A method of monitoring grassland degradation in fusion with animal activity

Abstract

The application discloses a kind of grassland degradation monitoring methods of fusing animal activity, it is related to grassland ecological degradation monitoring technical field, in target area deployment multi-source sensor and animal activity tracking equipment, continuously collect vegetation, soil and in-situ monitoring data of animal activity;Preliminary cleaning and fusion are carried out to in-situ monitoring data by end side equipment, and after compression, it is uploaded to cloud end;In cloud end, integrate multi-time remote sensing image data, generate regional remote sensing-monitoring space-time data set.The application is through end cloud cooperation architecture, in target area deployment multi-source sensor and animal activity tracking equipment, continuously collect vegetation, soil and animal activity data, and combine multi-time remote sensing image, generate regional remote sensing-monitoring space-time data set, effectively solve the problem of discontinuous time-space data in traditional monitoring method, can capture the dynamic process of grassland degradation, provide continuous and consistent monitoring result on space-time, provide data basis for accurate evaluation of grassland degradation degree.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of grassland ecological degradation monitoring, and particularly relates to a grassland degradation monitoring method fusing animal activities. BACKGROUND

[0002] Grassland is an important global ecosystem, which can support a variety of biological habitats and survival. The health of grassland is directly related to biodiversity and ecological balance. Animal activities (such as grazing, migration, etc.) have a direct impact on the structure and function of grassland, which can lead to grassland degradation. Grassland degradation not only affects the health of the ecosystem, but also directly relates to the livelihoods of local residents and agricultural production. Therefore, it is particularly important to monitor the degree of grassland degradation and the recovery effect.

[0003] Since grassland degradation is a dynamic process, it is affected by many factors such as seasonal changes, climate changes, and animal migrations. In the prior art, static or short-period monitoring data is often used, which is difficult to capture the spatiotemporal dynamic characteristics of grassland degradation. In the time dimension, there is a lack of analysis of long-term degradation trends of grassland. In the spatial dimension, the differences of grassland ecosystems in different regions are not fully considered, resulting in a lack of spatiotemporal continuity and comparability of the degradation grade division results, and it is difficult to accurately present the degree of grassland degradation. Therefore, how to use an end-cloud collaborative architecture to process multi-dimensional monitoring data of the target region grassland to construct a grassland degradation model based on spatiotemporal sequences, determine the degradation grade of the target region grassland, and accurately reflect the degree of grassland degradation is a problem to be solved by the present application. Therefore, a grassland degradation monitoring method fusing animal activities is proposed. SUMMARY

[0004] The present application aims to provide a grassland degradation monitoring method fusing animal activities to solve the problems in the background art.

[0005] To solve the above technical problems, the technical solution adopted by the present application is as follows:

[0006] A grassland degradation monitoring method fusing animal activities, comprising the following steps:

[0007] S1, deploying multi-source sensors (vegetation, soil) and animal activity tracking devices in a target region, and continuously collecting in-situ monitoring data of vegetation, soil and animal activities;

[0008] S2, performing preliminary cleaning and fusion of the in-situ monitoring data through an end-side device, and uploading the compressed data to a cloud;

[0009] S3, integrating multi-temporal remote sensing image data in the cloud to generate a regional remote sensing-monitoring spatiotemporal data set;

[0010] S4, fuse remote sensing image data and in-situ monitoring data in the spatio-temporal data set to construct a spatio-temporal feature sequence to quantify the dynamic degradation process;

[0011] S5, combine the spatio-temporal feature sequence, introduce animal activity trajectory and intensity data to form a grassland degradation feature for grassland degradation model construction;

[0012] S6, train a grassland degradation model based on the spatio-temporal sequence in the cloud, and iteratively update the model parameters through real-time feedback data from the terminal device;

[0013] S7, generate a spatio-temporal continuous grassland degradation dynamic atlas based on the grassland degradation model, and output repair suggestions combined with a preset management rule library.

[0014] The further improvement of the technical scheme of the application is that the S1 specifically comprises:

[0015] Based on the topography and ecological zoning of the target area, the spatial layout of the multi-source sensor is planned, the layout points and density of the multi-source sensor and the animal activity tracking device are determined, and a three-dimensional in-situ monitoring network infrastructure covering the whole region is constructed, wherein the multi-source sensor includes a vegetation canopy sensor and a soil erosion monitor;

[0016] According to the set time reference, the sensors and animal activity tracking devices are started synchronously, and in-situ monitoring data including vegetation physiological and ecological parameters, soil physicochemical property data, and animal activity trajectory and intensity data are continuously collected to ensure the synchronization and consistency of the data in the spatio-temporal dimension.

[0017] The further improvement of the technical scheme of the application is that the S2 specifically comprises:

[0018] The original in-situ monitoring data is verified for effectiveness by the terminal device including the multi-source sensor and the animal activity tracking device, abnormal values and noise caused by sensor failure or communication interference are automatically identified and removed, and different sources of data are preliminarily time-stamped and standardized in format;

[0019] With the unified time reference as the anchor point, the in-situ monitoring data that has been processed is associated and integrated to generate a comprehensive data package containing multi-dimensional information;

[0020] The fused data package is end-to-end encrypted and uploaded to the cloud storage cluster through the 5G slice channel, and the cloud receives and performs integrity check and decompression restoration, automatically triggering the data quality secondary evaluation process.

[0021] The further improvement of the technical scheme of the application is that the S3 specifically comprises:

[0022] The cloud automatically accesses a multi-satellite platform, obtains original remote sensing image data of a target area, calls a standardized algorithm process to pre-process the original remote sensing image data, eliminates sensor and environmental errors, and establishes a unified space-time index for all remote sensing image data to form an original remote sensing image data pool organized according to space-time dimensions;

[0023] Based on the pre-processed remote sensing image data, a batch of ecological parameter sequences including a vegetation index, a land surface temperature and soil moisture are generated by inversion, and a time series filtering and reconstruction algorithm is used to smooth and denoise the ecological parameter sequences, so that the ecological parameter sequences obtained by inversion are integrated with the spatial position and time stamp of in-situ monitoring data to form a regional remote sensing-monitoring space-time data set that is continuous in space-time and consistent in scale.

