Ecological environment monitoring device based on remote sensing image

Abstract

The invention relates to the technical field of ecological environment monitoring, in particular to an ecological environment monitoring device based on remote sensing images, which comprises a data acquisition and preprocessing module used for acquiring current and historical remote sensing images and meteorological data of a target area and constructing a time sequence data set; the land coverage identification module is used for identifying the land coverage type of the target area; the reference area selection module is used for delimiting a reference area in a preset range; the meteorological-spectral mapping model building module is used for building a mapping function from meteorological data to spectral reflection characteristics; the spectrum anomaly detection and region correction module is used for comparing the difference between the actual spectrum and the expected spectrum output by the model, and performing boundary cutting correction on an abnormal region to obtain a new target region after meteorological influence correction; and the ecological parameter inversion and output module is used for performing inversion and mapping of the ecological parameters. The adaptive capacity to different weather conditions can be improved, and the accuracy of a monitoring result is enhanced.

Description

Technical Field

[0001] This invention relates to the field of ecological environment monitoring technology, and specifically to an ecological environment monitoring device based on remote sensing images. Background Technology

[0002] With the increasing prominence of global ecological and environmental problems, dynamic and precise monitoring of large-scale surface ecosystems has become a crucial foundation for environmental management, ecological protection, and sustainable development. Traditional ground-based survey methods, limited by manpower, resources, and spatiotemporal coverage, struggle to meet the demands for large-scale, high-frequency, and continuous monitoring. Remote sensing technology, with its advantages of macroscopic, rapid, objective, and periodic observation, has become an indispensable tool for ecological and environmental monitoring. Currently, monitoring based on remote sensing images primarily relies on multispectral, hyperspectral, and radar image data acquired by satellites or airborne platforms. This data is then interpreted and analyzed to assess key ecological parameters such as vegetation cover, land use, water quality, and surface temperature.

[0003] Remote sensing monitoring relies on the interaction between surface targets and electromagnetic waves (primarily from solar radiation or active emission from sensors). Different land features (such as vegetation, water bodies, soil, and man-made structures) exhibit unique absorption, reflection, and emission characteristics to incident electromagnetic waves due to differences in material composition, structure, and moisture content, forming identifiable "spectral fingerprints." Sensors record these differentiated radiation signals to generate pixel brightness values ​​in remote sensing images, thereby retrieving various ecological parameters.

[0004] However, the acquisition of remote sensing data is highly dependent on atmospheric conditions and the instantaneous state of the land surface. Climate change and its resulting fluctuations in meteorological conditions introduce significant uncertainties and disturbances at the data source. For example, surface moisture after precipitation significantly alters the spectral reflectance characteristics of vegetation leaves and soil—moisture reduces the reflectance of vegetation in the near-infrared band, which can easily be misinterpreted as a decline in growth or a reduction in biomass; the reflectance of moist soil is also significantly lower than that of dry soil, affecting the accuracy of land surface classification and parameter inversion.

[0005] Existing monitoring methods largely rely on remote sensing images from specific time periods or use simple vegetation indices (such as NDVI), lacking adaptive modeling and compensation mechanisms for seasonal variations and interannual fluctuations. This leads to fluctuations in monitoring results between peak and off-peak vegetation seasons, and between dry and wet years, making it difficult to effectively separate long-term ecological trends from short-term climate noise. This reduces the comparability and decision support value of data over time series, and also limits the ability to continuously monitor under different meteorological conditions and the accuracy and dimensionality of ecological parameter inversion.

[0006] Given the shortcomings of existing technologies, there is an urgent need for a new, intelligent remote sensing image-based ecological environment monitoring device to improve its adaptability to different climatic conditions and enhance the accuracy and reliability of monitoring results. Summary of the Invention

[0007] To address the aforementioned problems, this invention provides an ecological environment monitoring device based on remote sensing images, which improves the accuracy of monitoring results.

[0008] To achieve the above objectives, the technical solution of the present invention is as follows: an ecological environment monitoring device based on remote sensing images, comprising:

[0009] The data acquisition and preprocessing module is used to acquire the current remote sensing images and current meteorological data of the target area and perform preprocessing, and simultaneously acquire historical remote sensing images and corresponding historical meteorological data to construct a time-series dataset.

[0010] The land cover identification module is used to identify the land cover type of a target area based on spectral reflectance characteristics and a classification algorithm.

