, sensor type, maximum flight height, etc. These parameters can be obtained from the technical manual of the UAV or the data provided by the manufacturer. According to the priority of the community, the monitoring area is divided into multiple monitoring areas according to the distribution range and priority of the vegetation community. The high-priority community area can be divided into smaller monitoring units to ensure more detailed monitoring; the low-priority area can be divided into larger monitoring units to improve monitoring efficiency; Based on the terrain data of the monitoring area, the heading overlap and the lateral overlap are calculated; the heading overlap is the proportion of the overlapping part of the images on the adjacent flight lines to a single image, and the lateral overlap is the proportion of the overlapping part of the images on the adjacent flight lines to a single image According to the calculated heading overlap and lateral overlap, the initial waypoint distribution is automatically generated. The waypoints should be evenly distributed in the monitoring area to ensure that the distance between adjacent waypoints meets the overlap requirement; According to the generated initial waypoint distribution, the UAV flight task is executed to collect data in the monitoring area. Ensure that the UAV flies according to the predetermined waypoints and takes pictures or videos at each waypoint, collect data in the range area according to the initial waypoint distribution, and analyze the overlap and splicing quality of the data; If the overlap and splicing quality do not meet the set threshold, adjust the waypoint density; According to the adjusted waypoint density, the monitoring path of the UAV is generated.

[0033] In some possible embodiments, the multi-spectral image data of the UAV is subjected to radiation correction and geometric correction: Radiation correction includes: Convert the pixel brightness value data of the original image to the surface reflectivity data outside the atmosphere, extract the radiation calibration parameters monitored by the sensor, and obtain the radiation calibration data.

[0034] Take the radiation value of the calibrated tarpaulin as the standard value, and perform reflectance correction on the radiation calibrated data to convert the radiation value data to reflectance data; Geometric correction includes: Use a semiconductor photosensitive element to image and obtain spectral information on a space line; Complete the acquisition of the entire space image and spectral data by mechanical push scanning; The image data of the entire area obtained according to the monitoring path of the UAV is subjected to multi-row band splicing to obtain UAV hyperspectral image data.

[0035] Through radiometric correction and geometric correction, the quality and usability of the unmanned aerial vehicle multi-spectral image data can be significantly improved. Radiometric correction can eliminate the influence of the sensor and atmospheric conditions to obtain accurate reflectance data; geometric correction can correct the geometric distortion of the image to ensure the spatial consistency between images. These corrected data can provide reliable data support for vegetation monitoring, land use analysis and other applications.

[0036] In some possible embodiments, the corrected image data is saved in a standard image format, and quality inspection is performed on the corrected image data, including the accuracy of the reflectance data and the precision of the geometric correction. The quality of the correction result can be evaluated by comparing with reference data (such as ground measured data and high-resolution images).

[0037] In some possible embodiments, the unmanned aerial vehicle hyperspectral image data is band-matched with the remote sensing image, the unmanned aerial vehicle hyperspectral image data is fused with the remote sensing image, the wetland distribution area is identified, and a wetland species distribution map is generated, specifically including: Based on the band information of the multi-band remote sensing image, the center wavelength and the band range of the band of the multi-band remote sensing image are obtained from the metadata of the multi-band remote sensing image, the center wavelength and the band range of the band of the unmanned aerial vehicle hyperspectral image are obtained from the metadata of the unmanned aerial vehicle hyperspectral image, the band ranges of the hyperspectral image and the remote sensing image are compared, the band range in the hyperspectral image closest to the band of the remote sensing image is determined, and the band range corresponding to the band information of the multi-band remote sensing image in the unmanned aerial vehicle hyperspectral image data is obtained; If the spectral resolution of the hyperspectral image is higher than that of the remote sensing image, the hyperspectral data needs to be spectrally resampled to make the spectral resolution consistent with that of the remote sensing image, that is, the spectral reflectance data corresponding to the band of the multi-band remote sensing image is extracted from the hyperspectral data, and the extracted spectral reflectance data is consistent with the band information of the remote sensing image in terms of wavelength range and resolution; According to the extracted spectral reflectance information of the hyperspectral data, the same vegetation index as the remote sensing vegetation index is calculated; The vegetation index calculated according to the band reflectance value of the multi-band remote sensing image is obtained; The calculated vegetation index of the unmanned aerial vehicle image is fused with the vegetation index of the remote sensing data; Based on the fused data, the fused data features are extracted; A classification model is constructed to classify the fused data features. Through the steps of band matching, vegetation index calculation and fusion, feature extraction and classification, the unmanned aerial vehicle hyperspectral image data can be effectively fused with the remote sensing image, the wetland distribution area can be identified, and a wetland species distribution map can be generated.

