Grassland ecological monitoring methods, devices, systems, and storage media

Abstract

This invention discloses a grassland ecological monitoring method, device, system, and storage medium, comprising: acquiring measured grassland quadrat data and near-surface environmental data; acquiring low-altitude remote sensing images and structural information of the grassland; completing spatial registration, feature extraction, data fusion, parameter inversion, and ecological evaluation of ground and aerial data; and generating grassland ecological monitoring maps, monitoring reports, and evaluation results. By employing the technical solution of this invention, efficient acquisition, scale conversion, fusion analysis, and comprehensive evaluation of grassland ecological parameters are achieved, thereby improving the accuracy, continuity, and practicality of grassland ecological monitoring.

Description

Technical Field

[0001] This invention belongs to the field of ecological monitoring technology, specifically relating to a grassland ecological monitoring method, device, system, and storage medium. Background Technology

[0002] Grassland ecosystems are an important component of terrestrial ecosystems, playing a vital role in maintaining biodiversity, ensuring livestock production, and regulating regional carbon cycles and hydrological processes. Current grassland ecological monitoring primarily focuses on indicators such as vegetation cover, grass height, aboveground biomass, species composition, community structure, and degree of degradation. Existing grassland ecological monitoring methods can be broadly categorized as follows.

[0003] The first category is the traditional ground quadrat survey method. This method typically involves technicians setting up quadrats within the monitoring area to manually measure and record indicators such as grassland cover, height, density, species composition, biomass, and soil moisture. While this method offers advantages such as high monitoring accuracy and clear ecological significance, it suffers from drawbacks including high labor intensity, low operational efficiency, limited spatial coverage, and high costs associated with repeated monitoring. It is particularly difficult to meet the demands for rapid, continuous, and large-scale application in large-scale grassland dynamic monitoring.

[0004] The second category is single-aerial remote sensing monitoring methods. Current technologies often acquire grassland images through satellite or UAV remote sensing, and then estimate grassland growth based on vegetation indices, texture features, or elevation models. These methods have advantages such as large acquisition range, good repeatability, and strong timeliness. However, relying solely on aerial imagery is easily affected by spatial resolution, mixed ground feature pixels, changes in lighting conditions, and surface background interference. Their ability to identify species composition, near-surface structural parameters, and local ecological conditions remains limited. Furthermore, aerial remote sensing inversion models often require calibration and verification using ground-based measured data; otherwise, their accuracy and generalization ability are difficult to guarantee.

[0005] The third category involves methods that separate the use of ground and aerial data. While some existing technologies employ both ground surveys and UAV remote sensing, ground data is typically used only for post-hoc verification. The lack of a unified sampling design, a unified spatial positioning benchmark, and a unified data fusion mechanism between the two methods makes it difficult to achieve effective collaboration between ground and aerial data, resulting in problems such as scale mismatch, low data utilization, poor temporal consistency, and insufficient stability of evaluation results.

[0006] In addition, existing grassland ecological monitoring technologies generally suffer from the following shortcomings: First, the monitoring indicator system is relatively fragmented, with parameters such as cover, height, biomass, and diversity often being measured separately, lacking a unified integrated evaluation process; second, monitoring results mostly remain at the level of single-indicator expression, making it difficult to form a comprehensive output for grassland ecological quality assessment; third, there is a lack of a unified temporal comparison mechanism between different monitoring phases, which is not conducive to the dynamic identification of key stages such as the greening period, growth period, and peak grassing period; fourth, the adaptability to complex grassland patches, heterogeneous communities, and degraded surface backgrounds is weak, making it difficult to achieve ecological monitoring that balances accuracy and efficiency. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention provides a grassland ecological monitoring method, device, system, and storage medium, which solves the problems of low efficiency of ground monitoring, insufficient accuracy of single aerial monitoring, difficulty in coordinating and integrating ground and aerial data, scattered evaluation of ecological indicators, and weak ability to express time-series dynamics in the prior art.

[0008] To achieve the above objectives, the present invention provides the following solution: A grassland ecological monitoring method, comprising: Obtain measured data from grassland quadrats and near-surface environmental data; Acquire low-altitude remote sensing images and structural information of grasslands; Based on the measured data of grassland quadrats, near-ground environmental data, low-altitude remote sensing images of grassland, and structural information, spatial registration, feature extraction, data fusion, parameter inversion, and ecological assessment of ground and aerial data are performed. Generate grassland ecological monitoring maps, monitoring reports, and evaluation results.

