A land ecological quality evaluation method and system based on linkage data collection

Abstract

This invention discloses a land ecological quality assessment method and system based on linked data acquisition, belonging to the field of ecological environment monitoring and management technology. It utilizes satellite remote sensing imagery to construct macroscopic ecological characteristics, identifies abnormal areas through time series decomposition, and calculates confidence levels. When the confidence level exceeds a threshold, it automatically triggers a drone to perform dynamic path planning and high-precision scanning, real-time calculation of canopy structure and texture features. If the drone determines that vegetation stress exists, it directionally wakes up ground sensors or dispatches robots to collect microscopic physicochemical data. It performs spatiotemporal registration of multi-source data from air, space, and ground, dynamically adjusts the weights of the evaluation model by calculating inter-scale deviations, and uses microscopic measured data to inversely calibrate remote sensing inversion parameters. This invention solves the problems of lack of linkage between different levels and difficulty in correcting conflicts between multi-source data in traditional monitoring, achieving low-cost, high-precision, and evolvable land ecological quality assessment and degradation attribution diagnosis.

Description

Technical Field

[0001] This invention belongs to the field of ecological quality assessment and management technology, and specifically relates to a land ecological quality assessment method and system based on linked data collection. Background Technology

[0002] With increasing global pressure on the ecological environment, land ecological quality monitoring has become a core component of land spatial planning, natural resource management, and ecological restoration projects. Particularly in mining reclamation areas, farmland protection zones, and ecologically fragile areas, timely and accurate understanding of land ecological conditions and their dynamic trends is crucial for preventing ecological risks and formulating restoration strategies. Traditional monitoring methods, primarily relying on manual patrols, are inefficient and unable to cover large areas, failing to meet the demands of modern ecological management for high timeliness and precision.

[0003] Current land ecological quality assessment technologies mainly fall into two categories: one is large-scale monitoring based on satellite remote sensing, which uses multispectral imagery to calculate indices such as NDVI and RSEI, suitable for large-scale macro-trend analysis; the other is micro-level monitoring based on ground sensors or manual sampling, which uses IoT nodes to acquire physicochemical parameters such as soil temperature, humidity, and pH, offering high accuracy but extremely limited coverage. Some advanced solutions attempt to introduce drones to supplement meso-scale monitoring, forming an integrated "air-space-ground" monitoring architecture. However, these existing technologies mostly operate independently at each level, with data collection typically following a fixed schedule (such as satellite transits and timed data transmissions), lacking real-time interaction and collaboration between data at different levels.

[0004] The main pain points of existing technologies lie in the lack of linkage and adaptability; specifically: 1. Satellites cannot automatically trigger subsequent high-precision verification after detecting anomalies, resulting in spatiotemporal lag in monitoring; 2. Drones and ground sensors usually perform fixed tasks and cannot dynamically adjust the acquisition strategy according to the degree of anomaly, resulting in energy waste or missing key data; 3. Multi-source data fusion often uses static weights. When macroscopic inversion and microscopic measurement conflict, the evaluation model cannot automatically correct itself, which can easily lead to distorted evaluation results and make it difficult to accurately locate the root cause of ecological degradation. Summary of the Invention

[0005] (a) Technical problems to be solved To address the problems in related technologies, this invention provides a land ecological quality assessment method and system based on linked data collection, thereby overcoming the aforementioned technical problems existing in existing related technologies.