[0038] In some possible implementations, a multi-source remote sensing combined with unmanned aerial vehicle wetland monitoring method can be directly applied to various wetland monitoring projects, whether it is freshwater wetland, coastal wetland or inland swamp wetland, etc. As long as it involves monitoring the needs of various characteristics of the wetland ecosystem (land cover, vegetation, water quality, biodiversity, etc.), the multi-source remote sensing data and intelligent sample plot collaborative monitoring and evaluation method of the present embodiment can be used. For example, in large river wetland monitoring, satellite remote sensing data can be used to monitor the distribution and changes of the entire river basin wetland, unmanned aerial vehicles can be used to conduct fine detection of key areas such as river beaches and river bars in river wetlands, and intelligent sample plots can be set up in different ecological function zones of river wetlands (such as estuary wetlands and floodplain wetlands) to monitor water quality, soil and biodiversity changes in real time. Through multi-source data fusion analysis, the health status and dynamic changes of the river wetland ecosystem can be comprehensively mastered.

[0039] Further, it can be applied to the fields of wetland ecological restoration effect evaluation and wetland ecological compensation mechanism research. In a wetland ecological restoration project, through long-term multi-source data monitoring and analysis, the changes in various indicators of the wetland ecosystem before and after restoration, such as vegetation coverage increase, water quality improvement, biodiversity recovery, etc. can be compared to accurately evaluate the effectiveness of the ecological restoration project and provide scientific basis for optimization of the restoration scheme and decision-making of capital investment. In the research of wetland ecological compensation mechanism, the monitoring method of the present embodiment is used to obtain data related to the value of wetland ecosystem services (such as water conservation, climate regulation, biodiversity maintenance, etc.), which provides data support for reasonably determining the ecological compensation standard and range.

[0040] Specifically, the present embodiment can be directly applied to water quality ecological monitoring and evaluation of wetlands. When monitoring the water quality of wetlands, by combining satellite remote sensing and unmanned aerial vehicle technology, not only the efficiency and accuracy of wetland monitoring are achieved, but also the monitoring range is further expanded to the field of water quality ecological monitoring, forming a comprehensive and efficient wetland and water quality ecological comprehensive monitoring scheme, which provides more powerful support for wetland protection and management: In the aspect of wetland monitoring, firstly, the satellite remote sensing image is preprocessed for spatiotemporal continuity and noise suppression. This process effectively improves the quality of the image through advanced algorithms and technical means, making the wetland information in the image clearer and more accurate. Subsequently, the preprocessed image is optimized using segmentation fusion processing to accurately extract the distribution range and community boundary of the wetland community. This allows monitoring personnel to accurately locate the key areas of the wetland and carry out targeted monitoring work.

[0041] Based on the accurately extracted wetland species and water source distribution range and boundary data, the monitoring path of the unmanned aerial vehicle is further planned. The unmanned aerial vehicle can quickly and efficiently monitor the wetland according to the planned path, greatly improving the monitoring efficiency. The high-resolution multispectral images obtained by the unmanned aerial vehicle during the monitoring process are fused with satellite remote sensing images after radiation correction and geometric correction. This fusion fully takes advantage of the wide coverage of satellite remote sensing and the high resolution of unmanned aerial vehicles, and finally generates a high-precision wetland species and water source distribution map. The wetland species and water source distribution map can clearly show the distribution of different species and water sources in the wetland.