[0009] As a preferred approach, grassland ecological data is acquired simultaneously, and ground-air integrated analysis is performed using a unified coordinate benchmark and quadrat mapping method. At the same time, the ground quadrat boundaries are mapped to the UAV orthophoto after being located by RTK or high-precision GNSS, so as to achieve one-to-one correspondence and registration between the measured data of the quadrat and the aerial image area.

[0010] As a preferred approach, aerial remote sensing spectral features, texture features, and structural features are extracted and fused with ground-acquired features such as cover, height, biomass, species composition, diversity, and soil moisture. Simultaneously, the fusion model is used to invert grassland cover, grass height, aboveground biomass, species richness, or diversity indicators, and further constructs comprehensive grassland ecological evaluation indicators.

[0011] The present invention also provides a grassland ecological monitoring device, comprising: Ground monitoring units are used to acquire measured data from grassland quadrats and near-ground environmental data. The aerial monitoring unit is used to acquire low-altitude remote sensing images and structural information of grasslands; The data processing unit is used to perform spatial registration, feature extraction, data fusion, parameter inversion, and ecological assessment of ground and aerial data based on grassland quadrat measured data, near-ground environmental data, low-altitude remote sensing images of grassland, and structural information. The results output unit is used to generate grassland ecological monitoring maps, monitoring reports, and evaluation results.

[0012] As a preferred approach, the ground monitoring unit and the aerial monitoring unit simultaneously acquire grassland ecological data and conduct integrated ground-air analysis through a unified coordinate benchmark and quadrat mapping method. At the same time, the ground quadrat boundaries are mapped to the UAV orthophoto after being located by RTK or high-precision GNSS, so as to achieve one-to-one correspondence and registration between the measured data of the quadrat and the aerial image area.

[0013] As a preferred approach, the data processing unit simultaneously extracts aerial remote sensing spectral features, texture features, and structural features, and integrates them with ground-acquired features such as cover, height, biomass, species composition, diversity, and soil moisture. At the same time, the fusion model is used to invert grassland cover, grass height, aboveground biomass, species richness, or diversity indicators, and further constructs comprehensive grassland ecological evaluation indicators.

[0014] The present invention also provides a grassland ecological monitoring system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program performs a grassland ecological monitoring method when executed by the processor.

[0015] The present invention also provides a storage medium storing a computer program, which executes a grassland ecological monitoring method when running.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention establishes a unified sampling, positioning, registration, and fusion mechanism between ground-based and aerial monitoring units to achieve collaborative acquisition and comprehensive analysis of grassland ecological parameters. This invention not only improves the monitoring accuracy of indicators such as grassland cover, grass height, aboveground biomass, species composition, and ecological quality, but also significantly enhances monitoring efficiency, the repeatability of monitoring results, and the ability to be applied regionally. It provides technical support for grassland resource surveys, degradation assessments, ecological restoration effectiveness analysis, and grassland management decision-making. Attached Figure Description

[0017] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the grassland ecological monitoring method according to an embodiment of the present invention; Figure 2 A schematic diagram illustrating the registration of ground quadrats with UAV orthophotos; Figure 3 A schematic diagram of grassland ecological feature integration and parameter inversion; Figure 4 This is a diagram illustrating the output and dynamic updating of monitoring results. Figure 5 This is a schematic diagram of the grassland ecological monitoring device according to an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example 1 like Figures 1 to 4 As shown, the present invention provides a grassland ecological monitoring method, comprising: Step S1: Delineation of monitoring area and layout of quadrats Within the grassland area to be monitored, monitoring zones are delineated based on topography, grassland type, utilization method, and vegetation heterogeneity. Several ground quadrats are set up using a stratified layout method, which includes stratification according to vegetation type, slope position, degradation level, or utilization intensity.

[0022] Each quadrat's center and boundary coordinates are recorded, and ground markers are placed to make the quadrat identifiable in UAV imagery. Preferably, high-contrast markers or reflective markers are placed at the four corners of the quadrat to improve the accuracy of subsequent ground-to-air registration.

[0023] Step S2: Ground Ecological Information Collection Ground ecological information was collected simultaneously in each quadrat. The ground ecological information includes one or more of the following: vegetation cover, average grass height, aboveground biomass, species composition and dominant species information, species richness, soil moisture content, soil temperature, surface exposure, and proportion of degraded patches.