[0011] The reference area selection module is used to select fixed reference objects in historical remote sensing images, delineate a reference area with the reference object as the center, and use it as the benchmark area for learning the meteorological-spectral relationship. It is assumed that the land cover type of this area remains unchanged in time.

[0012] The meteorological-spectral mapping model construction module is used to learn the dynamic relationship between meteorological data and spectral reflectance characteristics within a reference area using a deep learning model, and to establish a mapping function from meteorological data to spectral reflectance characteristics.

[0013] The spectral anomaly detection and regional correction module is used to input the current remote sensing image and current meteorological data into the trained meteorological-spectral mapping model, predict the expected spectrum of the reference area under the current meteorological conditions, compare the actual spectrum of the target area in the current remote sensing image with the expected spectrum output by the model, and if the difference exceeds the set threshold, it is determined to be an area where the spectral-meteorological relationship does not correspond. The boundary of the area is then cropped and corrected to obtain a new target area after meteorological impact correction.

[0014] The ecological parameter inversion and output module is used to invert and map ecological parameters based on the corrected new target area.

[0015] Furthermore, current and historical meteorological data include at least precipitation data, temperature data, air humidity data, and cloud cover data.

[0016] Furthermore, ecological parameters include NDVI, leaf area index, surface temperature, and vegetation cover.

[0017] Furthermore, the fixed reference objects are natural or man-made features that are spectrally and spatially stable in long-term remote sensing images.

[0018] Furthermore, it also includes a dynamic baseline construction module, which is used for long-term historical data to establish a dynamic ecological baseline library that changes with the seasons for different geographical units.

[0019] Furthermore, it also includes an intelligent early warning module, which is used to compare real-time ecological parameters with the dynamic baseline of the corresponding spatiotemporal location, and combine the parameter change trend analysis to conduct graded assessment and early warning of ecological anomalies.

[0020] Furthermore, it also includes a model adaptive update module, which is used to continuously collect remote sensing images and synchronous meteorological data during long-term monitoring, and to trigger incremental learning or retraining of the meteorological-spectral mapping model periodically or when the prediction error of the meteorological-spectral mapping model is detected to continuously exceed the preset range.

[0021] Furthermore, the model adaptive update module determines whether the prediction error of the meteorological-spectral mapping model continuously exceeds a preset range based on the following logic:

[0022] Within the reference area, the expected spectrum predicted by the model based on current meteorological data is compared with the actual spectrum extracted from the current remote sensing image.

[0023] Calculate the root mean square error between the two;

[0024] The root mean square error obtained from each processing step, along with its timestamp, is stored separately to form a continuous error time series.

[0025] If a statistical test is performed on the error time series and an upward trend in the error is detected, and the current error value is already at the historical preset level, then it is determined that the performance is continuously degrading.

[0026] Furthermore, the intelligent early warning module performs state comparison based on the following logic:

[0027] Obtain the parameter value P obtained from the inversion of the current spatiotemporal location;

[0028] The normal value range extracted from the dynamic ecological baseline database, corresponding to the current pixel geographic location and the current time period;

[0029] Determine whether the parameter value P falls within the normal range:

[0030] If P ∈ the normal value range, it is marked as normal and will not proceed to the subsequent warning process;

[0031] If P If the value falls within the normal range, it is marked as an abnormal state, and the deviation is calculated.

[0032] Furthermore, the intelligent early warning module performs trend analysis based on the following logic:

[0033] Input a short time series parameter sequence: obtain the time series parameter values ​​of the anomalous pixels for the most recent N days, with the current parameter value P as the end point of the sequence;

[0034] Trend fitting: Perform linear regression on short-time series data and calculate the trend slope S;

[0035] Decision matrix construction: Construct a decision matrix with deviation as the horizontal axis and trend slope S as the vertical axis, and predefine the warning level and urgency of each cell in the matrix.

[0036] Working principle:

[0037] The device first simultaneously acquires remote sensing images and corresponding meteorological data (current and historical) of the target area, forming a spatiotemporally aligned data foundation. Land cover identification is then used to understand the land surface background. A key step is selecting a fixed reference area whose spectrum and morphology are stable over a long time scale. This area is assumed to have an unchanging land cover type, and its spectral variations over time are primarily attributed to fluctuations in meteorological conditions, thus becoming an ideal natural laboratory for learning pure meteorological-spectral correlations.