[0042] In terms of water quality ecological monitoring, satellite remote sensing can analyze the spectral characteristics of water bodies to preliminarily judge the water quality, such as identifying the distribution of suspended solids, algae and other substances in the water, so as to roughly understand the eutrophication degree of the water body and other information. According to the eutrophication degree of the water body, the ecological environment of the water quality (such as the pollution state of the water quality and the growth of plants and animals) can be further analyzed. The unmanned aerial vehicle can carry professional water quality monitoring equipment and penetrate into different areas of the wetland to obtain more accurate water quality data, including water temperature, pH value, dissolved oxygen, chemical oxygen demand (COD), ammonia nitrogen and other key water quality indicators. Through the flexible flight and precise positioning of the unmanned aerial vehicle, all-around and multi-level monitoring of the wetland water body can be realized, making up for the shortcomings of traditional water quality monitoring methods in spatial coverage and monitoring frequency. At the same time, according to the spectral image of the remote sensing image, the water source distribution data can be further monitored after obtaining the water source distribution range and boundary, and the object distribution data of the water surface can be obtained, so as to further judge the species distribution and garbage distribution of the water source.

[0043] Specifically, specific band combinations of multispectral unmanned aerial vehicle images can be used to extract water quality parameter information, such as analyzing the spectral characteristics of near-infrared and red light bands to estimate the chlorophyll content and suspended solids concentration in the water, so as to accurately evaluate the nutritional status and pollution degree of the wetland water body. For small wetland patches or special topographic features (such as oxbow lakes, river islands, etc.), multispectral unmanned aerial vehicles can perform fine boundary extraction and morphological feature analysis due to their high-resolution advantage, obtaining detailed information such as area, perimeter, shape index, etc. These microscopic information is of great significance for in-depth understanding of the ecological structure and ecological process inside the wetland. The combination of wetland monitoring and water quality ecological monitoring forms a complete wetland and water quality ecological comprehensive monitoring system.

[0044] In practical applications, by preprocessing and segmentation fusion processing of satellite remote sensing images, the key areas and species distribution of wetlands are determined, providing key monitoring targets and areas for water quality ecological monitoring. Unmanned aerial vehicles then conduct targeted water quality monitoring based on this information, obtaining detailed water quality data. At the same time, combining water quality monitoring results with wetland species distribution maps can more comprehensively analyze the health status of the wetland ecosystem. For example, by comparing water quality data and species distribution in different areas, the impact of water quality changes on wetland biodiversity can be found, providing scientific basis for the development of wetland ecological restoration and protection measures. This comprehensive monitoring scheme fully leverages the advantages of satellite remote sensing in large-scale coverage and unmanned aerial vehicles in high resolution and flexibility, effectively addressing issues such as low data quality, inaccurate information extraction, and low monitoring efficiency in traditional monitoring methods. It not only provides precise data support for wetland protection and management, but also timely discovers the vegetation and water quality conditions in the wetland ecosystem, providing strong support for the sustainable development and protection of the wetland ecosystem.

[0045] As a possible implementation, satellite remote sensing data, with its macroscopic and periodic characteristics, enables continuous monitoring of large areas of wetlands. For example, by regularly obtaining multispectral and hyperspectral images from Beijing-3 series or Resource-1 series satellites, the distribution pattern of wetland land cover types in large-scale space and their dynamic changes over time can be clearly understood. For example, in the comparison of satellite images over the past few years, the mutual transformation between marsh areas and forest areas in wetlands can be visually observed, providing long-term basic data support for wetland land use planning and ecological protection. Multispectral unmanned aerial vehicle data, with its high spatial resolution and rich spectral information, can finely detect key areas within wetlands. In terms of wetland vegetation community structure analysis, multiple specific spectral bands (such as blue, green, red, and near-infrared bands) of multispectral unmanned aerial vehicle images can capture subtle differences in spectral reflectance of different vegetation species. Through object-oriented classification methods, the distribution range and community boundaries of various wetland vegetation can be accurately identified. For example, reeds and cattails, two common wetland plants, have different reflectance in the near-infrared band. Using multispectral unmanned aerial vehicle images, their growth areas can be clearly divided, and their distribution patterns and ecological functions in the wetland ecosystem can be studied in detail.