[0006] (II) Technical Solution To solve the above-mentioned technical problems, the present invention is achieved through the following technical solution: S1. Acquire satellite remote sensing images of the area to be evaluated and construct macro-ecological characteristics; based on the macro-ecological characteristics, identify suspected abnormal areas and calculate the anomaly confidence index of the suspected abnormal areas. S2. When the anomaly confidence index exceeds the linkage threshold, the scanning path is dynamically selected based on the spatial set characteristics of the suspected anomaly area, and a drone inspection task instruction is generated. S3. The drone scans the suspected abnormal area according to the drone inspection task instructions in S2; it calculates the collected data in real time, extracts the canopy structure parameters and surface texture features, and calculates the vegetation stress index and structural damage degree to obtain the calculation results; if the calculation results meet the micro-awakening conditions, a micro-sensing awakening signal is generated. S4 and broadcast S3 generate micro-sensing wake-up signals to directionally activate ground equipment located in suspected abnormal areas, collect soil physicochemical property parameters and micro-environment meteorological parameters, and obtain micro-measured data; S5. Perform spatiotemporal registration and normalization on the macroscopic ecological characteristics of S1, the canopy structure and texture characteristics of S3, and the microscopic measured data of S4; calculate the deviation measure between data at different scales. The weight coefficients of each data source in the ecological quality evaluation model are dynamically adjusted according to the deviation measurement, and the land ecological quality evaluation results are output. The ecological quality evaluation model includes a weighted ecological quality evaluation equation and a weight adjustment rule. The weight adjustment rule includes: if the inter-scale consistency deviation is greater than the tolerance limit threshold, the macro weight coefficient is reduced according to the exponential decay law, and the micro weight coefficient is increased accordingly. Preferably, step S1 includes the following steps: S11. Acquire multispectral satellite images of the area to be evaluated for the most recent N periods, perform radiometric calibration and atmospheric correction to obtain corrected image data; based on the corrected image data, calculate the normalized vegetation index, surface temperature and remote sensing ecological index to obtain macro-ecological characteristics. S12. Use the STL algorithm to decompose the time series of macro-ecological characteristics into seasonal, trend, and residual terms; for the trend term, calculate the deviation between the current period value and the historical average for the same period. S13. If the absolute value of the deviation is greater than the threshold multiple of the benchmark standard deviation. k If the pixel is identified as an anomaly, then a density clustering algorithm is used to spatially cluster the anomaly points to generate continuous suspected anomaly regions. S14. Calculate the mean of the deviation of all pixels within the ROI, and obtain the anomaly confidence index after normalization. Preferably, step S2 includes the following steps: S21. Set the linkage threshold; if the anomaly confidence index of the suspected anomaly area obtained in S1 is greater than the linkage threshold, then activate the path planning program. S22. After activating the path planning program, extract the coordinates of the polygon boundary vertices of the suspected abnormal region ROI; calculate the aspect ratio and area of ​​the suspected abnormal region; if the aspect ratio is less than the ratio threshold and the area is less than the area threshold, use a spiral centripetal scanning path; if the aspect ratio is greater than the ratio threshold or the area is greater than the area threshold, use a parallel reciprocating scanning path. S23. Calculate the required resolution based on the anomaly confidence index of suspected anomaly areas; calculate the optimal flight altitude based on the required resolution. S24. Generate UAV inspection mission instructions that include a comprehensive flight path point set, flight altitude parameters, and sensor exposure strategies. Preferably, step S3 includes the following steps: S31. The drone arrives over the suspected abnormal area ROI and uses a hyperspectral camera to collect spectral data within the wavelength range and a lidar to acquire point cloud data. S32, the airborne edge computing module performs first-order differential processing on the hyperspectral data, extracts the red edge position, and obtains the surface texture features; it uses lidar point cloud to construct a canopy height model, calculates the canopy statistical height and canopy porosity, and obtains the canopy structure parameters; S33. Based on canopy structure parameters and surface texture characteristics, calculate the vegetation stress index (VSI) and structural damage degree; If the vegetation stress index is greater than the vegetation stress index threshold or the structural damage degree is greater than the structural damage degree threshold, then it is determined that there is substantial ecological stress on the land surface, and a micro-sensing wake-up signal containing the ROI center coordinates and instruction type is generated. Preferably, step S4 includes the following steps: S41. Ground sensor nodes are pre-installed with LoRaWAN communication modules and are in heartbeat mode by default, sending a live heartbeat packet every 24 hours. S42. When the node receives the micro-sensing wake-up signal from S3 and its own GPS coordinates are within the suspected abnormal region ROI range, the node immediately parses the instruction type in the signal. In specific implementation, preferably, S42 specifically refers to: the LoRa gateway carried by the drone flying over the ROI A The wake-up frame is broadcast downwards; located in the ROI. A The internal sensor, designated Node 2025, received the signal, compared its GPS coordinates to confirm that it was within the target range, and was then activated. S43. Adjust the data collection method within the suspected abnormal ROI area where there are no fixed sensors according to the instruction type; if there are no fixed sensor nodes within the suspected abnormal ROI area, the system dispatches a nearby mobile sampling robot to the center coordinates of the ROI to perform soil sampling and portable XRF spectral analysis. Preferably, step S5 includes the following steps: S51. Using the Kriging interpolation method, the macroscopic ecological characteristics of S1, the canopy structure and texture characteristics of S3, and the microscopic measured data of S4 are resampled to a unified standard evaluation grid to obtain a multidimensional ecological data cube. S52. Construct an initial weighted ecological quality evaluation equation; read the sub-indicator data in the multi-dimensional ecological data cube and calculate the initial ecological quality index through the weighted ecological quality evaluation equation; the initial weighted ecological quality evaluation equation includes macro-weight coefficients, canopy structure and texture weight coefficients, and micro-weight coefficients; S53. Based on the sub-indicator data extracted in S52, calculate the inter-scale consistency deviation; if the inter-scale consistency deviation is less than or equal to the tolerance limit threshold, then the initial ecological quality index is used as the final ecological quality index. Otherwise, the weight coefficients in the initial weighted ecological quality evaluation equation are reduced according to the exponential decay law to obtain the adjusted weighted ecological quality evaluation equation; the ecological quality index is then calculated again using the adjusted weighted ecological quality evaluation equation to obtain the final ecological quality index. S54. Based on the final ecological quality index and the data source type of the factor with the highest weight, generate labels for the main causes of degradation to obtain the land ecological quality evaluation results. Preferably, step S53, which involves reducing and adjusting the weight coefficients in the initial weighted ecological quality evaluation equation according to an exponential decay law to obtain the adjusted weighted ecological quality evaluation equation, further includes a parameter reverse calibration step: S531. If the micro-weight coefficient in the adjusted weighted ecological quality evaluation equation is greater than the micro-weight coefficient threshold, then the macro-remote sensing inversion parameters are reverse-calibrated using micro-measured data, and the updated parameters will be used in step S1 of the next cycle. Preferably, in step S54, generating labels for the main causes of degradation based on the data source type of the factor with the highest weight, and obtaining the land ecological quality assessment results includes the following steps: S541. Construct an attribution feature vector that includes water stress feature components, pollution stress feature components, and physical damage feature components; S542. If the soil moisture content is lower than the wilting coefficient, the water stress characteristic component is set to 1; if the heavy metal ion concentration exceeds the standard or the pH value is abnormal, the pollution stress characteristic component is set to 1; if the surface roughness increases suddenly and the vegetation height variance exceeds the vegetation height variance threshold, it is set to 1. S543. Use a decision tree classifier to classify the attribution feature vectors and output diagnostic conclusions for drought, water shortage, chemical pollution, physical damage, or combined degradation. Preferably, a land ecological quality assessment system based on linked data acquisition implements a land ecological quality assessment method based on linked data acquisition, the system comprising: Macro monitoring module: used to receive and process satellite remote sensing images, identify suspected abnormal areas and calculate anomaly confidence; Joint Control Center: Includes an intelligent scheduling engine, used to receive signals from the macro monitoring module and generate UAV mission instructions and ground wake-up signals; Airborne data acquisition subsystem: It consists of a UAV carrying hyperspectral and lidar payloads and an airborne edge computing unit, and has the ability to fly autonomously and perform real-time data screening. Ground-based sensing subsystem: includes a distributed low-power IoT sensor node cluster and a mobile sampling robot, with a wake-up mechanism and multi-parameter acquisition capabilities; Comprehensive evaluation cloud platform: used to store multi-source heterogeneous data, run dynamic weight correction models and parameter inverse calibration algorithms, and generate ecological quality evaluation reports.

[0007] Preferably, the linkage control center further includes: Communication gateway: Supports heterogeneous network convergence access of satellite communication, 4G / 5G network and LoRaWAN narrowband IoT protocol; Digital Twin Engine: Utilizing data collected from S1 to S4, an ecological model of the area to be evaluated is constructed in real time in virtual space, which is used to visualize the ecological degradation process and simulate the effects of restoration plans.

[0008] (III) Beneficial Effects The present invention has the following beneficial effects: This invention changes the traditional model of independent data acquisition at each level by constructing a cascaded triggering mechanism of macroscopic anomaly detection, mesoscopic path planning, and microscopic sensing activation. The system only dispatches UAVs when satellites detect significant anomalies, and only activates ground sensors when the UAVs confirm the presence of stress. This on-demand response strategy significantly reduces ineffective UAV flights and standby power consumption of ground equipment, resolves the contradiction between cost and accuracy in large-area monitoring, and achieves optimal allocation of monitoring resources.

[0009] This invention proposes a dynamic weight correction algorithm based on inter-scale consistency deviation. During the evaluation process, the consistency between satellite inversion data and ground-based measured data is calculated in real time. When significant deviations occur, the weight of macroscopic data is automatically reduced and the weight of microscopic measured data is increased to ensure that the evaluation results always approximate the true values. By using the mechanism of reverse calibration of satellite inversion parameters using microscopic true values, the system has self-learning capabilities, and the accuracy of macroscopic monitoring will continuously improve over time.