[0024] The further improvement of the technical scheme of the present application is that the S4 specifically comprises:

[0025] The remote sensing image data and in-situ monitoring data are extracted from the regional remote sensing-monitoring space-time data set, and the in-situ monitoring data is spatially interpolated to be scaled to a remote sensing pixel grid based on a unified space-time reference to form a multi-dimensional data cube that is strictly consistent in space-time scale;

[0026] For long-time series data in each remote sensing pixel grid, a signal processing method is used to quantify the dynamic change characteristics, and then trend, seasonal and periodic indexes are extracted, and fluctuation and stability measures are analyzed to depict the degradation trajectory and evolution mode of each ecological parameter of the grassland in the time dimension;

[0027] Fusion of internal and external information in the remote sensing pixel grid constructs a feature set, wherein the lag correlation between each ecological parameter sequence is calculated, the spatial variation characteristics are extracted, and a set of space-time feature sequences containing time dynamics, internal correlation and spatial pattern are generated to quantitatively describe the degradation process.

[0028] The further improvement of the technical scheme of the present application is that the analysis of the fluctuation measures the fluctuation degree of the data relative to the average value through the coefficient of variation;

[0029] The stability measure analyzes the overall stability through the autocorrelation coefficient;

[0030] The lag correlation is used to analyze the lead-lag relationship between different ecological parameters in the same remote sensing pixel grid;

[0031] The spatial variation characteristics quantify the spatial variation degree of the local area around each remote sensing pixel grid by calculating the spatial variation standard deviation, and capture the spatial degradation mode of fragmentation and patchiness.

[0032] The further improvement of the technical scheme of the present application is that the S5 specifically comprises:

[0033] Integrate the spatiotemporal feature sequence and the animal activity trajectory data, unify the spatiotemporal reference, map the animal activity intensity including grazing frequency and migration path density to the remote sensing pixel grid through coordinate conversion and timestamp matching, ensure the accurate alignment of data in the spatiotemporal dimension, and form a multi-dimensional data fusion layer;

[0034] Analyzing the spatial distribution pattern of animal activity trajectory and the spatiotemporal correlation with grassland ecological parameters, determining the lag effect coefficient of animal activity intensity on grassland ecological parameters, extracting animal activity driven animal activity impact features, including ecological parameter degradation rate and spatial heterogeneity enhancement index in the activity intensive area, and then constructing an animal activity impact feature set;

[0035] Fusing the spatiotemporal feature sequence and the animal activity impact feature, generating coupled features through feature intersection, and screening grassland degradation features sensitive to grassland degradation by principal component analysis.

[0036] The further improvement of the technical scheme of the application is that the process of screening grassland degradation features sensitive to grassland degradation by principal component analysis is:

[0037] Fusing the extracted spatiotemporal feature sequence of grassland ecological parameters and animal activity impact features, generating high-dimensional coupled features by feature intersection method, and then constructing a high-dimensional coupled feature set containing original features and intersection features;

[0038] Performing principal component analysis on the high-dimensional coupled feature set, converting the original features to a new set of principal components, and arranging them in descending order according to their variance contribution, retaining the principal components whose cumulative variance contribution rate reaches a preset threshold (85%), while eliminating redundant information and reducing data dimension;

[0039] Analyzing the principal components that reach the preset threshold, determining the grassland degradation features, i.e., ecological parameter degradation rate and animal activity intensity synergy index, for evaluating the driving effect of animal activity on ecological degradation and spatial differentiation characteristics.

[0040] The further improvement of the technical scheme of the application is that S6 specifically includes:

[0041] In combination with the regional remote sensing-monitoring spatiotemporal data set in the historical period, a spatiotemporal sequence grassland degradation model is trained in the cloud, a deep learning framework is adopted, the grassland degradation features are taken as input, and supervised learning is performed with degradation level as label to generate parameters of the initial grassland degradation model and issue them to the end-side device;

[0042] The end-side device uses the deployed initial grassland degradation model to perform online inference on the newly collected real-time data, generates a preliminary degradation state evaluation result, and at the same time, packs the real-time data and the inference result into a feedback data packet and returns the feedback data packet to the cloud end through a secure channel.

[0043] After the cloud end receives the feedback data of the end-side device, an incremental learning process is started, the parameters of the initial grassland degradation model are fine-tuned and optimized by using the new data, the model weight is updated to adapt to the environmental changes, after the training is completed, a new grassland degradation model is formed and is redeployed to the end-side device, and a closed-loop iteration is completed.

[0044] The further improvement of the technical scheme of the present application is that the S7 specifically comprises:

[0045] Based on the iterative optimization of the grassland degradation model, the multi-source spatio-temporal data covered in the regional remote sensing-monitoring spatio-temporal data set is continuously inferred, the grassland degradation features at different time scales are extracted, the high-resolution grassland degradation dynamic atlas covering the target region is generated through spatial interpolation and trend analysis, and the degradation evolution process is quantified.

[0046] The generated grassland degradation dynamic atlas is input into a preset management rule library, combined with the degradation level, spatial distribution and driving factors, and the rule engine is used to automatically match and adapt the repair measures;

[0047] The output structured repair suggestion is synchronized to the end-side device and the cloud end, the rule library parameters are dynamically adjusted according to the field feedback data, and a closed-loop management chain of monitoring-decision-execution-optimization is formed.