[0046] In some possible embodiments, in terms of monitoring result visualization, the land cover type distribution, vegetation growth condition, water level change, and ecosystem health assessment result of the wetland are superimposed on the map in the form of thematic layers by means of a geographic information system (GIS) platform, to form an intuitive wetland monitoring thematic map. This enables the wetland management department, researchers, and other users to intuitively understand the overall spatial pattern of the wetland and the distribution characteristics of various ecological elements. For example, through the land cover type layer marked by different colors, the distribution range and mutual relationship of different land use types in the wetland can be quickly identified; according to the vegetation index NDVI rendering effect of the vegetation growth condition layer, the distribution of good and poor growth areas of the wetland vegetation and its relationship with the surrounding environment can be intuitively judged; according to the water level change layer, the fluctuation of the wetland water level in different seasons and years and its influence range on the wetland ecosystem can be clearly seen. At the same time, by drawing various charts such as line charts, column charts, and scatter charts, the change trend of various monitoring indexes of the wetland over time can be displayed, such as the annual change curve of the wetland area, which can clearly present the increase and decrease rate and time node of the wetland area, the seasonal change column chart of the vegetation index, which can reflect the seasonal regularity of the vegetation growth, and the time series scatter chart of the water quality parameters, which can help analyze the relationship between water quality change and time, season, and external interference factors (such as precipitation, surrounding human activities, etc.). These charts provide a powerful tool for in-depth analysis of the dynamic change law of the wetland ecosystem. For key areas or typical wetland landforms, three-dimensional models are constructed using unmanned aerial vehicle images for stereoscopic display, such as the three-dimensional model of the kettle lake in the wetland, which can observe the topographic and geomorphic features, surrounding vegetation distribution details, and water body shape change in detail, providing an immersive visualization tool for in-depth study of the microstructure and ecological function of the wetland, greatly improving the readability and usability of the wetland monitoring data, and facilitating efficient data analysis, decision making, and scientific research by relevant personnel.

[0047] As a possible embodiment, the present embodiment provides a wetland monitoring of a certain national nature reserve, as shown in the following table: Figure 2 As shown in the following table, in a certain national nature reserve wetland, according to the ecological function zoning and topographic and geomorphic features of the wetland, five intelligent sample plot positions are selected, which are marked as A, B, C, D, and E, and are evenly distributed in the core area, buffer zone, and experimental area of the wetland. Complete monitoring equipment is installed in each intelligent sample plot for monitoring.

[0048] According to the monitoring needs and geographical location of the wetland, ZY1E satellite and BJ3N satellite are determined as the main satellite remote sensing data sources. The orbit characteristics of ZY1E satellite enable it to cover the target wetland once every 16 days. Its multi-spectral band spatial resolution is 10 meters, the panchromatic band can reach 2.5 meters, and it has rich spectral band information, covering visible light, near-infrared and other bands, which can be used for monitoring various features of the wetland. The revisit period of ZY1E satellite is 2 days, and the multi-spectral band spatial resolution is 10 meters, which has high accuracy in vegetation monitoring. According to the time interval of once every quarter, the image data of the past 5 years of the two satellites is obtained to ensure that the wetland can be dynamically monitored for a long time sequence.

[0049] After obtaining the satellite remote sensing image data, professional remote sensing image processing software ENVI is used for preprocessing. First, radiation calibration is performed according to the radiation calibration parameter file provided by ZY1E satellite and BJ3N satellite to convert the original DN value of the image to radiance value or reflectivity value, so as to eliminate the influence of sensor itself and atmospheric scattering, absorption and other factors on data accuracy. In the geometric correction link, through the collection of evenly distributed ground control points in the wetland area (the coordinates of these control points are obtained by high-precision GPS measurement), the image is corrected to the unified CGCS2000 geographic coordinate system by using polynomial fitting method to ensure that the geometric accuracy of the image meets the requirements of subsequent analysis. Finally, according to the boundary vector data of the wetland, the image is cropped to extract the regional image of the target wetland, and the preprocessed satellite remote sensing image data is obtained to prepare for the subsequent wetland information extraction.

[0050] According to the macro features of the wetland reflected by ZY1E satellite remote sensing data and the distribution of the intelligent sample plot, a detailed unmanned aerial vehicle flight plan is developed. The longitudinal and horizontal CW-15 unmanned aerial vehicle platform is selected, which has good stability and load capacity and can carry various high-precision remote sensing equipment. The cameras carried include a 6-channel multi-spectral resolution of 0.08-meter-long optical camera to obtain multi-spectral image data, and a Rui Pi DG6M full-frame five-lens oblique photography camera to obtain five-lens RGB images of the wetland for making real scene three-dimensional model and digital orthographic image.