[0024] Vegetation cover can be obtained using the needle acupuncture method, visual estimation, or near-ground photography; average grass height can be obtained using a measuring rod; aboveground biomass can be obtained using the harvesting and weighing method or an empirical conversion method; species richness is the number of plant species within the quadrat; and species diversity is preferably characterized using the Shannon-Wiener index. H′ = -Σ(pi·lnpi) Where pi is the relative importance, relative coverage, or relative biomass of the i-th species in the quadrat.

[0025] Step S3: Aerial Remote Sensing Data Acquisition and DSM Construction On the same day or within a similar timeframe after the completion of the ground quadrat survey, low-altitude aerial surveys were conducted over the monitoring area using an aerial monitoring unit. The UAV flew along a pre-set route, acquiring RGB and multispectral images covering the entire monitoring area. To minimize errors introduced by temporal differences, flights were preferably conducted during clear, low-wind periods. The multispectral camera underwent standard reflector calibration before and after flight to improve the consistency of band reflectivity data.

[0026] If it is necessary to obtain 3D structural information of the grass layer, high-overlap photogrammetry or low-altitude point cloud acquisition should be performed simultaneously to construct a digital surface model (DSM). The construction of the DSM is preferably implemented using a photogrammetric workflow based on SfM-MVS, which includes feature point extraction, image matching, aerial triangulation, dense point cloud reconstruction, and surface rasterization. This part can preferably be implemented using existing mature photogrammetric software or algorithms.

[0027] Let the observed coordinates of the j-th feature point in the i-th image be xij, and the corresponding spatial point be Xj. The camera projection model is as follows: Where Ki is the camera intrinsic parameter matrix, Ri and ti are the extrinsic parameters of the i-th image, and π(·) is the projection function. Aerial triangulation is accomplished by minimizing the reprojection error: After aerial triangulation optimization, a dense point cloud P={(xk,yk,zk)} is generated based on multi-view image matching. Projecting the point cloud onto the raster cells Ωuv allows for the construction of a DSM. Based on this, a construction process can be formed: "original image → feature extraction and matching → aerial triangulation → dense point cloud → DSM / DEM → DOM / CHM".

[0028] Step S4: Ground-to-Air Data Preprocessing The acquired raw ground and aerial data undergo preprocessing. Preprocessing of aerial remote sensing data includes lens distortion correction, radiometric correction, image stitching, orthorectification, and multispectral band registration; when the terrain is significantly undulating, terrain correction is also included. If photogrammetry is used to acquire 3D information, an orthophoto map (DOM), a digital surface model (DSM), and a digital elevation model (DEM) are further generated, and a grass height model (CHM) is constructed using the difference between the DSM and DEM. Preprocessing of ground data includes standardizing quadrat numbering, coordinate format standardization, outlier removal, unit standardization, and missing value correction.

[0029] Lens distortion correction preferably employs a pinhole camera model and radial and tangential distortion models. Let the normalized image point coordinates be (x, y), and let r² = x² + y², then the distorted image point coordinates (xd, yd) can be expressed as: Where k1, k2, and k3 are radial distortion coefficients, and p1 and p2 are tangential distortion coefficients. After obtaining the intrinsic parameter matrix and distortion parameters through camera calibration, lens distortion correction can be completed.

[0030] Radiation correction is preferably performed using a reflectivity plate and / or an airborne illumination sensor. If the reflectivity plate method is used, assuming the known reflectivity of the calibration plate is ρpanel,λ, the average surface density (DN) is DN̄panel,λ, and the dark current is DNdark,λ, then the calibration coefficient Cλ is: The reflectivity of a target pixel can be expressed as: If the radiance / irradiance method is used, then the surface reflectance can be expressed as: Where Lλ is the radiance and Eλ is the downlink irradiance.

[0031] Multispectral band registration is preferably based on a reference band, establishing geometric transformation relationships from other bands to the reference band. Let the pixel coordinates of the band to be registered be... , The transformed coordinates p̃j can then be expressed as: Here, Hj is the homography matrix or affine transformation matrix obtained from the control points. By minimizing the control point error and solving for Hj, the alignment of each band can be achieved.

[0032] During image stitching and orthorectification, for each pixel in the output DOM, it is back-projected onto the original image according to the elevation constraints provided by the DSM / DEM and then fused according to weights. The DOM pixel value can be expressed as: Where ωi is a weight related to viewpoint, sharpness, and occlusion status, and Ii(·) is the pixel value of the i-th original image at the corresponding projection position.