[0010] Unlike traditional methods that only output a general ecological score, this invention constructs an attribution diagnostic model by integrating multi-dimensional features such as spectral, structural, and physicochemical data. The system can automatically distinguish different types of degradation causes, such as drought and water shortage, chemical pollution, and physical damage, and generate intelligent reports that include pollution source location and remediation suggestions. This provides land managers with support from problem discovery to cause diagnosis and decision support, greatly improving the scientific nature and pertinence of ecological governance.

[0011] Of course, any product implementing this invention does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0012] To more clearly illustrate the technical solutions of the embodiments of the invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, the drawings can be obtained from these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating a land ecological quality assessment method based on linked data acquisition according to the present invention. Figure 2 This is a schematic diagram of a land ecological quality evaluation system based on linked data acquisition according to the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the invention, and not all embodiments. Based on the embodiments of the invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the invention.

[0015] To address the technical problems raised in the background section, please refer to [link / reference]. Figure 1 This invention provides a method and system for evaluating land ecological quality based on linked data collection, comprising: S1. Acquire satellite remote sensing images of the area to be evaluated and construct macro-ecological characteristics; based on the macro-ecological characteristics, identify suspected abnormal areas and calculate the anomaly confidence index of the suspected abnormal areas. S2. When the anomaly confidence index exceeds the linkage threshold, the scanning path is dynamically selected based on the spatial set characteristics of the suspected anomaly area, and a drone inspection task instruction is generated. S3. The drone scans the suspected abnormal area according to the drone inspection task instructions in S2; it calculates the collected data in real time, extracts the canopy structure parameters and surface texture features, and calculates the vegetation stress index and structural damage degree to obtain the calculation results; if the calculation results meet the micro-awakening conditions, a micro-sensing awakening signal is generated. S4 and broadcast S3 generate micro-sensing wake-up signals to directionally activate ground equipment located in suspected abnormal areas, collect soil physicochemical property parameters and micro-environment meteorological parameters, and obtain micro-measured data; S5. Perform spatiotemporal registration and normalization on the macroscopic ecological characteristics of S1, the canopy structure and texture characteristics of S3, and the microscopic measured data of S4; calculate the deviation measure between data at different scales. The weight coefficients of each data source in the ecological quality evaluation model are dynamically adjusted according to the deviation measurement, and the land ecological quality evaluation results are output. The ecological quality evaluation model includes a weighted ecological quality evaluation equation and a weight adjustment rule. The weight adjustment rule includes: if the inter-scale consistency deviation is greater than the tolerance limit threshold, the macro weight coefficient is reduced according to the exponential decay law, and the micro weight coefficient is increased accordingly. The above embodiments utilize satellite remote sensing imagery to construct macroscopic ecological features, identify abnormal areas and calculate confidence levels through time series decomposition; when the confidence level exceeds a threshold, the UAV is automatically triggered to perform dynamic path planning and high-precision scanning, and canopy structure and texture features are calculated in real time; if the UAV determines that vegetation stress exists, it further activates ground sensors or dispatches robots to collect microscopic physicochemical data; the air-space-ground multi-source data are spatiotemporally registered, the weights of the evaluation model are dynamically adjusted by calculating the scale deviation, and the remote sensing inversion parameters are calibrated inversely using microscopic measured data; this solves the problems of lack of linkage between different levels and difficulty in correcting conflicts between multi-source data in traditional monitoring.

[0016] The above embodiment S1 includes the following steps: S11. Acquire multispectral satellite images of the area to be evaluated for the most recent N periods, perform radiometric calibration and atmospheric correction to obtain corrected image data; based on the corrected image data, calculate the Normalized Difference Vegetation Index (NDVI), Land Surface Temperature (LST), and Remote Sensing Ecological Index (RSEI) to obtain macro-ecological characteristics. In specific implementation, the above embodiment S11 is as follows: In this embodiment, a large mining area reclamation site is selected as the area to be evaluated; the system automatically downloads the Sentinel-2 multispectral images and Landsat-8 thermal infrared images of the most recent 3 years (N=108, one period every 10 days) from the data center; atmospheric correction is performed on the images using the 6S atmospheric radiative transfer model to eliminate the influence of aerosols and water vapor and obtain the true surface reflectance; NDVI = (NIR - Red) / (NIR + Red) is calculated, where NIR represents the surface reflectance in the near-infrared band and Red represents the surface reflectance in the red band. Based on the corrected thermal infrared brightness temperature, the surface emissivity estimated from multispectral data, and atmospheric water vapor content data, the surface temperature LST (thermal component) is retrieved using a single-window algorithm. Principal component analysis (PCA) is used to fuse NDVI, Wet (humidity component), LST, and NDBSI (dryness component), and the first principal component is extracted as the remote sensing ecological index RSEI to obtain a macro-ecological feature dataset. S12. Use the STL algorithm to decompose the time series of macro-ecological characteristics into seasonal, trend, and residual terms; for the trend term, calculate the deviation between the current period value and the historical average for the same period. In specific implementation, the above embodiment S12 is as follows: Considering that vegetation growth has obvious seasonality, direct comparison of values ​​is prone to misjudgment; the system performs RSEI time series analysis on each pixel. Y t Perform STL decomposition: Y t = S t + T t + R t ;in, S t This is a seasonal cycle item (reflecting natural phenology). T t This represents the long-term trend (reflecting substantial changes in ecological quality). R t This represents the residual term (random noise); in this embodiment, the Loess smoothing window is set to 13 (corresponding to quarterly variations), based on the trend term of seasonal fluctuations. T t Extract the current trend item T current ; Calculate the deviation D =( T current - m history ) / s history ;in, m history This represents the average of the trend items for the same period in history. s history This represents the standard deviation for the same period in history. S13. If the absolute value of the deviation is greater than the threshold multiple of the benchmark standard deviation. k If the pixel is identified as an anomaly, then a density clustering algorithm is used to spatially cluster the anomaly points to generate continuous suspected anomaly regions. In specific implementation, the above embodiment S13 specifically involves: setting k =2.5 (i.e., confidence level 98.7%); according to the 3σ rule of normal distribution, in a normal distribution, the probability interval covered by 2.5σ is approximately 98.76%; setting k=2.5 means only filtering out extreme deviation events with a probability less than 1.24%, which can effectively filter out slight noise caused by weather fluctuations or measurement errors, ensuring that areas identified as abnormal have a very high probability of true degradation; if a certain pixel D A value less than -2.5 indicates a significant non-seasonal degradation of the ecological quality at that point, and is marked as an anomalous pixel. This embodiment utilizes the DBSCAN algorithm, setting a search radius... E ps =60 meters, minimum number of points Min Pts =10; the clustering results identified 3 independent suspected anomalous regions (ROIs). A ROI B ROI C ); S14. Calculate the mean of the deviation of all pixels within the ROI, and obtain the anomaly confidence index after normalization. In specific implementation, the above embodiment S14 specifically involves: calculating the ROI of suspected abnormal regions respectively. A ROI B ROI C The abnormal confidence index; with ROI A For example, ROI A abnormal confidence index Where M represents the suspected anomaly region ROI. A The total number of pixels in the middle, D i Indicates the first i The deviation of each pixel; the total number of pixels within the ROI; calculated in this embodiment. C roiA =3.2; This index directly quantifies the severity of regional ecological degradation, providing a basis for coordinated decision-making in the subsequent S2 steps; The above-described embodiment S1 constructs a macro-monitoring network covering the entire region by integrating long-term satellite remote sensing imagery with the STL time-series decomposition algorithm. Its core advantage lies in its ability to isolate seasonal disturbances caused by natural phenology and accurately identify ecological quality trends caused by non-natural factors. Compared to traditional single-period image comparison methods, this method significantly reduces the misjudgment rate caused by seasonal phase differences, achieving low-cost, high-confidence initial screening of ecological anomalies over large areas, and providing precise spatial guidance for subsequent high-cost UAV and ground sensor interventions.