[0051] The UAV data collection work is carried out in spring and autumn when the weather conditions are suitable (wind speed less than 5 meters per second, no precipitation, cloud coverage less than 10%). According to the location of the wisdom sample plot and the key monitoring areas found in satellite remote sensing within the wetland (such as areas with complex changes in wetland vegetation community, areas with dynamic changes in water boundary, etc.), flight routes are planned to ensure that the UAV can fully cover and shoot these key areas. The flight height is set between 100-150 meters according to different monitoring needs to obtain image data with sufficient resolution. During the flight, the flight control system of the UAV automatically records key parameters such as flight trajectory, flight height, shooting angle, and shooting time.

[0052] After the UAV data collection is completed, PhotoScan and Reconstructor software are used for preprocessing. The software first removes fog and noise from the large number of original images taken by the UAV, performs radiation correction, and then performs aerial triangulation. Using the high-precision positioning information provided by the UAV and a small number of ground control points selected on the images (some of these control points can be shared with the satellite remote sensing geometric correction control points to ensure the same coordinate system), the images are corrected to the same CGCS2000 geographic coordinate system as the satellite remote sensing images. Finally, radiation correction is performed to convert the gray value of the image to the actual radiation brightness value or reflectivity value based on the camera's radiation calibration parameters, thereby obtaining preprocessed UAV image data that can be fused and analyzed with satellite remote sensing data and wisdom sample plot data.

[0053] Using the spectral characteristics of the preprocessed ZY1E satellite remote sensing images, the wetland land cover type classification is performed. By analyzing the spectral characteristics of a large number of sample areas with known land cover types, for example, wetland water has low reflectivity in the near-infrared band, while vegetation has high reflectivity in the near-infrared band, different land cover types such as wetland water, marsh vegetation, mudflat, forest, and grassland are gradually distinguished by setting different threshold conditions for reflectivity in different bands. After classification processing, the wetland land cover type classification results are obtained, and the area of different types and its change over time are counted.

[0054] The normalized difference vegetation index (NDVI) is calculated to monitor the growth of wetland vegetation. By analyzing the changes in NDVI values of satellite remote sensing images at different times, the NDVI time series curve is drawn to intuitively understand the growth season changes, interannual changes, and long-term growth trends of wetland vegetation. For example, if the NDVI value continues to rise in a certain period of time, it indicates that the vegetation growth in that area is good and the coverage is increasing; otherwise, there may be vegetation degradation.

[0055] By utilizing the texture and shape features of high-resolution UAV imagery, an object-oriented classification method is employed to analyze the wetland vegetation community structure. First, the image is segmented at multiple scales. Based on the actual distribution of wetland vegetation communities, appropriate segmentation scale parameters are selected to divide the image into objects of different sizes and shapes. Then, according to the spectral features (such as color differences of different vegetation in multispectral images), texture features (such as texture patterns formed by leaf arrangement), and shape features (such as the contour shape of vegetation communities), combined with the vegetation species information obtained from field surveys, the distribution range and community boundaries of different vegetation species are identified. For example, for the common reed community in wetlands, it presents specific texture and higher vegetation height in the image, which can be accurately distinguished from other vegetation and its distribution boundary can be accurately drawn.

[0056] For some small wetland patches or special topographic features (such as oxbow lakes, river islands, etc.) in wetlands, UAV images are used for fine boundary extraction. By using edge detection algorithms (such as Canny edge detection algorithm), the boundary contours of these special topographic features are identified by analyzing the gray level changes of image pixels, and then the detailed information such as area, perimeter, shape index, etc. is calculated. These micro-information is of great significance for in-depth study of the ecological structure and function of wetlands.

[0057] The soil moisture data collected by the smart plot are correlated with the vegetation growth conditions reflected in satellite remote sensing and UAV images. For example, in smart plot A, through long-term monitoring, it is found that when the soil moisture is between 25%-35%, satellite remote sensing and UAV images show that the wetland vegetation in this area grows most vigorously, with high vegetation coverage and NDVI value. Further analysis shows that when the soil water content is too high (more than 40%) or too low (less than 20%), the growth of vegetation will be inhibited, showing a decrease in NDVI value and vegetation coverage. Through this fusion analysis, the influence mechanism of soil water on wetland vegetation growth can be further explored, providing scientific basis for wetland water resource management and vegetation protection.