[0038] Using a deep learning model, the complex and nonlinear dynamic relationship between historical meteorological data (such as precipitation, temperature, and humidity) and multi-band spectral reflectance is analyzed within a selected fixed reference area. Through learning, the model establishes a mapping function (meteorological-spectral mapping model) that can predict the spectral reflectance characteristics that the reference area should exhibit based on input meteorological conditions. Essentially, this model quantifies the interference patterns of meteorological factors on the spectral response of specific land surface types.

[0039] During monitoring, current meteorological data is input into a pre-trained model to calculate the expected spectrum of the reference area under the current weather conditions. Then, the model-predicted spectrum is compared pixel-by-pixel with the actual spectrum of the entire target area in the current remote sensing image. Theoretically, if a pixel within the target area is only affected by meteorology (and the land cover remains unchanged), its actual spectrum should match the spectral pattern predicted by the model. If the actual spectrum of a certain area differs significantly from the expected spectrum (exceeding a threshold), it indicates that the spectral changes in that area cannot be fully explained by the current meteorological conditions, suggesting possible real changes in the land surface caused by non-meteorological factors (such as vegetation degradation, land use change, etc.) or that the model is not applicable in that local area. The device identifies this area as a region with a mismatch between spectral and meteorological relationships and temporarily excludes or marks it from the accurate area used for ecological parameter inversion through boundary clipping, thus obtaining a corrected image of the target area where meteorological influences are identified and isolated.

[0040] Based on the corrected imagery, ecological parameters such as NDVI and leaf area index are retrieved. These parameters, having minimized instantaneous meteorological interference, better reflect the true ecological condition of the land surface. Combined with a dynamic baseline database, the real-time retrieved parameters are compared with historical normal ranges under corresponding spatiotemporal conditions. By analyzing short-term trends, intelligent ecological monitoring is achieved, enabling detection of anomalies and early warning of changing trends.

[0041] Through the model adaptive update module, the device can utilize continuously accumulated new data to periodically test and optimize the meteorological-spectral mapping model, ensuring that it can adapt to potentially long-term slow changes (such as the new normal under the background of climate change) or capture new disturbance patterns, thus maintaining the long-term accuracy and robustness of the device.

[0042] The above approach has the following beneficial effects:

[0043] 1. This scheme, by constructing and applying a meteorological-spectral mapping model, enables the system to intelligently separate instantaneous variation components caused by meteorological fluctuations such as precipitation and humidity from spectral signals, and to identify and correct pixels or regions affected by these fluctuations. This solves the misjudgment problem caused by the inability of traditional methods to distinguish between meteorological pseudo-changes and real ecological changes, making the retrieved ecological parameters such as vegetation index, biomass, and surface humidity more realistically reflect the inherent attributes and long-term state of surface targets, and significantly improving the accuracy of basic data.

[0044] 2. This method does not rely on absolutely clear imagery. Instead, by learning the relationship between meteorology and spectral data, it possesses the ability to conduct continuous and effective monitoring under various meteorological conditions (such as after rain or before and after cloudy weather). This overcomes the problems of monitoring interruptions or large fluctuations in results caused by data quality or meteorological interference in traditional methods.

[0045] 3. This solution, combining a dynamic baseline database and an intelligent early warning module, enables the device not only to determine whether the current ecological state is abnormal, but also to assess the severity and urgency of the anomaly through trend analysis. This transforms monitoring from post-event description to pre-event early warning and in-event tracking, providing more timely and refined information support for ecological protection and restoration decisions.

[0046] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0047] Figure 1 This is a system block diagram of the ecological environment monitoring device based on remote sensing images according to the present invention. Detailed Implementation

[0048] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0049] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0050] The following detailed description illustrates the specific implementation method:

[0051] Example:

[0052] As attached Figure 1 As shown: An ecological environment monitoring device based on remote sensing images includes a data acquisition and preprocessing module, which is used to acquire current remote sensing images of the target area from the data center of multi-source Earth observation satellites, acquire synchronous meteorological data from the meteorological station observation network, and perform preprocessing, and synchronously acquire historical remote sensing images and corresponding historical meteorological data to construct a time-series dataset; the current meteorological data and historical meteorological data include at least precipitation data, temperature data, air humidity data and cloud cover data.