[0017] The above embodiment S2 includes the following steps: S21. Set the linkage threshold. T link If the anomaly confidence index of the suspected anomaly region obtained in S1 C roi >Linkage threshold T link If so, the path planning program will be activated; In specific implementation, the above embodiment S21 specifically involves: setting a linkage threshold. T link =2.0; System judgment: C roiA =3.2>2.0, triggering linkage; C roiB =1.5 < 2.0, drones will not be triggered for the time being, and monitoring will continue only at the satellite level; the setting of 2.0 is based on a balance between statistical significance and transportation costs. C roi The mean of the standardized deviation is 2.0. When the mean exceeds 2.0, it means that the overall ecological quality of the area has deviated from the historical mean by more than two standard deviations, which is a statistically significant anomaly. At this time, the input-output ratio of launching drone inspection is the highest, which avoids overreacting to slight fluctuations and ensures timely response to severe degradation. S22. After activating the path planning program, extract the coordinates of the polygon boundary vertices of the suspected abnormal region ROI; calculate the aspect ratio and area of ​​the suspected abnormal region; if the aspect ratio is less than the ratio threshold and the area is less than the area threshold, use a spiral centripetal scanning path; if the aspect ratio is greater than the ratio threshold or the area is greater than the area threshold, use a parallel reciprocating scanning path. In specific implementation, the above embodiment S22 specifically involves: extracting the ROI. A Calculate the boundary of ROI A Geometric features; ROI in this embodiment A The area is approximately circular, with an area of ​​50 mu (approximately 3.3 hectares) and an aspect ratio of 1.2. The aspect ratio of 1.2 is less than the preset ratio threshold of 2.0, and the area of ​​50 mu is less than the preset area threshold of 100 mu (approximately 6.7 hectares). A spiral centripetal scanning path is selected. This path minimizes the energy consumption of the drone during acceleration and deceleration at turns, making it suitable for concentrated observation of small to medium-sized areas. The ratio threshold of 2.0 is based on the drone's turning efficiency. If the aspect ratio is too large (for narrow and elongated plots), using a spiral path will lead to frequent sharp turns and low efficiency. The area threshold of 100 mu is based on empirical values ​​of single-flight battery life. For small plots less than 100 mu, using a spiral centripetal scanning path can reduce the number of acceleration and decelerations, saving approximately 15% of power. For large or narrow plots, using a parallel reciprocating line is more conducive to segmentation tasks and image stitching. S23. Calculate the required resolution based on the anomaly confidence index of suspected anomaly regions; calculate the optimal flight altitude based on the required resolution. H opt ; In specific implementation, the above embodiment S23 specifically involves: a higher anomaly confidence level means a more serious problem and a higher required resolution (i.e., a lower flight altitude); setting the baseline parameter: GSD base =10cm (corresponding to) C roi Resolution requirement when =1), camera focal length f =25mm, pixel size S w =4.8μm; Required ground resolution GSD req The calculation formula is GSD req =GSD base / C roi ; ROI A of C roiA Substituting 3.2 into the formula for calculating the required ground resolution GSDreq, the required ground resolution is obtained. GSD req =10 / 3.2≈3.1cm; Calculate the optimal flight altitude based on the required ground resolution. =161m; S24. Generate UAV inspection mission instructions that include a comprehensive flight path point set, flight altitude parameters, and sensor exposure strategies. In specific implementation, the above embodiment S24 specifically involves: generating an XML format task file containing 30 key waypoints of the spiral route, including waypoint coordinates, flight altitude, gimbal pitch angle and sensor exposure parameters, and sending it to the unmanned aerial vehicle (UAV) automated hangar deployed on the edge of the mining area. The above embodiment S2 constructs a UAV collaborative data collection triggering mechanism based on anomaly confidence; the core innovation lies in the on-demand allocation of transport capacity and accuracy, that is, generating inspection instructions only for areas where satellites detect significant anomalies, and adjusting the instructions according to the severity of the anomaly (C). roi The system reverse-engineers flight altitude and path strategies. This linkage mechanism avoids indiscriminate and blind inspections of the entire area, significantly reduces the ineffective flight time and energy consumption of drones, solves the pain points of limited drone endurance and data redundancy in large-area monitoring, and achieves the optimal balance between monitoring efficiency and accuracy.