[0048] Due to the adoption of the above technical scheme, the present application has the following technical progress compared with the prior art:

[0049] 1. The present application provides a grassland degradation monitoring method fusing animal activities, through an end-cloud collaborative architecture, multi-source sensors and animal activity tracking devices are deployed in a target region, vegetation, soil and animal activity data are continuously collected, and combined with multi-time remote sensing images, a regional remote sensing-monitoring spatio-temporal data set is generated, effectively solving the problem of discontinuous spatio-temporal data in traditional monitoring methods, capturing the dynamic process of grassland degradation, providing continuous and consistent monitoring results in space and time, and providing a data basis for accurately evaluating the degree of grassland degradation.

[0050] 2. The present application provides a grassland degradation monitoring method fusing animal activities, by fusing remote sensing image data and in-situ monitoring data, and introducing animal activity trajectory and intensity data, a spatio-temporal feature sequence containing time dynamics, internal correlation and spatial pattern is constructed, so that the extracted grassland degradation features are more comprehensive and accurate, and the comprehensive influence of animal activities on the grassland ecosystem can be more accurately reflected, and the accuracy and reliability of grassland degradation monitoring are improved. BRIEF DESCRIPTION OF DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments will be briefly introduced as follows. Obviously, the drawings in the following description only represent some embodiments described in the present application, and other drawings can also be obtained by those skilled in the art based on these drawings.

[0052] Figure 1 A working flow diagram of a grassland degradation monitoring method fusing animal activities according to the present application;

[0053] Figure 2 A method flow diagram of a grassland degradation monitoring method fusing animal activities according to the present application. DETAILED DESCRIPTION

[0054] In order to make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0055] Embodiment 1, as shown in the present application, a grassland degradation monitoring method fusing animal activities is provided, comprising the following steps: Figure 1 、 Figure 2 Embodiment 1, as shown in the present application, a grassland degradation monitoring method fusing animal activities is provided, comprising the following steps:

[0056] S1, deploying multi-source sensors (vegetation, soil) and animal activity tracking equipment in a target area, continuously collecting in-situ monitoring data of vegetation, soil and animal activities, planning the spatial layout of multi-source sensors based on the terrain and ecological zoning of the target area, determining the layout points and density of multi-source sensors and animal activity tracking equipment, and constructing a three-dimensional in-situ monitoring network infrastructure covering the whole region, wherein the multi-source sensors include vegetation canopy sensors and soil erosion monitoring instruments, starting each sensor and animal activity tracking equipment synchronously according to the set time reference, continuously collecting in-situ monitoring data including vegetation physiological and ecological parameters, soil physicochemical property data and animal activity trajectory and intensity data, and ensuring the synchronization and consistency of data in the time and space dimensions;

[0057] Specific work content is: based on the target area of the terrain, hydrological characteristics and ecological function zoning, multi-source sensor spatial layout planning, according to the environmental heterogeneity hierarchical sampling layout, in the ecological fragile area encryption deployment of vegetation canopy sensor and soil erosion monitor, to capture the micro-scale degradation process, animal activity tracking equipment layout needs to combine with species habitat preference and migration corridor, preferential in livestock drinking water point, wild animal migration corridor key node deployment GPS collar and infrared camera array, form a three-dimensional in-situ monitoring network infrastructure covering the whole domain; With global navigation satellite system time as the unified reference, configure high-precision clock synchronization module for all sensors and animal activity tracking equipment, ensure that the timestamp error of data collection is less than 1 millisecond, synchronize the start of all deployed sensors and animal activity tracking equipment, start continuous automated data collection work, for the dynamic change characteristics of vegetation physiological and ecological parameters (photosynthetically active radiation, transpiration rate, etc.) and soil physical and chemical properties (humidity, temperature, etc.), set up minute-level high-frequency collection mode; Animal activity trajectory and intensity data use event triggering and periodic sampling combined way, record instantaneous behavior (chewing, running, etc.) and long-time stay trajectory, through edge computing node to the multi-source in-situ monitoring data preliminary time alignment, using sliding window algorithm to detect and correct the clock drift between devices, ensure the consistency of spatio-temporal data on daily, seasonal and interannual scales;

[0058] S2, through the end side device, the preliminary cleaning and fusion of in-situ monitoring data, after compression, upload to the cloud, use the end side device including multi-source sensor and animal activity tracking equipment to verify the effectiveness of the original in-situ monitoring data, automatically identify and eliminate abnormal values and noise caused by sensor failure or communication interference, at the same time, preliminary time stamp alignment and format standardization of data from different sources, take the unified time reference as the anchor point, associate and integrate the in-situ monitoring data that has completed processing, generate comprehensive data package containing multi-dimensional information, encrypt the fused data package end to end, upload to the cloud storage cluster through 5G slice channel, after receiving by the cloud, execute integrity check and decompression restoration, automatically trigger the secondary evaluation process of data quality;

[0059] The specific work content is: through the end-side device including multi-source sensors and animal activity tracking equipment, performing effectiveness verification process on original in-situ monitoring data, using the rule engine built-in the end-side device to detect data range, change rate and logical consistency in real time, automatically identifying abnormal values and noise caused by sensor failure or communication interference, smoothing the noise through sliding window filtering, median filtering and other algorithms, and marking suspicious data segments for subsequent review, on this basis, the end-side device aligns the time stamps of data from different sources according to the unified time protocol, eliminates the time sequence misplacement caused by the difference in collection frequency, and structures and encapsulates according to the standard data format; taking the high-precision time reference provided by the global navigation satellite system as the anchor point, the end-side device performs spatio-temporal correlation and integration on the processed in-situ monitoring data, matches the vegetation physiological and ecological parameters, soil physicochemical property data and animal activity information according to time and spatial coordinates through constructing a data mapping table, generates a comprehensive data package containing multi-dimensional information of ecology-environment-biology, fills in the short missing data in the integration process by using dynamic interpolation algorithm, generates a unique identifier for each data package through hash algorithm to ensure data traceability and non-tamperability, and stores the output data package in encrypted compressed format; the end-side device uploads the encrypted comprehensive data package to the storage cluster in the cloud through the 5G slice network, uses end-to-end encryption and transport layer security protocol to ensure data confidentiality during transmission, the cloud performs integrity verification on the received comprehensive data package, verifies whether the data package is tampered during transmission through digital signature, restores the original data structure by decompression, and then automatically triggers the secondary data quality evaluation process, analyzes the spatio-temporal continuity and logical rationality of the data by using the computing resources of the cloud and combining historical data distribution, generates a quality evaluation report and feeds back to the end-side device, forming a closed-loop optimization mechanism of collection-verification-feedback to ensure the reliability of the final data in the warehouse;