[0033] The preferred method for generating a DEM is to first classify the dense point cloud into ground point sets G, and then generate a surface elevation model using inverse distance weighted interpolation or triangulation interpolation. When using inverse distance weighted interpolation: Where zi is the elevation of the nearest ground point, di is the distance from the ground point to the center of the grid, and p is the distance power exponent.

[0034] The grass height model CHM is obtained by comparing the DSM and DEM: Wherein, CHM′ is the grass height model after outlier truncation. The above preprocessing process can preferably be implemented using existing mature photogrammetry and image processing techniques.

[0035] Step S5: Ground-to-Air Space Registration By mapping the boundaries of the ground quadrat to the coordinate system of the orthophoto, a precise correspondence between the ground quadrat and the aerial image is achieved. The aforementioned "ground quadrat boundary" is derived from the coordinates of the four corner points directly measured in step S1, or from the coordinates of the four corner points calculated from the center point, side length, and azimuth.

[0036] The specific process of the "mapping" includes: first, transforming the sample plot boundary from the geographic coordinate system to the projected coordinate system used by the DOM; second, establishing a coarse registration relationship based on RTK coordinates and DOM coordinates; third, performing fine registration correction using sample plot corner markers, local image features, or template matching results; and finally, generating a sample plot polygon mask and cropping it to obtain the corresponding image sample area.

[0037] In the coarse registration stage, a two-dimensional similarity transformation is preferred: Where (x,y) are RTK coordinates, (u,v) are image coordinates, s is the scale factor, φ is the rotation angle, and tx and ty are translation amounts.

[0038] In the fine registration stage, the corner points or local feature points of the quadrat markers are used as control points to minimize control point errors: Where pi represents ground control points, qi represents image control points, and T represents the affine transformation or homography matrix. If outlier matching points exist, robust estimation using RANSAC is preferred.

[0039] The quadrat mask can be represented as: Here, Ωplot represents the registered quadrat region. Through the above process, each ground quadrat corresponds to a unique aerial image sample area, thus forming an integrated ground-air analysis unit.

[0040] Step S6: Extraction of grassland ecological features For the integrated ground-air analysis unit obtained in step S5, aerial remote sensing features and ground survey features are extracted. Aerial remote sensing features include spectral features, texture features, and structural features; ground survey features include vegetation cover, grass height, aboveground biomass, number of species, proportion of dominant species, Shannon-Wiener diversity index, soil moisture content, and degradation characteristics.

[0041] The preferred spectral characteristics include one or more of NDVI, EVI, SAVI, OSAVI, GNDVI, and NDRE, and their calculation formulas are as follows: Texture features are preferably calculated using the gray-level co-occurrence matrix (GLCM). Let the co-occurrence matrix be P(i,j), then contrast, entropy, and homogeneity can be expressed as: Structural features are preferably extracted from CHM, DSM, or point clouds, including average grass height, canopy relief, height standard deviation, patch area, bare ground percentage, and spatial fragmentation degree. Among these: Where h̄ is the average height, σh is the standard deviation of height, Rb is the proportion of bare land, Abare is the area of ​​bare land, and Aplot is the area of ​​the quadrat. To improve robustness, it is preferable to further statistically analyze the mean, median, standard deviation, quantiles, and area proportions within the quadrat mask area. The above extraction process can be implemented using existing mature remote sensing feature extraction algorithms.

[0042] Step S7: Fusion of Ground and Space Features and Inversion of Ecological Parameters The aerial remote sensing features extracted in step S6 are fused with ground survey features to construct a grassland ecological parameter inversion model. To ensure that the "preferred adoption" method has a clear implementation method, in a specific embodiment, the measured cover, average height, aboveground biomass, and species richness of the sample plot are used as label variables y, and the spectral features, texture features, structural features, and ground environmental features extracted from the corresponding sample plot area are combined to form the input feature vector x.

[0043] The overall feature vector can be represented as: Before modeling, it is preferable to use range normalization for standardization: The preferred method is to use a random forest regression model to invert vegetation cover, grass height, aboveground biomass, and species richness, respectively. The random forest predictions can be expressed as: Where B is the number of regression trees, and Tb is the prediction result of the b-th regression tree. In a preferred embodiment, B can be between 100 and 500. In addition to random forests, existing mature algorithms such as multiple linear regression, partial least squares regression, support vector regression, or gradient boosting trees can also be used.