[0018] The above embodiment S3 includes the following steps: S31. The drone arrives over the suspected abnormal area ROI and uses a hyperspectral camera to collect spectral data in the wavelength range of 400nm-1000nm, and uses a lidar to acquire point cloud data. In specific implementation, the above embodiment S31 specifically refers to: the drone in H opt To fly at an altitude of 161 meters, the hyperspectral camera collected data in 150 bands, and the lidar scanned the ground at a rate of 200,000 points per second. S32, the airborne edge computing module performs first-order differential processing on the hyperspectral data to extract the red edge position (REP) and obtain the surface texture features; it uses the lidar point cloud to construct the canopy height model (CHM) to calculate the canopy statistical height and canopy porosity and obtain the canopy structure parameters; In specific implementation, the above embodiment S32 specifically involves: the airborne NVIDIA Jetson module processing the data stream in real time; spectral analysis processing: performing first-order differentiation on the 670nm-760nm band to find the position with the fastest reflectance change as the red edge position (REP). In this embodiment, the current red edge position is... REP current =695 nm; the red edge position of normal vegetation on the land in this embodiment. REP ref The height was 720 mm, showing a significant "blue shift" (movement towards shorter wavelengths), indicating heavy metal stress or chlorophyll deficiency. Structural analysis was performed: a digital surface model (DSM) and a digital elevation model (DEM) were generated from the point cloud data; the difference between the two yielded the canopy height model (CHM). Statistical analysis of the ROI was conducted. A For all pixels within the range with a CHM value greater than 0.2 meters (excluding ground noise), calculate their arithmetic mean to obtain the current canopy statistical height. H current = 0.5 meters, the current historical baseline canopy height of the land. H ref The height is 1.2 meters; on CHM, the height value is less than 0.5 meters. Pixels with a span of 0.5m (0.25m) are identified as canopy gaps (or the gap ratio is calculated using multi-echo information); the number of gap pixels is counted and compared with the ROI. A The current canopy porosity is calculated by comparing the total number of pixels. P current =0.4 (this value is approximately 0.2 under historical healthy conditions); S33. Based on canopy structure parameters and surface texture characteristics, calculate the vegetation stress index (VSI) and structural damage degree; If the vegetation stress index is greater than the vegetation stress index threshold or the structural damage degree is greater than the structural damage degree threshold, then it is determined that there is substantial ecological stress on the land surface, and a micro-sensing wake-up signal containing the ROI center coordinates and instruction type is generated. In specific implementation, the above embodiment S33 is as follows: In this embodiment, the vegetation stress index (VSI) threshold is set to 0.3, which is derived from an empirical model based on the red edge position (REP) offset. When VSI > 0.3, the corresponding REP blue shift exceeds 10 nm, indicating that the vegetation chlorophyll content has decreased by more than 20%, which is the early stage of disease visible to the naked eye. The structural damage degree (SDI) threshold is set to 0.3, which is derived from the statistical deviation of the canopy height from the historical baseline. SDI > 0.3 means that the biomass loss exceeds 30%, corresponding to human logging, lodging, or severe dieback. These two thresholds together constitute the critical value for judging substantial ecological damage. Calculate the vegetation stress index VSI=( REP ref- REP current ) / REP ref The structural damage index (SDI) is calculated to be 0.5. (1-( H current / H ref ))+0.5 P current = 0.5 0.42 + 0.2 = 0.41; Set vegetation stress index threshold = 0.3; Current SDI > 0.3, meeting the wake-up condition; This indicates severe growth inhibition in the area, likely involving changes in soil physicochemical properties, requiring ground sensor intervention; Generate signal: Command type = "Heavy metal / disease stress check", target area is ROI A Center coordinates; this signal will be directly used for device activation in step S4. The above-described embodiment S3 introduces airborne edge computing and mesoscopic feature extraction technology, enabling the UAV to have aerial preliminary diagnostic capabilities. By calculating the hyperspectral red edge position and lidar canopy height in real time, the UAV can instantly determine whether the vegetation is under substantial stress during flight. This on-flight calculation mode breaks the time lag drawback of traditional first-collection-then-processing, and can generate microscopic wake-up signals in the first instance, providing a decision basis for the accurate activation of ground sensors, and realizing the technical effect from morphological monitoring to physiological diagnosis.

[0019] The above embodiment S4 includes the following steps: S41. Ground sensor nodes are pre-installed with LoRaWAN communication modules and are in heartbeat mode by default, sending a live heartbeat packet every 24 hours. In specific implementation, the above embodiment S41 is as follows: Several buried multi-parameter soil sensors are deployed in the mining area. They are powered by batteries and are usually in a deep sleep state with a current of only 5μA. S42. When the node receives the micro-sensing wake-up signal from S3 and its own GPS coordinates are within the suspected abnormal region ROI range, the node immediately parses the instruction type in the signal. In specific implementation, the above embodiment S42 specifically involves: the LoRa gateway carried by the drone flying over the ROI A The wake-up frame is broadcast downwards; located in the ROI. A The internal sensor, designated Node 2025, received the signal, compared its GPS coordinates to confirm that it was within the target range, and was then activated. S43. Adjust the data collection method within the suspected abnormal ROI area where there are no fixed sensors according to the instruction type; if there are no fixed sensor nodes within the suspected abnormal ROI area, the system dispatches a nearby mobile sampling robot to the center coordinates of the ROI to perform soil sampling and portable XRF spectral analysis. In specific implementation, S43 of the above embodiment is as follows: Node 2025 parses the instruction generated in S34 as "heavy metal / disease stress"; turns on the high-power ion-sensitive field-effect transistor (ISFET) sensor and electrochemical heavy metal probe (normally turned off to save power); collects soil pH value of 4.5 (strongly acidic), and cadmium ion (Cd2+) activity is significantly increased; immediately sends the data packet to the base station through the LoRaWAN network; if the instruction type is "drought stress", the node activates the soil moisture sensor and adjusts the sampling frequency to once every 10 minutes; The above embodiment S4 constructs an event-driven IoT wake-up mechanism; by using microscopic sensing signals broadcast by drones, dormant sensors in a specific area are activated or mobile robots are scheduled, enabling on-demand operation of ground monitoring equipment; this mechanism solves the problems of short battery life and high maintenance costs caused by high-frequency acquisition when traditional IoT sensors are deployed in large-scale outdoor areas, extending sensor life by 3-5 times, while ensuring that high-frequency, high-precision physicochemical data can be obtained when ecological anomalies occur, achieving dual optimization of energy efficiency and timeliness.