[0060] S3, integrating multi-time sequence remote sensing image data in the cloud to generate a regional remote sensing-monitoring spatio-temporal data set, the cloud automatically accesses multiple satellite platforms to obtain original remote sensing image data of the target region, calls a standardized algorithm process to pre-process the data by radiation calibration, atmospheric correction and geometric registration, eliminates sensor and environmental errors, and establishes a unified spatio-temporal index for all remote sensing image data to form an original remote sensing image data pool organized according to spatio-temporal dimensions, based on the pre-processed remote sensing image data, batch-inversion generates ecological parameter sequences including vegetation index, land surface temperature and soil moisture, and uses time series filtering and reconstruction algorithm to smooth and denoise the ecological parameter sequences, and then integrates the ecological parameter sequences obtained by inversion with the spatial position and time stamp of the in-situ monitoring data to form a regional remote sensing-monitoring spatio-temporal data set with continuous time and space and consistent scale;

[0061] Specific work content is: cloud through the standardized interface automatic access to multi-satellite platform (Landsat, Sentinel, MODIS, etc.), continuous access to the original remote sensing image data of the target area, after data access, call the preset standardization algorithm process to execute radiation calibration (convert DN value to radiance value), atmospheric correction (eliminate aerosol, water vapor and other atmospheric scattering effects) and geometric registration (unified projection coordinate system to WGS84 / UTM, plane accuracy <0.5 pixel), in order to eliminate the error caused by the sensor itself response difference, atmospheric scattering absorption and terrain distortion and other factors, all the preprocessed image data are established according to the time and space range of the unified space-time index, forming the original remote sensing image data pool covering the target area; Based on the preprocessed remote sensing image data, the cloud deployment of automatic inversion engine, batch generation of ecological parameters sequence including vegetation index, land surface temperature and soil moisture, wherein, the inversion algorithm adopts the mixed strategy of physical model (PROSAIL radiation transfer model) and empirical model (TVDI temperature vegetation drought index), combined with in-situ monitoring data to dynamically calibrate model parameters, improve the parameter inversion accuracy, for the generated ecological parameter sequence, the application of time series filtering algorithm to suppress random noise, and using time series reconstruction algorithm (HANTS harmonic analysis) to fill in the missing values caused by cloud cover, ensure the space-time continuity of the ecological parameter sequence, output the ecological parameter time series data set covering the study area, time resolution up to day / week level, the cloud will integrate the ecological parameter sequence obtained by inversion with in-situ monitoring data in space-time, eliminate the difference of sampling frequency, generate a space-time continuous, scale consistent regional remote sensing-monitoring space-time data set;

[0062] S4, fusion of remote sensing image data and in-situ monitoring data in regional remote sensing-monitoring space-time data set, construct time series to quantify dynamic degradation process, extract remote sensing image data and in-situ monitoring data from regional remote sensing-monitoring space-time data set, based on unified space-time reference, implement spatial interpolation of in-situ monitoring data to scale up to remote sensing pixel grid, form a multi-dimensional data cube with strict consistency in space-time scale, for the long time series data in each remote sensing pixel grid, use signal processing method to quantify its dynamic change characteristics, then extract trend, seasonality and periodicity index, and analyze fluctuation and stability measure to describe the degradation trajectory and evolution mode of each ecological parameter of grassland in time dimension, fuse internal and external information of remote sensing pixel grid to construct feature set, wherein, calculate the lag correlation between each ecological parameter sequence, extract spatial variation characteristics in space, and then generate a set of time-space feature sequence containing time dynamics, internal correlation and spatial pattern, to quantify the degradation process comprehensively;

[0063] Fluctuation analysis, the coefficient of variation measures the fluctuation degree of data relative to the average value;

[0064] The calculation expression of coefficient of variation is as follows:

[0065]

[0066] In the formula: CV is the coefficient of variation, which measures the degree of fluctuation of data relative to the mean; σ is the standard deviation of the time series, and μ is the mean of the time series; an increase in CV indicates a decrease in ecosystem stability, greater sensitivity to external disturbances (such as drought and grazing), and a more severe response, which is an important precursor to degradation; a decrease in CV or a stable CV indicates that the ecosystem has strong resistance and less fluctuation in state.

[0067] Stability measurement, analyzing overall stability through autocorrelation coefficient;

[0068] The formula for calculating the autocorrelation coefficient is as follows:

[0069]

[0070] In the formula: R(l) is the autocorrelation coefficient with a lag time of l, l is the lag time step, and x t x t+l Let N be two observations in a time series that are l time steps apart, and N be the length of the time series.

[0071] Lag correlation is used to analyze the lead-lag relationship between different ecological parameters within the same remote sensing pixel grid.