[0044] If support vector regression is used, its optimization objective can be expressed as: The preferred metrics for evaluating model accuracy include the coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). Subsequently, the parameters obtained from the inversion, such as cover, height, biomass, and diversity, were normalized to construct a comprehensive grassland ecological evaluation index E. A weighted summation method was preferred. In one specific implementation, C′, H′, B′, and D′ can represent the normalized values ​​of cover, height, biomass, and diversity, respectively. Among them, the weight wk can be preferably determined by the analytic hierarchy process, the entropy weight method, or the expert weighting method.

[0045] Step S8: Grassland Ecological Status Assessment Based on the comprehensive grassland ecological evaluation indicators obtained in step S7, the ecological status of the monitoring area is determined. The determination includes one or more of the following: grassland ecological quality level, grassland degradation degree, grassland growth status, key growth phases, and restoration effectiveness evaluation.

[0046] The preferred evaluation indicators include vegetation cover, average height, aboveground biomass, species richness, Shannon-Wiener diversity index, bare land ratio, soil moisture content, and comprehensive ecological index E. The evaluation can preferably be based on current standards, including GB 19377-2003 "Classification Indicators for Degradation, Desertification, and Salinization of Natural Grassland", GB / T 41282-2022 "Verification of the Authenticity of Vegetation Cover Remote Sensing Products", QX / T 494—2019 "Evaluation Grades for Meteorological and Ecological Quality Monitoring of Terrestrial Vegetation", and LY / T 3318-2022 "Technical Specifications for Monitoring and Evaluation of the Benefits of Grassland Ecological Construction Projects". These standards can be used for degradation degree classification, verification of the authenticity of vegetation cover remote sensing results, evaluation of vegetation ecological quality grades, and monitoring and evaluation of the effectiveness of grassland ecological construction and restoration.

[0047] In one specific implementation, if the comprehensive index Normalized to Based on the range, the ecological state can be divided into four levels: For assessing the degree of degradation, it can be further divided into four levels: non-degraded, slightly degraded, moderately degraded, and severely degraded. For multi-temporal monitoring tasks, it can also be used to identify the greening-up period, rapid growth period, and peak vegetation period based on changes in cover, height, biomass, and diversity at different times. The above-mentioned threshold levels can also be locally calibrated based on historical samples and regional standards of the monitoring area.

[0048] Step S9: Monitoring Result Output and Dynamic Update The results of step S8 are output as a thematic map of grassland ecological monitoring, a quadrat statistical table, and a regional evaluation report. When monitoring is repeated in the same area at different time periods, a time series database is established to dynamically compare grassland ecological parameters, generating ecological change trend curves and zonal early warning results.

[0049] In a preferred embodiment, the method is used for monitoring typical grassland degradation. Specifically, multiple 1 m × 1 m quadrats are set up in the target grassland to collect data on cover, height, aboveground biomass, species number, and soil moisture content. Simultaneously, a multi-rotor UAV equipped with RGB and multispectral cameras is used for aerial surveying to obtain orthophotos and multispectral images. Then, based on the ground-to-air spatial registration results, features such as NDVI, NDRE, texture entropy, grass height model, and bare land ratio corresponding to the quadrats are extracted. Next, a random forest model is used to invert regional cover, biomass, and species richness. Finally, a degradation level distribution map and an ecological quality distribution map are output based on comprehensive evaluation indicators.

[0050] In another preferred embodiment, the method is used for dynamic monitoring of grassland during the growing season. Continuous monitoring is conducted during the greening-up stage, the growth stage, and the peak grassing stage. By using a unified formula, a unified flight path, and a unified processing procedure, multi-temporal data can be compared, enabling dynamic analysis of grassland growth progress and differences in growth status.

[0051] Compared with the prior art, the present invention has at least the following beneficial effects.

[0052] First, by establishing a unified sampling and registration mechanism between ground monitoring units and aerial monitoring units, the true collaborative application of ground-measured data and UAV remote sensing data has been realized, overcoming the problems of separate use of ground data and aerial data, scale mismatch, and difficulty in fusion in existing technologies.

[0053] Secondly, this invention combines the high precision advantage of ground surveys with the wide coverage advantage of low-altitude remote sensing. It can ensure the accuracy of ecological indicators such as grassland cover, height, biomass, and diversity, and significantly improve monitoring efficiency, making it suitable for large-scale, continuous, and batch grassland ecological monitoring tasks.

[0054] Third, this invention constructs a unified process for extracting and comprehensively evaluating grassland ecological features, integrating and analyzing spectral, texture, structural, and ground ecological information, thus realizing the transformation from single-indicator monitoring to comprehensive ecological evaluation, which can more comprehensively reflect the ecological status of grassland.