[0020] The above embodiment S5 includes the following steps: S51. Using the Kriging interpolation method, the macroscopic ecological characteristics of S1, the canopy structure and texture characteristics of S3, and the microscopic measured data of S4 are resampled to a unified standard evaluation grid to obtain a multidimensional ecological data cube. In specific implementation, the above embodiment S51 is as follows: Since the satellite data resolution of S1 is 30m, the UAV data resolution of S3 is 0.1m, and the ground data of S4 is discrete point data, the spatial scale is not uniform, and algebraic operations cannot be directly performed; the system establishes a coverage ROI. A A 1m×1m standard evaluation grid was established for the region. Using Kriging interpolation, the point-like pH and heavy metal data collected by the Node and robot in step S4 were expanded into a continuous areal data layer and mapped onto the standard grid to generate a microscopic measured layer. The UAV data from step S3 was resampled and aggregated, and mapped to generate a mesoscopic structure layer. The satellite data from step S1 was downscaled and projected, and mapped to generate a macroscopic remote sensing layer. Finally, a multidimensional ecological data cube was formed, where each pixel (x,y) in the grid simultaneously contains the normalized value of macroscopic ecological characteristics Isat(x,y), the normalized value of canopy structure and texture characteristics Iuav(x,y), and the normalized value of microscopic measured data. This standard evaluation grid provides a unified data foundation for subsequent calculations in S52, verification in S53, and attribution. S52. Construct an initial weighted ecological quality evaluation equation; read the sub-indicator data in the multi-dimensional ecological data cube and calculate the initial ecological quality index through the weighted ecological quality evaluation equation; the initial weighted ecological quality evaluation equation includes macro-weight coefficients, canopy structure and texture weight coefficients, and micro-weight coefficients; In specific implementation, the above embodiment S52 is as follows: Initial empirical weights are set: macroscopic weight coefficient ws = 0.5 (wide coverage), canopy structure and texture weight coefficient wu = 0.3, and microscopic weight coefficient wg = 0.2. The macroscopic weight coefficient ws = 0.5 is taken into account that satellite data has full coverage and temporal continuity, forming the basic framework for evaluation; the canopy structure and texture weight coefficient wu = 0.3 is because UAV data can reflect the three-dimensional physical state of vegetation and is an important supplement to the mesoscopic level; the microscopic weight coefficient wg = 0.2 is set relatively low because ground points are sparse and their representativeness is relatively weak. This initial setting conforms to the conventional monitoring logic from surface to point, but it needs to be adjusted dynamically when data conflicts are found; therefore, the subsequent S53 is introduced. Ecological Quality Index (EQI) = ws Isat+wu Iuav+wg Ignd; Satellite RSEI indicates that the area is only under "mild stress" (because satellites are not sensitive to heavy metals in soil). Reading the macroscopic remote sensing layer, the normalized value Isat=0.7 (high RSEI, indicating mild stress); reading the mesoscopic structural layer, the normalized value Iuav=0.5 (indicating moderate stress); reading the microscopic measured layer, the normalized value Ignd=0.2 (indicating severe pollution); substituting into the initial weighted ecological quality assessment equation, the calculation is: EQI=0.5×0.7+0.3×0.5+0.2×0.2=0.54; since it is mainly dominated by high-weighted satellite data, it may mask the true pollution on the ground. Therefore, the system must extract the two key sub-indicators Isat and Iuav and transmit them to S53 for logical conflict detection. S53. Based on the sub-indicator data extracted in S52, calculate the inter-scale consistency deviation; if the inter-scale consistency deviation is less than or equal to the tolerance limit threshold, then the initial ecological quality index is used as the final ecological quality index. Otherwise, the weight coefficients in the initial weighted ecological quality evaluation equation are reduced according to the exponential decay law to obtain the adjusted weighted ecological quality evaluation equation; the ecological quality index is then calculated again using the adjusted weighted ecological quality evaluation equation to obtain the final ecological quality index. In specific implementation, the above embodiment S53 is as follows: calculate the inter-scale consistency deviation Delta=|Isat-Iuav|=|0.7-0.2|=0.5; the tolerance threshold set in this embodiment is 0.2; the tolerance threshold of 0.2 is set based on the error distribution characteristics of normalized data; within the normalized space of 0-1, the observation error of different sensors for the same object is usually within 0.1; setting 0.2 as the threshold (i.e., a 20% deviation) means that the difference between the two has exceeded the normal systematic error range, and is very likely caused by inversion failure or a sudden change in the properties of the monitored object (such as "different objects of the same spectrum"), so weight intervention must be performed; 0.5 > 0.2; The system determines that "there is a significant conflict between macroscopic data and microscopic true values," indicating a local failure in satellite inversion. Based on Delta = 0.5 calculated by S53, the system penalizes and compensates the initial weights according to an exponential law: the corrected microscopic weight coefficient wg′ = wg × e Delta =0.2×e 0.5 ≈0.33; Corrected macroscopic weight coefficient: ws′=ws×e−Delta=0.5×e−0.5≈0.30; After normalizing the weights, the final weight allocation is adjusted as follows: microscopic weight coefficient wg′′≈0.6, macroscopic weight coefficient ws′′≈0.15, canopy structure and texture weight coefficient wu′′≈0.25; The S51 mesh data is read again using the corrected weights for calculation: EQIfinal =0.15×0.7+0.25×0.5+0.6×0.2=0.35; Analyze the final weight contribution of each item in the S54 calculation; Identify that the corrected ground data has the highest weight (0.6) and its corresponding value is the lowest (0.2, representing the worst quality); The system logic determines that the ecological weakness of this area lies in the "ground physical and chemical properties" rather than the "macro vegetation cover"; Automatically generate diagnostic labels: "Main causes of degradation: soil acidification and heavy metal pollution"; In the above embodiment S53, the weight coefficients in the initial weighted ecological quality evaluation equation are reduced and adjusted according to the exponential decay law to obtain the adjusted weighted ecological quality evaluation equation. The process also includes a parameter inverse calibration step. S531. If the micro-weight coefficient in the adjusted weighted ecological quality evaluation equation is greater than the micro-weight coefficient threshold, then the macro-remote sensing inversion parameters are reverse-calibrated using micro-measured data, and the updated parameters will be used in step S1 of the next cycle. In specific implementation, the above embodiment is as follows: Construct a regression loss function Loss=(F(Refsat,P)−Valgnd) 2 Valgnd is taken from the ground-measured pH value (4.5) in the S51 grid, and Refsat is taken from the satellite band reflectance in the S51 grid; when the micro-weight coefficient > the micro-weight coefficient threshold, it is determined that the pH inversion algorithm has a huge error in this area; the pH=4.5 measured by the Node is used as the Label, and the Sentinel-2 band reflectance of the corresponding coordinate is used as the Feature; the regression loss function Loss is specifically the mean square error function; the inversion function F is initialized as a linear regression model trained based on the historical data of the entire domain, taking the satellite multispectral band reflectance vector as input and the predicted pH value as output, and P is the regression coefficient vector of the linear regression model; in During calibration, the Adam optimizer was used with an initial learning rate of 0.001, coupled with an exponential decay strategy (decreasing to 0.95 times the original value every 10 iterations). Each calibration used a small batch of data, consisting of satellite reflectance samples and ground-measured pH values ​​from the previous cycle, matched to spatial locations. The coefficients P were iteratively updated by minimizing the mean square error function until the loss converged or the preset number of iterations was reached. This calibrated the macroscopic inversion model and established a clear mapping relationship between spectral and physicochemical properties. The updated pH inversion coefficients P were saved. In the next satellite scan of the region, even without ground sensor data, the satellite could more accurately estimate the lower pH value based on the corrected parameters, ensuring the system's accuracy. S54. Based on the final ecological quality index and the data source type of the factor with the highest weight, generate labels for the main causes of degradation to obtain the land ecological quality evaluation results. In the above embodiment S54, generating labels for the main causes of degradation based on the data source type of the factor with the highest weight, and obtaining the land ecological quality assessment results includes the following steps: S541. Construct an attribution feature vector that includes water stress feature components, pollution stress feature components, and physical damage feature components; S542. If the soil moisture content is lower than the wilting coefficient, the water stress characteristic component is set to 1; if the heavy metal ion concentration exceeds the standard or the pH value is abnormal, the pollution stress characteristic component is set to 1; if the surface roughness increases suddenly and the vegetation height variance exceeds the vegetation height variance threshold, it is set to 1. S543. Use a decision tree classifier to classify the attribution feature vectors and output diagnostic conclusions for drought, water shortage, chemical pollution, physical damage, or combined degradation. In specific implementation, the above embodiment is as follows: The system revisits the multidimensional ecological data cube constructed by S51, extracts features to construct the attribution feature vector Vcause=[Vwater,Vpollution,Vphysical]; Vwater, Vpollution, and Vphysical are the attribution feature vectors of water stress feature components, pollution stress feature components, and physical damage feature components, respectively; water data is read from the micro-measurement layer, which is higher than the wilting coefficient, so Vwater=0; heavy metal data is read from the micro-measurement layer, where cadmium ions exceed the standard, so Vpollution=1; structural damage data (i.e., the SDI value calculated by S3) is read from the meso-structure layer, showing vegetation dwarfing, so Vphysical=1; the vector [0,1,1] is input into the pre-trained decision tree classifier; the output conclusion is: "Secondary vegetation structural damage caused by chemical pollution"; finally, the EQI is... final The index and detailed diagnostic conclusions are integrated to generate a report that includes "severe pollution warning" and "pollution source location (ROI)". A The system generates intelligent reports with the following information: "Drought", "Pollution", and "Remediation Plan (Sprinkle Lime to Neutralize Acidity)" and sends them to land managers. A training set containing historical case studies is constructed, with sample features of [Vwater, Vpollution, Vphysical] and labels indicating the degradation type as determined by experts (e.g., "Drought", "Pollution", "Physical Damage"). The system is trained using the CART (Classification and Regression Tree) algorithm, employing the Gini coefficient as the splitting criterion. Cross-validation is used to optimize the maximum tree depth and the minimum number of samples per leaf node to prevent overfitting. The trained decision tree classifier can quickly output classification results based on the input feature vector, achieving a classification accuracy of over 92% on the test set. The above embodiment S5 establishes a dynamic fusion and adaptive evaluation system for multi-source heterogeneous data; it solves the problem of spatiotemporal registration and logical conflicts between multi-scale data from space, air, and ground; by calculating the consistency deviation between scales, it dynamically adjusts the weights of each data source, prioritizes the trust of high-precision microscopic measured data, and effectively corrects the inversion error of satellite remote sensing in complex scenarios; at the same time, it uses microscopic data to inversely calibrate macroscopic inversion parameters, realizing the evolution and accuracy improvement of the evaluation model, and the final output attribution diagnosis report provides scientific and accurate decision support for ecological restoration; A land ecological quality assessment system based on linked data acquisition, implementing the aforementioned land ecological quality assessment method based on linked data acquisition, the system comprising: Macro monitoring module: used to receive and process satellite remote sensing images, identify suspected abnormal areas and calculate anomaly confidence; Joint Control Center: Includes an intelligent scheduling engine, used to receive signals from the macro monitoring module and generate UAV mission instructions and ground wake-up signals; Airborne data acquisition subsystem: It consists of a UAV carrying hyperspectral and lidar payloads and an airborne edge computing unit, and has the ability to fly autonomously and perform real-time data screening. Ground-based sensing subsystem: includes a distributed low-power IoT sensor node cluster and a mobile sampling robot, with a wake-up mechanism and multi-parameter acquisition capabilities; Comprehensive evaluation cloud platform: used to store multi-source heterogeneous data, run dynamic weight correction models and parameter inverse calibration algorithms, and generate ecological quality evaluation reports.