[0072] The expression for calculating lagged correlation is as follows:

[0073]

[0074] ρ X，Y (l)=ρ Y，X (-l), for l<0;

[0075] In the formula: ρ X，Y (l) represents the cross-correlation coefficient between variables X and Y at a lag time of l, i.e., the lagged correlation between ecological parameter sequences; X t Y t For two different ecological parameter sequences within the same remote sensing pixel grid at time t; μ X μ Y ρ represents the mean of variables X and Y. X，Y (l) > 0 indicates a positive correlation, meaning that an increase in one variable tends to be accompanied by an increase in the other variable after a lag time; ρ X，Y (l) < 0 indicates a negative correlation, meaning that an increase in one variable tends to be accompanied by a decrease in the other variable after a lag time; it indicates no linear correlation.

[0076] The spatial variation feature is quantified by calculating the spatial variation standard deviation, which captures the spatial degradation pattern of fragmentation and patchiness.

[0077] The calculation expression of the spatial variation feature is as follows:

[0078]

[0079] In the formula, σ K is the standard deviation of all remote sensing pixel grid values in a specified window; z i is the observation value of the i-th remote sensing pixel grid in the window, is the average value of all remote sensing pixel grid values in the window, and n is the total number of pixels in the window; σ K tends to 0, indicating high homogeneity in the window; σ K increases, indicating high heterogeneity in the window, and the remote sensing pixel grid value changes dramatically. High spatial variation is often a sign of degradation in the middle stage;

[0080] The specific work content is as follows: from the regional remote sensing-monitoring spatio-temporal data set, the remote sensing image data and in-situ monitoring data are extracted, the in-situ monitoring data is spatially interpolated based on the unified spatio-temporal reference, the in-situ monitoring data is upscaled from discrete point observations to continuous remote sensing pixel grids through interpolation algorithms, ensuring that the two are matched in the time and space dimensions, and then a multi-dimensional data cube that is aligned in the time and space dimensions is constructed. The multi-dimensional data cube ensures that each remote sensing pixel grid contains completely matched multi-source observation values at each time node; for the long time series data in each remote sensing pixel grid in the multi-dimensional data cube, signal processing methods are used for quantitative analysis to extract key indicators that characterize the time evolution pattern: including the trend component that reveals the long-term change direction, the seasonal component that reflects the annual regularity change, and the periodic component that captures the longer time period fluctuation. At the same time, the fluctuation and stability measures are analyzed, where the coefficient of variation is used to measure the fluctuation degree of the data relative to the average value, and the autocorrelation coefficient is used to analyze the overall stability. The degradation trajectory and evolution pattern of the health status of the grassland in the time dimension are accurately described in a numerical way by precisely depicting the various ecological parameters of the grassland. The internal and surrounding multi-dimensional information of the remote sensing pixel grid is fused to construct a comprehensive feature set. In the internal of the remote sensing pixel grid, the lag correlation between the sequences of various ecological parameters is calculated to reveal the mutual influence relationship of different parameters in time; in the spatial dimension, the spatial variation feature is extracted to reflect the distribution difference and change of the grassland ecology in space, and then a set of time-space feature sequences containing time dynamics, internal correlation and spatial pattern are generated by combining the internal correlation information and spatial pattern features;

[0081] S5, combine the spatiotemporal feature sequence, introduce the animal activity trajectory and intensity data, and form a grassland degradation feature for constructing a grassland degradation model;

[0082] S6, train the grassland degradation model based on the spatiotemporal sequence in the cloud, and iteratively update the model parameters in real time through the end-side device;

[0083] S7, generate a spatiotemporally continuous grassland degradation dynamic atlas based on the grassland degradation model, and output a repair suggestion combined with a preset management rule library.

[0084] In the embodiment 1, the application provides a technical solution as shown in the embodiment 2: Figure 1 、 Figure 2 S5 specifically includes:

[0085] Integrate the spatiotemporal feature sequence and the animal activity trajectory data, unify the spatiotemporal reference, map the animal activity intensity including the grazing frequency and the migration path density to the remote sensing pixel grid through coordinate conversion and time stamp matching, ensure the accurate alignment of the data in the spatiotemporal dimension, form a multi-dimensional data fusion layer, analyze the spatial distribution pattern of the animal activity trajectory and the spatiotemporal correlation between the animal activity trajectory and the grassland ecological parameter, determine the lag effect coefficient of the animal activity intensity on the grassland ecological parameter, extract the animal activity impact feature driven by the animal activity, cover the ecological parameter degradation rate and the spatial heterogeneity enhancement index of the activity intensive area, and then construct an animal activity impact feature set, fuse the spatiotemporal feature sequence and the animal activity impact feature, generate coupled features through feature intersection, and use principal component analysis to screen the grassland degradation features sensitive to grassland degradation.

[0086] In addition, the process of using principal component analysis to screen the grassland degradation features sensitive to grassland degradation is as follows:

[0087] Fuse the extracted spatiotemporal feature sequence of the grassland ecological parameter and the animal activity impact feature, generate high-dimensional coupled features through feature intersection method, and then construct a high-dimensional coupled feature set containing original features and intersection features, perform principal component analysis on the high-dimensional coupled feature set, convert the original features to a new set of principal components, and arrange them in descending order according to their variance contribution, retain the principal components whose cumulative variance contribution rate reaches a preset threshold (85%), eliminate redundant information, and reduce the data dimension, analyze the principal components that reach the preset threshold, determine the grassland degradation features, i.e., the ecological parameter degradation rate and the animal activity intensity synergy index, and use them to evaluate the driving effect of animal activity on ecological degradation and the spatial differentiation characteristics.