[0058] By combining water quality data from the smart plots with satellite remote sensing monitored water area changes and water body morphological features in UAV images, the pollution diffusion path and ecological self-purification capacity of wetland water bodies are analyzed. For example, smart plot C is located near the mouth of a major river in the wetland. The water quality monitoring data of this plot shows that the chemical oxygen demand and ammonia nitrogen concentration in the water body increase during the rainy season. At the same time, satellite remote sensing shows that the water area in this region has increased significantly, and UAV images clearly show the flow direction and water body diffusion form at the river mouth. By analyzing these data comprehensively, it can be determined that the main diffusion path of pollutants in the wetland is along the water flow from the river mouth to the interior of the wetland, and according to the growth status of vegetation in different regions and the differences in water body self-purification capacity, the degradation and purification effect of the wetland on pollutants can be evaluated.

[0059] According to the biodiversity monitoring data of the smart plots and the vegetation community distribution information in satellite remote sensing and UAV images, the stability and health status of the wetland ecosystem are evaluated. For example, the biodiversity monitoring data of smart plot E shows that the number and variety of birds and insects have decreased in recent years. Combined with satellite remote sensing and UAV image analysis, it is found that the vegetation community structure around this area has changed, and part of the wetland vegetation has been cultivated into farmland, resulting in habitat loss and ecological connectivity destruction. Through this multi-source data fusion analysis, the health status of the wetland ecosystem can be comprehensively evaluated, ecological problems can be found in time, and appropriate protection measures can be taken.

[0060] An index system for evaluating the health of the wetland ecosystem is established, which comprehensively considers the wetland ecological information extracted from multi-source data. Including land cover type diversity index, vegetation coverage and health index (combining satellite remote sensing and UAV image vegetation information, evaluating vegetation coverage area proportion and NDVI average value, etc.), water quality comprehensive index (according to the water quality data of the smart plot, using the water quality comprehensive pollution index method to combine multiple water quality parameters into a dimensionless index), biodiversity index, etc.

[0061] The weights of each index are determined by the analytic hierarchy process, and the relative importance weight values of each index in the evaluation system are calculated through expert scoring and the construction of a judgment matrix. For example, after analysis, the weight of the land cover type diversity index is 0.25, the weight of the vegetation coverage and health index is 0.3, the weight of the water quality comprehensive index is 0.2, and the weight of the biodiversity index is 0.25. Then, the fuzzy comprehensive evaluation method is used to quantitatively evaluate the health status of the wetland ecosystem. First, the evaluation set is determined, which is divided into five levels: healthy, sub-healthy, mild degradation, moderate degradation, and severe degradation. According to the actual monitoring values of each index and the weights, the membership degree of the wetland ecosystem to each evaluation level is calculated through fuzzy transformation, and finally the health level of the wetland ecosystem is determined. For example, after calculation, the evaluation result of the health status of the wetland ecosystem is sub-healthy, among which the vegetation coverage and health index and the biodiversity index are relatively low, indicating that the wetland vegetation and biodiversity are under certain pressure and need to be protected and managed.

[0062] Using the ArcGIS geographic information system platform, the information such as the distribution of land cover types, the growth status of vegetation, the change of water level, and the evaluation result of the ecosystem health of the wetland are displayed on the map in the form of thematic layers, and a wetland monitoring thematic map is made. On the map, different land cover types are represented by different colors, such as wetland water bodies represented by blue, marsh vegetation represented by green, and forest land represented by brown, etc., which intuitively shows the overall land use pattern of the wetland. The growth status of vegetation is displayed through the rendering of NDVI value classification, and the deeper the color, the higher the NDVI value and the more vigorous the vegetation growth. The change of water level is displayed by superimposing the water level contour line layers of different periods, which clearly presents the dynamic change of the water level of the wetland. The evaluation result of the ecosystem health is marked on the map with different symbols or colors, which facilitates quick identification of the health status of different areas of the wetland.

[0063] Through drawing line graphs, column charts, scatter plots, etc., the trends of various monitoring indicators of the wetland over time are displayed. For example, the annual change line graph of the wetland area is drawn to intuitively reflect the increase and decrease of the wetland area in the past few years; the seasonal change column chart of the vegetation index is drawn to show the growth vigor change of the wetland vegetation in different seasons; the scatter plot of water quality parameters and time is drawn to analyze the fluctuation law of water quality indicators over time and the relationship with factors such as season and precipitation. These charts help to deeply analyze the dynamic change process of the wetland ecosystem and provide data support for wetland management decision-making.