[0021] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0022] The preferred embodiments of the invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.

Claims

1. A method for evaluating land ecological quality based on linked data collection, characterized in that, Includes the following steps: S1. Obtain satellite remote sensing images of the area to be evaluated and construct macroscopic ecological characteristics; Based on macro-ecological characteristics, suspected abnormal areas are identified, and an anomaly confidence index for the suspected abnormal areas is calculated. S2. When the anomaly confidence index exceeds the linkage threshold, the scanning path is dynamically selected based on the spatial set characteristics of the suspected anomaly area, and a drone inspection task instruction is generated. S3. The drone scans the suspected abnormal area according to the drone inspection task instructions in S2. The collected data is processed in real time, canopy structure parameters and surface texture features are extracted, and vegetation stress index and structural damage degree are calculated to obtain the calculation results; if the calculation results meet the micro-awakening conditions, a micro-sensing awakening signal is generated. S4 and broadcast S3 generate micro-sensing wake-up signals to directionally activate ground equipment located in suspected abnormal areas, collect soil physicochemical property parameters and micro-environment meteorological parameters, and obtain micro-measured data; S5. Spatiotemporal registration and normalization are performed on the macroscopic ecological characteristics of S1, the canopy structure and texture characteristics of S3, and the microscopic measured data of S4. Calculate the deviation measure between data at different scales; The weight coefficients of each data source in the ecological quality assessment model are dynamically adjusted based on the deviation metric, and the land ecological quality assessment results are output. The ecological quality assessment model includes a weighted ecological quality assessment equation and a weight adjustment rule. The weight adjustment rule includes: if the inter-scale consistency deviation is greater than the tolerance limit threshold, the macro-weight coefficient is reduced according to the exponential decay law, and the micro-weight coefficient is increased accordingly.

2. The land ecological quality assessment method based on linked data acquisition as described in claim 1, characterized in that, S1 includes the following steps: S11. Acquire multispectral satellite images of the area to be evaluated for the most recent N periods, perform radiometric calibration and atmospheric correction to obtain corrected image data; based on the corrected image data, calculate the normalized vegetation index, surface temperature and remote sensing ecological index to obtain macro-ecological characteristics. S12. Use the STL algorithm to decompose the time series of macro-ecological characteristics into seasonal, trend, and residual terms; for the trend term, calculate the deviation between the current period value and the historical average for the same period. S13. If the absolute value of the pixel deviation is greater than the threshold multiple of the baseline standard deviation. k If the pixel is identified as an anomaly, then a density clustering algorithm is used to spatially cluster the anomaly points to generate continuous suspected anomaly regions. S14. Calculate the mean deviation of all pixels within the ROI, and obtain the anomaly confidence index after normalization.