[0088] The calculation expression of the ecological parameter degradation rate is as follows:

[0089]

[0090] where R is the rate of ecological parameter degradation, t j is the time of the jth year, y j is the observed value of the ecological parameter in the jth year, μ is the mean value of the time series, is the mean value of the ecological parameter; R < 0 indicates that the ecological parameter decreases over time, indicating ecosystem degradation; R > 0 indicates that the ecological parameter increases over time, indicating ecosystem recovery; R tends to 0, indicating that the ecological parameter is stable, and the ecosystem is in a balanced state;

[0091] The calculation expression of the animal activity intensity synergy index is as follows:

[0092]

[0093] where AAIS is the animal activity intensity synergy index, used to quantify the degree of synergy between different animal individuals or groups; Q is the number of animal individuals or groups participating in the synergy analysis; q and w represent index variables for traversing all animal individuals or groups, q from 1 to Q-1, w from q+1 to Q, ensuring that each pair of animal individuals or groups is calculated only once, avoiding repeated calculation; r qw is the correlation coefficient between the activity intensity of animal individual q and animal individual w, r qw = 1 indicates complete positive correlation, i.e., the trend of activity intensity change of the two animals is completely consistent, r qw = -1 indicates complete negative correlation, i.e., the trend of activity intensity change of the two animals is completely opposite, r qw = 0 indicates no correlation, i.e., there is no obvious linear relationship between the activity intensity change of the two animals; a q and a w represent the activity intensity values of animal individual q and animal individual w, respectively; |a q -a w | is the absolute difference between the activity intensity values of animal individual q and animal individual w, used to measure the difference in activity intensity between the two animals; max(a q ,a w ) represents taking the maximum value of the activity intensity values of animal individual q and animal individual w, used for normalization processing of the activity intensity difference; is an adjustment factor, used to adjust the influence of activity intensity difference on the synergy index, when the activity intensity difference between the two animals is smaller, the adjustment factor is closer to 1, when the activity intensity difference between the two animals is larger, the adjustment factor is closer to 0; represents the sum of all animal individual pairs The sum of all animal individual pairs activity intensity synergy degrees is obtained by summing up all animal individual pairs is the number of combinations, representing the number of combinations of selecting 2 individuals from Q animal individuals, used to average the sum to obtain the average animal activity intensity synergy index; as the correlation between the activity intensity of animal individuals or groups increases (i.e., r qw tends to 1 or -1), and the difference in activity intensity decreases (i.e. tends to 0, the value of AAIS tends to 1 or -1, indicating that the degree of synergy of animal activity intensity is improved; if the correlation between the activity intensity of animal individuals or groups decreases (i.e., r qw tends to 0), or the difference in activity intensity increases (i.e. tends to 1), the value of AAIS tends to 0, indicating that the degree of synergy of animal activity intensity is reduced;

[0094] The specific work content is: coordinate transformation is performed on the animal activity trajectory data (migration path recorded by the GPS collar, distribution of grazing points, etc.), which is projected from the original coordinate system to the grid coordinate system of the remote sensing image, so that the spatial positions are strictly aligned, at the same time, through timestamp matching, the discrete animal activity observation and the time sequence data of the remote sensing pixel grid are synchronized to the same time resolution, and spatial interpolation is used to map the animal activity intensity to the continuous remote sensing pixel grid, to generate an activity intensity distribution map consistent with the spatial and temporal resolution of the ecological parameters, forming a multi-dimensional data fusion layer containing animal behavior and grassland ecological parameters; based on the multi-dimensional data fusion layer, spatial statistical analysis method is used to describe the spatial distribution characteristics of the animal activity trajectory, to identify the high-frequency activity area and its spatial coupling relationship with the grassland ecological parameters, and through time sequence correlation analysis, to clarify the lag response relationship between the animal activity intensity and the ecological parameters, to determine the key time delay, to further extract the animal activity driving animal activity impact characteristics, including the ecological parameter degradation rate and the spatial heterogeneity enhancement index of the activity intensive area, to construct a feature set reflecting the ecological impact of animal behavior; the spatiotemporal feature sequence and the animal activity impact characteristics are fused, the spatiotemporal feature sequence and the animal activity impact characteristics of the grassland ecological parameters are coupled through feature cross method, to construct a high-dimensional coupling feature set, and principal component analysis is used to reduce the dimension of the coupling features, to screen the grassland degradation features sensitive to grassland degradation, i.e., the ecological parameter degradation rate and the animal activity intensity synergy index, to quantify the synergy effect strength of animal activity and ecological degradation;

[0095] S6 specifically includes:

[0096] In combination with the regional remote sensing-monitoring spatio-temporal dataset of the historical period, the spatio-temporal sequence grassland degradation model is trained in the cloud, the deep learning framework is adopted, the grassland degradation features are taken as the input, the degradation level is taken as the label for supervised learning, the parameters of the initial grassland degradation model are generated and distributed to the end-side device, the end-side device uses the deployed initial grassland degradation model to perform online inference on the newly collected real-time data, generates a preliminary degradation state evaluation result, and meanwhile, the real-time data and the inference result are packaged into a feedback data packet and returned to the cloud through a secure channel, after the cloud receives the feedback data from the end-side device, an incremental learning process is started, the new data is used to fine-tune and optimize the parameters of the initial grassland degradation model, the model weight is updated to adapt to the environmental changes, after the training is completed, a new grassland degradation model is formed and redeployed to the end-side device, and the closed-loop iteration is completed;

[0097] Specific work content is: the cloud constructs a regional remote sensing-monitoring spatio-temporal dataset based on the historical period, adopts a deep learning framework to train a model, takes the extracted grassland degradation features as the input, takes the degradation level determined by expert interpretation as the label, performs supervised learning, adjusts the network weight through optimization of the loss function during the training process, generates a parameter set of the initial grassland degradation model, captures the nonlinear relationship between the ecological parameter evolution law and the influence of animal activities, after the training is completed, the cloud distributes the model parameters and structure files to the computing nodes of the end-side device through a secure channel; after the end-side device receives the initial grassland degradation model, the model is loaded into the built-in inference engine, online analysis is performed on the real-time collected data covered by the regional remote sensing-monitoring spatio-temporal dataset, data preprocessing, feature extraction and degradation level inference are automatically performed, the grassland degradation state evaluation result at the current time is generated, meanwhile, the end-side device adds a timestamp and a device identifier to the complete data packet containing the original data, the preprocessed features and the inference result, and then returns the data packet to the storage cluster of the cloud through an encrypted communication link; after the cloud receives the data packet fed back by the end-side device, an incremental learning process is started, the feedback data is quality checked and labeled verified, the feedback data is added to the training sample set, the incremental learning algorithm is adopted to fine-tune and optimize the parameters of the initial grassland degradation model, the network layer weight sensitive to environmental changes in the model is adjusted, the grassland degradation model is adapted to the factors such as vegetation phenology change, climate change and grazing strategy adjustment, a new version of the grassland degradation model is generated after the training is completed, and the model is redeployed to the end-side device after performance evaluation, and the model iteration upgrade is completed;