[0064] For some key areas or typical wetland topography, a three-dimensional model is constructed using UAV images for stereoscopic display. For example, for a large oxbow lake area in the wetland, a three-dimensional model of the oxbow lake is constructed using five-lens image data captured by a UAV through three-dimensional modeling software (Reconstruction Master), which can intuitively observe the topographic and geomorphic features, surrounding vegetation distribution, and water body morphology of the oxbow lake, providing a powerful tool for in-depth study of the ecological function and evolution process of the oxbow lake. At the same time, the monitoring data of the smart plot is displayed on the monitoring platform in the form of real-time data charts, such as in the smart plot monitoring interface, the real-time changes of meteorological data (temperature, humidity, precipitation, etc.) are displayed in the form of dynamic line charts, and the real-time monitoring values of soil moisture and water quality parameters are displayed in the form of column charts, which facilitates wetland management personnel to view the dynamic changes of in-situ monitoring information of the wetland at any time, and to take appropriate measures in a timely manner.

[0065] Through the above specific implementation cases, the practical application process of the wetland monitoring method based on multi-source remote sensing data and smart plots is demonstrated in detail, from data collection, processing, analysis to final result evaluation and visualization display, each link is closely connected, fully embodying the effectiveness and practicality of the method, and providing comprehensive and accurate technical support for wetland monitoring and protection.

[0066] As a possible implementation manner, the embodiment provides an electronic device, including a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements a wetland monitoring method combining multi-source remote sensing and UAV when executing the program.

[0067] As a possible implementation manner, the embodiment provides a computer readable storage medium, which stores a computer program, and the program is executed by a processor to implement a wetland monitoring method combining multi-source remote sensing and UAV.

[0068] The above specific embodiments further illustrate the purpose, technical solutions and advantages of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application shall be included in the protection scope of the present application.​

Claims

1. A method for monitoring wetlands using multi-source remote sensing combined with unmanned aerial vehicles, characterized by, The method comprises the following specific steps: acquiring multi-band remote sensing images of a target detection area based on satellite remote sensing; preprocessing the multi-band remote sensing images of the target detection area to obtain multi-band remote sensing images with continuous time and space and noise suppression; performing segmentation and fusion processing on the multi-band remote sensing images with continuous time and space and noise suppression to obtain optimized multi-band remote sensing images; extracting spectral features and spatial features according to the band information of the optimized multi-band remote sensing images; extracting species distribution ranges and community boundaries based on the spectral features and the spatial features to obtain community distribution data of the target detection area; planning a monitoring path of a UAV based on the community distribution data of the target detection area; monitoring the target area using the UAV to obtain UAV multi-spectral image data, wherein the UAV multi-spectral image data comprises multiple bands similar to the wavelength range of the remote sensing image data; performing radiation correction and geometric correction on the UAV multi-spectral image data to obtain UAV hyperspectral image data; performing band matching on the UAV hyperspectral image data and the remote sensing image, fusing the UAV hyperspectral image data and the remote sensing image, identifying a wetland distribution area, and generating a wetland species distribution map.

2. The method of claim 1, wherein, The preprocessing of the multi-band remote sensing images of the target detection area specifically comprises: performing interpolation operation and resampling processing on the spatial features of the coarse-resolution multi-band remote sensing images; performing pixel-level matching on the time series data of the coarse-resolution images after the interpolation operation and resampling processing and the cloud-free high-resolution multi-band remote sensing images to obtain corrected time series; performing time-space modeling filling on the missing values in the original multi-band remote sensing images to restore the complete time series; performing smoothing processing on the time series using a weighted seasonal-trend decomposition filter to output the multi-band remote sensing images with continuous time and space and noise suppression.

3. The method for monitoring wetlands using multi-source remote sensing integrated with UAV as claimed in claim 1, wherein, The segmentation and fusion processing of the multi-band remote sensing images with continuous time and space and noise suppression specifically comprises: randomly selecting a single pixel in the multi-band remote sensing images; segmenting homogeneous regions and heterogeneous regions in the target area based on the pixel; merging the homogeneous regions into a new object; segmenting the heterogeneous regions and merging the homogeneous regions in the heterogeneous regions into the new object until the segmentation of the heterogeneous regions is completed to obtain segmented multi-band remote sensing images; integrating the segmented multi-band remote sensing images to obtain the optimized multi-band remote sensing images.