3. The land ecological quality assessment method based on linked data acquisition as described in claim 2, characterized in that, S2 includes the following steps: S21. Set the linkage threshold; if the anomaly confidence index of the suspected anomaly area obtained in S1 is greater than the linkage threshold, then activate the path planning program. S22. After activating the path planning program, extract the coordinates of the polygon boundary vertices of the suspected abnormal region ROI; calculate the aspect ratio and area of ​​the suspected abnormal region; if the aspect ratio is less than the ratio threshold and the area is less than the area threshold, use a spiral centripetal scanning path; if the aspect ratio is greater than the ratio threshold or the area is greater than the area threshold, use a parallel reciprocating scanning path. S23. Calculate the required resolution based on the anomaly confidence index of suspected anomaly areas; calculate the optimal flight altitude based on the required resolution. S24. Generate UAV inspection mission instructions that include a comprehensive flight path point set, flight altitude parameters, and sensor exposure strategies.

4. The land ecological quality assessment method based on linked data acquisition as described in claim 3, characterized in that, S3 includes the following steps: S31. The drone arrives over the suspected abnormal area ROI and uses a hyperspectral camera to collect spectral data within the wavelength range and a lidar to acquire point cloud data. S32. Perform first-order differential processing on the hyperspectral data to extract the red edge position and obtain the surface texture features; use lidar point cloud to construct a canopy height model, calculate the canopy statistical height and canopy porosity, and obtain the canopy structure parameters; S33. Based on canopy structure parameters and surface texture characteristics, calculate the vegetation stress index (VSI) and structural damage degree; If the vegetation stress index is greater than the vegetation stress index threshold or the structural damage degree is greater than the structural damage degree threshold, then it is determined that there is substantial ecological stress on the land surface, and a micro-sensing awakening signal containing the ROI center coordinates and instruction type is generated.

5. The land ecological quality assessment method based on linked data acquisition as described in claim 4, characterized in that, S4 includes the following steps: S41. Ground sensor nodes are pre-installed with LoRaWAN communication modules and are in heartbeat mode by default, sending a live heartbeat packet every 24 hours. S42. When the node receives the micro-sensing wake-up signal from S3 and its own GPS coordinates are within the suspected abnormal region ROI range, the node immediately parses the instruction type in the signal. In specific implementation, preferably, S42 specifically refers to: the LoRa gateway carried by the drone flying over the ROI A The wake-up frame is broadcast downwards; located in the ROI. A The internal sensor, designated Node 2025, received the signal, compared its GPS coordinates to confirm that it was within the target range, and was then activated. S43. Adjust the data collection method for areas without fixed sensors within the suspected abnormal ROI region according to the instruction type; if there are no fixed sensor nodes within the suspected abnormal ROI region, the system dispatches a nearby mobile sampling robot to the center coordinates of the ROI to perform soil sampling and portable XRF spectral analysis.

6. The land ecological quality assessment method based on linked data acquisition as described in claim 5, characterized in that, S5 includes the following steps: S51. Using the Kriging interpolation method, the macroscopic ecological characteristics of S1, the canopy structure and texture characteristics of S3, and the microscopic measured data of S4 are resampled to a unified standard evaluation grid to obtain a multidimensional ecological data cube. S52. Construct an initial weighted ecological quality evaluation equation; read the sub-indicator data in the multi-dimensional ecological data cube and calculate the initial ecological quality index through the weighted ecological quality evaluation equation; the initial weighted ecological quality evaluation equation includes macro-weight coefficients, canopy structure and texture weight coefficients, and micro-weight coefficients; S53. Based on the sub-indicator data extracted in S52, calculate the inter-scale consistency deviation; if the inter-scale consistency deviation is less than or equal to the tolerance limit threshold, then the initial ecological quality index is used as the final ecological quality index. Otherwise, the weight coefficients in the initial weighted ecological quality evaluation equation are reduced according to the exponential decay law to obtain the adjusted weighted ecological quality evaluation equation; the ecological quality index is then calculated again using the adjusted weighted ecological quality evaluation equation to obtain the final ecological quality index. S54. Based on the final ecological quality index and the data source type of the factor with the highest weight, generate labels for the main causes of degradation to obtain the land ecological quality evaluation results.

7. The land ecological quality assessment method based on linked data acquisition as described in claim 6, characterized in that, The step S53, which involves reducing and adjusting the weight coefficients in the initial weighted ecological quality evaluation equation according to the exponential decay law to obtain the adjusted weighted ecological quality evaluation equation, also includes a parameter inverse calibration step. S531. If the micro-weight coefficient in the adjusted weighted ecological quality evaluation equation is greater than the micro-weight coefficient threshold, then the macro-remote sensing inversion parameters are reverse-calibrated using micro-measured data, and the updated parameters will be used in step S1 of the next cycle.

8. The land ecological quality assessment method based on linked data acquisition as described in claim 6, characterized in that, The steps in S54 to generate labels for the main causes of degradation based on the data source type of the factor with the highest weight, and to obtain the land ecological quality assessment results, include the following: S541. Construct an attribution feature vector that includes water stress feature components, pollution stress feature components, and physical damage feature components; S542. If the soil moisture content is lower than the wilting coefficient, the water stress characteristic component is set to 1; if the heavy metal ion concentration exceeds the standard or the pH value is abnormal, the pollution stress characteristic component is set to 1; if the surface roughness increases suddenly and the vegetation height variance exceeds the vegetation height variance threshold, it is set to 1. S543. Use a decision tree classifier to classify the attribution feature vectors and output diagnostic conclusions for drought, water shortage, chemical pollution, physical damage, or combined degradation.

9. The land ecological quality assessment method based on linked data acquisition as described in claim 1, characterized in that, A land ecological quality assessment system based on linked data acquisition, implementing any one of the methods in claims 1-8, the system comprising: Macro monitoring module: used to receive and process satellite remote sensing images, identify suspected abnormal areas and calculate anomaly confidence; Joint Control Center: Includes an intelligent scheduling engine, used to receive signals from the macro monitoring module and generate UAV mission instructions and ground wake-up signals; Airborne data acquisition subsystem: It consists of a UAV carrying hyperspectral and lidar payloads and an airborne edge computing unit, and has the ability to fly autonomously and perform real-time data screening. Ground-based sensing subsystem: includes a distributed low-power IoT sensor node cluster and a mobile sampling robot, with a wake-up mechanism and multi-parameter acquisition capabilities; Comprehensive evaluation cloud platform: used to store multi-source heterogeneous data, run dynamic weight correction models and parameter inverse calibration algorithms, and generate ecological quality evaluation reports.

10. The land ecological quality assessment method based on linked data acquisition as described in claim 9, characterized in that, The linkage control center also includes: Communication gateway: Supports heterogeneous network convergence access of satellite communication, 4G / 5G network and LoRaWAN narrowband IoT protocol; Digital Twin Engine: Utilizing data collected from S1 to S4, an ecological model of the area to be evaluated is constructed in real time in virtual space, which is used to visualize the ecological degradation process and simulate the effects of restoration plans.

Images