[0055] Fourth, this invention has strong temporal extension capabilities and can be applied to monitoring key growth periods such as the greening period, growth period, and peak grass growth period, providing continuous support for grassland dynamic change analysis and ecological restoration effect evaluation.

[0056] Fifth, this invention has good versatility and scalability. This method is not dependent on a single grassland type and allows for flexible adjustment of quadrat size, flight parameters, and evaluation indicators based on the grassland conditions of different regions. It is applicable to monitoring natural grasslands, degraded grasslands, artificial grasslands, and ecological restoration areas.

[0057] Example 2 like Figure 5 As shown, the present invention also provides a grassland ecological monitoring device, comprising: Ground monitoring units are used to acquire measured data from grassland quadrats and near-ground environmental data. The aerial monitoring unit is used to acquire low-altitude remote sensing images and structural information of grasslands; The data processing unit is used to complete spatial registration of ground and aerial data, feature extraction, data fusion, parameter inversion, and ecological assessment. The results output unit is used to generate grassland ecological monitoring maps, monitoring reports, and evaluation results.

[0058] The grassland ecological monitoring device is implemented according to the process of "monitoring area deployment - ground data acquisition - aerial data acquisition - data preprocessing - spatial registration - feature fusion - parameter inversion - comprehensive evaluation - result output".

[0059] As one embodiment of the present invention, the ground monitoring unit includes a quadrat frame, a positioning module, a ground sensing module, and a survey recording module. The quadrat frame is used to determine the grassland survey area and is generally a square aluminum alloy or PVC quadrat frame with sides of 0.5 m, 1 m, or 2 m. The quadrat area can also be adjusted according to the actual situation of the monitored object. The positioning module preferably uses an RTK positioning instrument or a high-precision GNSS positioning instrument to record the coordinates of the quadrat center point and boundary points. The "ground quadrat boundary" is obtained in two ways: first, by directly measuring the coordinates P1, P2, P3, and P4 of the four corner points of the quadrat using RTK or high-precision GNSS; second, by first measuring the coordinates C(xc, yc), side length L, and azimuth θ of the quadrat center point, and then calculating the coordinates of the four corner points through geometric transformation.

[0060] When the center point estimation method is used locally, the coordinates of the four corner points of the sample plot can be expressed as: Where C is the center point of the quadrat, R(θ) is the rotation matrix, and di are the local coordinates of the four corner points relative to the center of the quadrat. This forms the quadrat boundary polygon required for subsequent spatial registration.

[0061] The ground sensing module includes one or more of the following: soil moisture sensor, soil temperature sensor, air temperature and humidity sensor, and light sensor, used to synchronously collect near-surface environmental information. The survey recording module records the quadrat number, time, location, vegetation cover, average height, dominant species, number of species, aboveground biomass, and degradation characteristics. This can be achieved using paper recording forms, handheld terminals, tablets, or mobile survey apps. Preferably, the survey recording module is time-synchronized with the positioning module to ensure consistency between ground data and aerial imagery.

[0062] In one embodiment of the present invention, the aerial monitoring unit includes a drone platform, a visible light camera, a multispectral camera, and a flight control module. The drone platform is preferably a multi-rotor drone to meet the requirements for low-altitude, low-speed, and high-precision grassland image acquisition. The visible light camera is used to acquire high spatial resolution RGB images, and the multispectral camera is used to acquire multispectral images including blue, green, red, red-edge, and near-infrared bands.

[0063] The flight control module includes a mission planning submodule, a status awareness submodule, a trajectory tracking submodule, a trigger acquisition submodule, and a safety management submodule. The mission planning submodule generates a flight path based on the monitoring area boundary, flight altitude, sensor field of view, and target overlap. The status awareness submodule fuses GNSS, IMU, barometer, and compass data to obtain the UAV's real-time position, speed, heading angle, and altitude. The trajectory tracking submodule controls the UAV to fly along a preset trajectory. The trigger acquisition submodule automatically calculates the exposure interval based on the forward overlap and triggers the camera to capture images. The safety management submodule executes return-to-home, hovering, or landing in cases of excessive wind speed, insufficient battery power, or abnormal positioning.