[0053] Specifically, the preprocessing steps include:

[0054] Remote sensing image preprocessing: Radiometric calibration and atmospheric correction (such as using FLAASH or 6S models) are performed on the acquired raw images to obtain the true surface reflectance, and geometric fine correction is performed to ensure accurate spatial registration of all temporal images.

[0055] Meteorological data preprocessing: Spatiotemporal interpolation (such as kriging interpolation) is performed on the meteorological data to match its resolution with that of the remote sensing imagery and to unify it to the same spatiotemporal coordinate system. Finally, the preprocessed, spatiotemporally aligned current and historical remote sensing images and meteorological data are organized in time series to construct the time-series dataset required by the instrument.

[0056] The land cover identification module is used to identify the land cover type of a target area based on spectral reflectance characteristics and employing classification algorithms. Specifically, this module classifies the land cover of the target area based on the spectral reflectance characteristics of each pixel in the image at different wavelengths (such as visible light, near-infrared, and shortwave infrared), using supervised classification algorithms (such as random forests and support vector machines) or deep learning semantic segmentation models (such as U-Net). The classification system includes at least the following basic types: cultivated land, forest land, grassland, water bodies, construction land, and unused land. The output of this module is a land cover classification map, providing baseline surface information for subsequent analysis.

[0057] The reference area selection module is used to select fixed reference objects in historical remote sensing images, delineate a reference area with the reference object as the center, and assume that the land cover type of the area remains unchanged in time, serving as the benchmark area for learning the meteorological-spectral relationship.

[0058] Specifically, this module automatically identifies spectrally and spatially stable natural or man-made features as fixed references in long-term remote sensing imagery (such as all available imagery from the past 1-2 years). Examples include long-established utility poles in woodlands and exposed, stable rock formations on grasslands. Centered on the selected reference feature, a pre-defined reference area is delineated within the corresponding land cover type (e.g., woodland) on the map. The device assumes that the land cover type of this reference area remains constant throughout the time series; therefore, its spectral variations over time are primarily attributed to fluctuations in meteorological conditions, thus establishing it as a benchmark area for learning and verifying meteorological-spectral correlations.

[0059] This embodiment uses a vegetation scene as an example.

[0060] Within the target vegetation area, identify and select long-term stable man-made features as the core reference, such as: the towers of high-voltage transmission lines crossing forests or grasslands, hardened forest fire lookout towers, or fixed meteorological observation stations themselves. These targets have stable and unique spectral and geometric characteristics (high reflectivity, regular shape) in long-term imagery, making them easy to automatically identify. If no man-made features are available, select the edges of large areas of exposed stable rock or permanent water bodies within the vegetation area. Alternatively, if neither of the above conditions is met, use the center of the target vegetation area as the geometric reference.

[0061] Reference area delineation: Centered on the selected tower (assuming it is located within a "deciduous broad-leaved forest" type), within the homogeneous vegetation patch containing the tower, a region much larger than the tower itself but much smaller than the entire vegetation patch (e.g., 30m x 30m) is delineated. A key premise is that the vegetation type (deciduous broad-leaved forest) and its growth state (mature forest) within this small area remain stable throughout the historical sequence. The primary driver of spectral variation in this area is meteorological fluctuation; therefore, it is established as the baseline region for learning the meteorological-spectral relationship.

[0062] The meteorological-spectral mapping model construction module is used to learn the dynamic relationship between meteorological data and spectral reflectance characteristics within a reference area using a deep learning model, and to establish a mapping function from meteorological conditions to changes in spectral reflectance.

[0063] Specifically, this module is the core modeling unit. It extracts multi-band spectral reflectance data of the reference area across all historical time phases (as model output labels) and corresponding synchronous meteorological data (as model input features). Using a deep learning model (e.g., a deep neural network with multiple fully connected layers and Dropout layers, or a time series model such as LSTM), it learns the complex nonlinear mapping relationship between meteorological conditions and spectral reflectance. The training objective is to enable the model to accurately predict the spectral reflectance characteristics that the reference area should exhibit under current conditions (with unchanged land cover type), based on a set of input meteorological conditions (such as cumulative precipitation and daily average temperature over the past week). After training, a dynamic mapping function from meteorological conditions to changes in spectral reflectance (meteorological-spectral mapping model) is established. This model essentially quantifies the interference patterns of meteorological factors on the spectral response of a specific land surface type.