4. The method for monitoring wetlands using multi-source remote sensing integrated with UAV as claimed in claim 1, wherein, The extraction of the spectral features and the spatial features according to the band information of the optimized multi-band remote sensing images specifically comprises: extracting the spectral reflectance curve of each pixel to obtain spectral features reflecting the types of vegetation; acquiring the spatial resolution of the multi-band remote sensing images, calculating the texture features, and obtaining spatial features reflecting the vegetation community. 5.The method of claim 1, wherein, The extraction of the species distribution ranges and the community boundaries based on the spectral features and the spatial features to obtain the community distribution data of the target detection area specifically comprises: calculating vegetation indices according to the band information of the optimized multi-band remote sensing images; extracting boundary information of each vegetation community unit based on the spectral features and the spatial features to obtain the species distribution ranges and the community boundaries. According to the vegetation index, the species distribution range and the community boundary, a threshold is set to separate the vegetation area from the non-vegetation area; Using a region growing algorithm, the vegetation area is divided into multiple community units to obtain community distribution data of the target detection area.

6. The method for monitoring wetlands using multi-source remote sensing integrated with UAV as claimed in claim 1, wherein, Based on the community distribution data of the target detection area, the monitoring path of the unmanned aerial vehicle is planned, specifically including: Based on the community distribution data of the target detection area, according to the requirements of the target detection area, the types of vegetation communities that need to be monitored are determined, and priorities are assigned to different communities; Obtaining the performance parameters of the unmanned aerial vehicle, combining the community priorities, and dividing the monitoring area; Based on the terrain data of the monitoring area, the heading overlap degree and the lateral overlap degree are calculated; Based on the overlap degree requirement, an initial flight point distribution is automatically generated; According to the initial flight point distribution, the data of the range area is collected, and the overlap degree and the splicing quality of the data are analyzed; If the overlap degree and the splicing quality do not meet the set threshold, the flight point density is adjusted; According to the adjusted flight point density, the monitoring path of the unmanned aerial vehicle is generated.

7. The method for monitoring wetlands using multi-source remote sensing integrated with UAV as claimed in claim 1, wherein, The multispectral image data of the unmanned aerial vehicle is radiometrically and geometrically corrected: The radiometric correction includes: Converting the pixel brightness value data of the original image to atmospheric outer surface reflectivity data, extracting the radiation calibration parameters monitored by the sensor to obtain the radiation calibration data; Taking the radiation value of the calibration tarpaulin as the standard value, the radiation calibration data is subjected to reflectance correction, and the radiation value data is converted into reflectivity data; The geometric correction includes: Using a semiconductor photosensitive element to image, the spectral information on a space line is obtained; Through mechanical push scanning, the entire space image and spectral data are collected; The image data of the entire area obtained according to the monitoring path of the unmanned aerial vehicle is spliced into multiple rows of bands to obtain unmanned aerial vehicle hyperspectral image data.

8. The method for monitoring wetlands using multi-source remote sensing integrated with UAV as claimed in claim 1, wherein, The unmanned aerial vehicle hyperspectral image data is matched with the remote sensing image, the unmanned aerial vehicle hyperspectral image data and the remote sensing image are fused, the wetland distribution area is identified, and a wetland species distribution map is generated, specifically including: Based on the band information of the multi-band remote sensing image, the corresponding band range of the hyperspectral data is determined; From the hyperspectral data, the spectral reflectivity data corresponding to the bands of the multi-band remote sensing image is extracted, and the extracted spectral reflectivity data is consistent with the band information of the remote sensing image in wavelength range and resolution; According to the extracted spectral reflectivity information of the hyperspectral data, the same vegetation index as the remote sensing vegetation index is calculated; Obtaining the vegetation index calculated according to the band reflectivity value of the multi-band remote sensing image; Fusing the calculated vegetation index of the unmanned aerial vehicle image with the vegetation index of the remote sensing data; Based on the fused data, the fused data features are extracted; A classification model is constructed to classify the fused data features, identify the wetland distribution area, and generate a wetland species distribution map.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to implement the wetland monitoring method of multi-source remote sensing combined with an unmanned aerial vehicle according to any one of claims 1 to 8.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the wetland monitoring method of multi-source remote sensing combined with an unmanned aerial vehicle according to any one of claims 1 to 8.

Images