[0064] In a preferred embodiment, let the flight altitude be H, the camera's field of view along the flight direction be α, and the lateral field of view be β. Then, the coverage length Lf and lateral coverage width Ls of a single image along the flight direction are respectively: Let the forward overlap be rf ​​and the lateral overlap be rs, then the line spacing Ds and the adjacent exposure point spacing Df are respectively: When the drone's flight speed is v, the camera exposure time interval Δt is: Trajectory tracking is preferably implemented using existing waypoint planning and closed-loop control algorithms, such as PID control or Model Predictive Control (MPC). In PID control, the control input u(t) can be expressed as: Where e(t) represents the error between the current position and the target trajectory, and Kp, Ki, and Kd are the proportional, integral, and derivative coefficients, respectively. Flight altitude control can also adjust for altitude errors. The same closed-loop control method is adopted. In the flight control stage, this invention preferably uses existing mature autopilot technology, without requiring a specific deep neural network as a necessary limitation. Preferably, the directional overlap is not less than 75%, the lateral overlap is not less than 70%, and the flight altitude is 20 m to 120 m.

[0065] As one embodiment of the present invention, the data processing unit includes: a data preprocessing module, a spatial registration module, a feature extraction module, a fusion modeling module, and an ecological evaluation module.

[0066] Furthermore, the data preprocessing module standardizes the original RGB imagery, multispectral imagery, and ground sample data, outputting a standard dataset that can be used for subsequent registration and modeling. This module includes sub-modules for lens distortion correction, radiometric correction, band registration, image stitching and orthorectification, and point cloud reconstruction. Camera distortion correction can be implemented using a pinhole camera model and radial / tangential distortion models; the OpenCV official documentation lists the camera matrix and radial distortion coefficients. and tangential distortion coefficient The standard expression.

[0067] For normalized image points ,make distorted image points It can be represented as: By solving the camera intrinsic parameter matrix Distortion coefficient vector and external references This allows for lens distortion correction.

[0068] Furthermore, the spatial registration module is used to accurately map the boundaries of ground quadrats to the DOM coordinate system. Its inputs include the coordinates of the four corner points of the ground quadrats, the coordinates of the quadrat center point, the DOM / DSM product, and the image features of the marker board. The output is an image analysis unit corresponding to each quadrat. A two-stage registration approach is preferred: first, coarse registration based on coordinates, followed by fine registration based on image features.

[0069] In the coarse registration stage, a two-dimensional similarity transformation is used: in, For RTK coordinates, For orthophoto coordinates, As a scale factor, The rotation angle is... This represents the translation amount.

[0070] In the fine registration stage, the corner points or local feature points of the quadrat markers are used as control points to minimize the reprojection error. in, For ground control points, For image control points, This represents an affine transformation or homography matrix. If local distortion is significant, the homography matrix can be estimated under RANSAC constraints. .

[0071] Furthermore, the feature extraction module is used to extract spectral features, texture features, structural features, and ground quadrat features from each integrated ground-air analysis unit. Its output is a multi-dimensional feature vector: Preferred spectral characteristics include: Texture features can be based on gray-level co-occurrence matrix Calculations, for example: Structural features are preferably extracted from CHM, DSM, or point clouds, including mean height, standard deviation, undulation, bare ground percentage, and patch fragmentation. Among these, mean height... with height standard deviation It can be represented as: bare land ratio for: In ground quadrat characteristics, the Shannon-Wiener index can be preferentially used to assess species diversity. in, For the first The relative importance or relative coverage of a species in a quadrat.

[0072] Furthermore, the fusion modeling module is used to establish an ecological parameter inversion model based on ground-space feature vectors. Feature standardization preferably employs range normalization. When the target variable is vegetation cover, average height, biomass, or species richness, a random forest regression model is preferred. The predicted values ​​for random forest are: in, For the number of decision trees, For the first A regressive tree.

[0073] As alternative implementation methods, multiple linear regression, partial least squares regression, support vector regression, or gradient boosting trees can also be used. If support vector regression is used, its optimization objective can be written as: and satisfy - Insensitive constraints.

[0074] The preferred evaluation index for model accuracy includes the coefficient of determination. Root mean square error (RMSE) and mean absolute error (MAE): Furthermore, the ecological assessment module is used to construct a comprehensive evaluation index and classify its levels based on the inversion results. A weighted summation method is preferred for constructing the comprehensive ecological index. : in, The standardized score for each individual indicator. The weights are the indicator weights, which can be determined by the analytic hierarchy process, entropy weighting method, or expert weighting method.

[0075] As one embodiment of the present invention, the result output unit is used to output the cover distribution map, height distribution map, biomass estimation map, diversity spatial distribution map, ecological quality classification map and corresponding statistical reports of the monitoring area, and supports result storage, export and time series comparison analysis.