[0098] S7 specifically includes:

[0099] Based on the iterative optimization of the grassland degradation model, the multi-source spatio-temporal data in the regional remote sensing-monitoring spatio-temporal dataset is continuously inferred, the grassland degradation features at different time scales are extracted, the high-resolution grassland degradation dynamic atlas covering the target region is generated through spatial interpolation and trend analysis, the degradation evolution process is quantified, the generated grassland degradation dynamic atlas is input into the preset management rule library, the degradation level, spatial distribution and driving factors are combined, the adaptive repair measures are automatically matched through the rule engine, the structured repair suggestions are output, and the management chain of monitoring-decision-execution-optimization is formed.

[0100] The specific work content is: based on the iterative optimization of the grassland degradation model, the multi-source spatio-temporal data in the regional remote sensing-monitoring spatio-temporal dataset is continuously inferred, the grassland degradation features at different time scales are extracted through the deep learning framework, the data gaps are filled by using the spatial interpolation technology, the degradation evolution rate is quantified by combining the trend analysis, the high-resolution grassland degradation dynamic atlas covering the target region is generated, the spatial distribution pattern of grassland degradation and the nonlinear law of its evolution with time are reflected; the generated grassland degradation dynamic atlas is input into the preset management rule library, the rule library integrates the correlation rules of degradation level division standard, spatial distribution characteristics and driving factors, the pattern matching is performed on the atlas data through the rule engine, the degradation hot spot area and its dominant factor are automatically identified, and the adaptive repair measures are deduced according to the preset logic, such as, for the light degradation area, the rule engine recommends rotational grazing or improved reseeding; for the severe degradation area, the intervention strategy such as fence enclosure or artificial grass planting is triggered, and the output result is a structured repair suggestion list, which clearly shows the measure type and implementation range, and is synchronously pushed to the cloud and the terminal device; the closed-loop management chain of monitoring-decision-execution-optimization is established, the implementation effect data is collected in real time after the terminal device executes the repair measures, and is transmitted back to the cloud through the encrypted link, the feedback data is quality checked in the cloud, the measure effect label is corrected by expert evaluation, the rule library parameters are dynamically updated, and the grassland degradation model is iteratively optimized based on the new data, so that the feature extraction accuracy and the trend prediction ability are improved.

[0101] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed in the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for monitoring grassland degradation that integrates animal activity, characterized in that, Includes the following steps: S1. Deploy multi-source sensors and animal activity tracking devices in the target area to continuously collect in-situ monitoring data on vegetation, soil and animal activity; S2. Perform preliminary cleaning and fusion of in-situ monitoring data through end-side equipment, compress the data, and upload it to the cloud. S3. Integrate multi-time series remote sensing image data in the cloud to generate regional remote sensing-monitoring spatiotemporal datasets, including: automatically accessing multiple satellite platforms in the cloud to obtain raw remote sensing image data of the target area, calling standardized algorithm processes to perform radiometric calibration, atmospheric correction and geometric registration preprocessing on the data, and batch inverting and generating ecological parameter sequences including vegetation index, surface temperature and soil moisture based on the preprocessed remote sensing image data. S4. Integrate remote sensing imagery and in-situ monitoring data from the regional remote sensing-monitoring spatiotemporal dataset to construct a spatiotemporal feature sequence to quantify the dynamic degradation process, specifically including: Remote sensing image data and in-situ monitoring data are extracted from the regional remote sensing-monitoring spatiotemporal dataset. Based on a unified spatiotemporal benchmark, spatial interpolation is performed on the in-situ monitoring data to scale it up to the remote sensing pixel grid, forming a multidimensional data cube with strict spatiotemporal scale consistency. For long-term data within each remote sensing pixel grid, signal processing methods are used to quantify its dynamic change characteristics, thereby extracting trend, seasonal and periodic indicators, and analyzing volatility and stability measures to characterize the degradation trajectory and evolution pattern of various grassland ecological parameters in the time dimension. A feature set is constructed by integrating internal and external information of remote sensing pixel grids. Internally, the lag correlation between the sequences of various ecological parameters is calculated, and spatial variation features are extracted spatially. This generates a set of spatiotemporal feature sequences that simultaneously contain temporal dynamics, internal correlations, and spatial patterns. The lag correlation is used to analyze the lead-lag relationship between different ecological parameters within the same remote sensing pixel grid. S5. Combining spatiotemporal feature sequences, animal activity trajectory and intensity data are introduced to form grassland degradation features for grassland degradation model construction; S6. Train a grassland degradation model based on spatiotemporal sequences in the cloud and iteratively update the model parameters through real-time data feedback from edge devices; S7. Generate a spatiotemporally continuous dynamic map of grassland degradation based on the grassland degradation model, and output restoration suggestions in combination with the preset management rule base.

2. The grassland degradation monitoring method integrating animal activity according to claim 1, characterized in that: S1 specifically includes: Based on the topography and ecological zoning of the target area, the spatial layout of multi-source sensors is planned, the deployment points and density of multi-source sensors and animal activity tracking devices are determined, and a three-dimensional in-situ monitoring network infrastructure covering the entire area is constructed. Among them, multi-source sensors include vegetation canopy sensors and soil erosion monitors. According to the set time benchmark, all sensors and animal activity tracking devices are activated simultaneously to continuously collect in-situ monitoring data, including vegetation physiological and ecological parameters, soil physicochemical properties data, and animal activity trajectory and intensity data.

3. The grassland degradation monitoring method integrating animal activity according to claim 1, characterized in that: S2 specifically includes: The validity of raw in-situ monitoring data is verified by using end-side devices including multi-source sensors and animal activity tracking devices. It automatically identifies and removes outliers and noise caused by sensor failure or communication interference, and performs preliminary timestamp alignment and format standardization on data from different sources. Using a unified time reference as an anchor point, the processed in-situ monitoring data are correlated and integrated to generate a comprehensive data package containing multi-dimensional information; The merged data packets are encrypted end-to-end and uploaded to the cloud storage cluster via the 5G slicing channel. After receiving the data, the cloud performs integrity verification and decompression and restoration, automatically triggering a secondary data quality assessment process.

4. The grassland degradation monitoring method integrating animal activity according to claim 3, characterized in that: S3 further includes: Establish a unified spatiotemporal index for all remote sensing image data to form a pool of raw remote sensing image data organized according to spatiotemporal dimensions; Temporal filtering and reconstruction algorithms are used to smooth and denoise the ecological parameter sequence. Then, the inverted ecological parameter sequence is integrated with the spatial location and timestamp of the in-situ monitoring data to form a spatiotemporally continuous and scale-consistent regional remote sensing-monitoring spatiotemporal dataset.

5. The grassland degradation monitoring method integrating animal activity according to claim 1, characterized in that: The volatility analysis measures the degree of data fluctuation relative to the mean using the coefficient of variation. The stability measure is analyzed using the autocorrelation coefficient to assess overall stability. The spatial variation feature quantifies the degree of spatial variation in the local area surrounding each remote sensing pixel grid by calculating the standard deviation of spatial variation, and captures spatial degradation patterns of fragmentation and patchiness.

6. The grassland degradation monitoring method integrating animal activity according to claim 1, characterized in that: S5 specifically includes: By integrating spatiotemporal feature sequences and animal activity trajectory data, and unifying the spatiotemporal benchmark, animal activity intensity, including grazing frequency and migration path density, is mapped to remote sensing pixel grids through coordinate transformation and timestamp matching, forming a multi-dimensional data fusion layer. This study analyzes the spatial distribution patterns of animal activity trajectories and their spatiotemporal correlation with grassland ecological parameters, clarifies the lagged impact coefficient of animal activity intensity on grassland ecological parameters, extracts the characteristics of animal activity-driven impacts, covers the degradation rate of ecological parameters and the spatial heterogeneity enhancement index in densely populated areas, and then constructs a set of animal activity impact characteristics. By integrating spatiotemporal feature sequences with animal activity impact features, coupled features are generated through feature cross-referencing, and principal component analysis is used to screen grassland degradation features that are sensitive to grassland degradation.

7. The grassland degradation monitoring method integrating animal activity according to claim 6, characterized in that: The process of using principal component analysis to screen grassland degradation features that are sensitive to grassland degradation is as follows: The spatiotemporal feature sequences of extracted grassland ecological parameters are fused with animal activity impact features, and a feature cross-method is used to generate high-dimensional coupled features, thereby constructing a high-dimensional coupled feature set containing original features and cross-features. Principal component analysis is performed on the high-dimensional coupled feature set to transform the original features into a new set of principal components, which are then sorted in descending order of their variance contribution. Principal components whose cumulative variance contribution rate reaches a preset threshold are retained. Principal components that reach a preset threshold are analyzed to determine grassland degradation characteristics, namely the ecological parameter degradation rate and the synergistic index of animal activity intensity, which are used to assess the driving effect of animal activity on ecological degradation and spatial differentiation characteristics.

8. The grassland degradation monitoring method integrating animal activity according to claim 7, characterized in that: S6 specifically includes: By combining historical regional remote sensing-monitoring spatiotemporal datasets, a spatiotemporal sequence-based grassland degradation model is trained in the cloud. A deep learning framework is used, with grassland degradation features as input and degradation level as label for supervised learning, to generate parameters for the initial grassland degradation model and distribute them to the edge device. The edge device uses the deployed initial grassland degradation model to perform online inference on the new data collected in real time, and generates preliminary degradation status assessment results. At the same time, the real-time data and its inference results are packaged into a feedback data packet and transmitted back to the cloud through a secure channel. After receiving feedback data from the edge devices, the cloud initiates an incremental learning process. It uses the new data to fine-tune and optimize the parameters of the initial grassland degradation model, updates the model weights to adapt to environmental changes, and after training, forms a new grassland degradation model and redeploys it to the edge devices, completing the closed-loop iteration.

9. The grassland degradation monitoring method integrating animal activity according to claim 8, characterized in that: Specifically, S7 includes: Based on the iteratively optimized grassland degradation model, continuous inference is performed on the multi-source spatiotemporal data covered in the regional remote sensing-monitoring spatiotemporal dataset to extract grassland degradation features at different time scales. Through spatial interpolation and trend analysis, a high-resolution dynamic map of grassland degradation covering the target area is generated to quantify the degradation evolution process. The generated grassland degradation dynamic map is input into the preset management rule base. Combined with the degradation level, spatial distribution and driving factors, the rule engine automatically matches the appropriate restoration measures. It outputs structured repair suggestions, synchronizes them to edge devices and the cloud, and dynamically adjusts the rule base parameters based on on-site feedback data, forming a closed-loop management chain of monitoring-decision-execution-optimization.

Images