[0076] This invention employs two main methods: First, it acquires grassland ecological data simultaneously using ground and aerial monitoring units, and establishes an integrated ground-air analysis unit through a unified coordinate benchmark and quadrat mapping method. Second, it maps the boundaries of ground quadrats to UAV orthophotos after positioning them using RTK or high-precision GNSS, achieving a one-to-one correspondence between the measured quadrat data and the aerial image area. Third, it simultaneously extracts aerial remote sensing spectral, texture, and structural features, and integrates them with ground-acquired features such as cover, height, biomass, species composition, diversity, and soil moisture for modeling. Fourth, it uses the fusion model to invert grassland cover, grass height, aboveground biomass, species richness, or diversity indicators, and further constructs comprehensive grassland ecological evaluation indicators. Fifth, it uses the comprehensive evaluation indicators to identify and output grassland ecological quality, degradation degree, growth status, or multi-temporal ecological changes. Sixth, it uses unified quadrats, unified flight routes, unified temporal phases, and unified processing procedures to achieve continuous monitoring and dynamic updating of key grassland growth stages. Example 3 The present invention also provides a grassland ecological monitoring system, comprising: a memory and a processor, wherein the memory stores a computer program executed by the processor, and the computer program performs a grassland ecological monitoring method when executed by the processor.

[0077] Example 4 The present invention also provides a storage medium storing a computer program, which executes a grassland ecological monitoring method when running.

[0078] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A grassland ecological monitoring method, characterized in that, include: Obtain measured data from grassland quadrats and near-surface environmental data; Acquire low-altitude remote sensing images and structural information of grasslands; Based on the measured data of grassland quadrats, near-ground environmental data, low-altitude remote sensing images of grassland, and structural information, spatial registration, feature extraction, data fusion, parameter inversion, and ecological assessment of ground and aerial data are performed. Generate grassland ecological monitoring maps, monitoring reports, and evaluation results.

2. The grassland ecological monitoring method as described in claim 1, characterized in that, Simultaneously acquire grassland ecological data and conduct integrated ground-air analysis using a unified coordinate benchmark and quadrat mapping method; at the same time, map the ground quadrat boundaries to the UAV orthophoto after positioning them using RTK or high-precision GNSS, so as to achieve one-to-one correspondence and registration between the measured data of the quadrat and the aerial image area.

3. The grassland ecological monitoring method as described in claim 1, characterized in that, Simultaneously, aerial remote sensing spectral features, texture features, and structural features are extracted and fused with ground-acquired features such as cover, height, biomass, species composition, diversity, and soil moisture. Furthermore, the fusion model is used to invert grassland cover, grass height, aboveground biomass, species richness, or diversity indicators, and further constructs comprehensive grassland ecological evaluation indicators.

4. A grassland ecological monitoring device, characterized in that, include: Ground monitoring units are used to acquire measured data from grassland quadrats and near-ground environmental data. The aerial monitoring unit is used to acquire low-altitude remote sensing images and structural information of grasslands; The data processing unit is used to perform spatial registration, feature extraction, data fusion, parameter inversion, and ecological assessment of ground and aerial data based on grassland quadrat measured data, near-ground environmental data, low-altitude remote sensing images of grassland, and structural information. The results output unit is used to generate grassland ecological monitoring maps, monitoring reports, and evaluation results.

5. The grassland ecological monitoring device as described in claim 4, characterized in that, Ground monitoring units and aerial monitoring units simultaneously acquire grassland ecological data and conduct integrated ground-air analysis using a unified coordinate benchmark and quadrat mapping method. At the same time, the boundaries of ground quadrats are mapped to UAV orthophotos after being located by RTK or high-precision GNSS, achieving one-to-one correspondence and registration between the measured data of the quadrats and the aerial image area.

6. The grassland ecological monitoring device as described in claim 5, characterized in that, The data processing unit simultaneously extracts aerial remote sensing spectral features, texture features, and structural features, and integrates them with ground-acquired features such as cover, height, biomass, species composition, diversity, and soil moisture. At the same time, it uses a fusion model to invert grassland cover, grass height, aboveground biomass, species richness, or diversity indicators, and further constructs comprehensive grassland ecological evaluation indicators.

7. A grassland ecological monitoring system, characterized in that, include: A memory and a processor, wherein the memory stores a computer program executed by the processor, the computer program performing the grassland ecological monitoring method as described in any one of claims 1-3 when executed by the processor.

8. A storage medium, characterized in that, The storage medium stores a computer program, which executes the grassland ecological monitoring method as described in any one of claims 1-3 when it runs.

Images