[0064] The spectral anomaly detection and regional correction module is used to input the current remote sensing image and current meteorological data into a trained meteorological-spectral mapping model, predict the expected spectrum of the reference area under the current meteorological conditions, compare the actual spectrum of the target area in the current image with the expected spectrum output by the model, and if the difference exceeds a set threshold, it is determined to be an area where the spectral-meteorological relationship does not correspond. The boundary of the area is then cropped and corrected to obtain a new target area after meteorological impact correction.

[0065] Specifically, the current remote sensing imagery (after preprocessing) and current meteorological data are input into a pre-trained meteorological-spectral mapping model. The model first predicts the expected spectrum (a set of expected reflectance values) of the reference area under the current weather conditions based on the current meteorological data.

[0066] Subsequently, the module performs a pixel-by-pixel comparison between the model-predicted expected spectrum (as a baseline) and the actual spectrum of each pixel in the entire target region of the current image. It calculates the difference between the actual and expected reflectance of each pixel in each band. If the spectral difference of a pixel or pixel cluster exceeds a set threshold (e.g., a spectral angle greater than 0.1 radians), the region is determined to be a region with a spectral-meteorological mismatch. For these anomalous regions, the module performs boundary clipping correction on the vector boundary of the target region, temporarily excluding them from the reliable region in this round of analysis. Finally, a more accurate new target region is obtained after meteorological influence identification and filtering.

[0067] Theoretically, if a pixel within the target area is only affected by meteorology (and the land cover remains unchanged), its actual spectrum should be consistent with the spectral pattern predicted by the model. If the actual spectrum of a certain area differs significantly from the expected spectrum (exceeding a threshold), it indicates that the spectral changes in that area cannot be fully explained by the current meteorological conditions, suggesting possible real changes in the land surface caused by non-meteorological factors (such as vegetation degradation, land use change, etc.) or that the model is not applicable in that local area. The device identifies this area as a region with a mismatch between spectral and meteorological relationships and temporarily excludes or marks it from the area used for ecological parameter inversion through boundary clipping, thereby obtaining a corrected image of the target area where meteorological influences are identified and isolated.

[0068] The ecological parameter inversion and output module is used to invert and map ecological parameters based on the corrected new target area.

[0069] Specifically, this module, based on the corrected new target area imagery obtained in the previous step, employs mature physical models or empirical / semi-empirical algorithms to perform inversion calculations of various ecological parameters. Based on the corrected imagery, ecological parameters such as NDVI and leaf area index are inverted. These parameters, having minimized instantaneous meteorological interference, better reflect the true ecological condition of the land surface. The inverted parameters include, but are not limited to: Normalized Difference Vegetation Index (NDVI), Leaf Area Index (LAI), Land Surface Temperature (LST), and FVC. After inversion, the module generates standardized thematic maps of ecological parameters, which can be visualized and rendered using mapping tools, adding legends, scale bars, and other elements. Finally, the monitoring results are output as image files or in a GIS-compatible format for subsequent analysis or publication.

[0070] Preferably, it also includes a dynamic baseline construction module: using long-term historical data, a dynamic ecological baseline database that varies with the seasons is established for different geographical units (such as zoning by land use type), storing the normal value range of each ecological parameter in the same period of previous years.

[0071] Preferably, the system also includes an intelligent early warning module: This module compares the real-time retrieved ecological parameter values ​​with the dynamic baseline at the corresponding spatiotemporal location. If the parameter value P falls outside the normal range of the baseline, it is marked as an abnormal state, and the deviation is calculated. Further, the system obtains the short-term parameter values ​​of the abnormal pixels corresponding to parameter value P over the most recent N days (e.g., 30 days) for trend fitting and calculates the trend slope S. Combining the deviation and trend slope, a graded assessment is performed using a predefined decision matrix, triggering different levels (e.g., attention, alert, severe) of ecological anomaly early warning.

[0072] Preferably, it also includes a model adaptive update module: during long-term operation, it continuously collects new remote sensing-meteorological data pairs. Periodically (e.g., annually) or when the model's prediction error (e.g., root mean square error) in the reference area is detected to continuously exceed a preset range (determined by trend verification of the error time series), it automatically triggers incremental learning or full retraining of the meteorological-spectral mapping model to ensure that the model adapts to long-term environmental changes.

[0073] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. An ecological environment monitoring device based on remote sensing images, characterized in that, include: The data acquisition and preprocessing module is used to acquire the current remote sensing images and current meteorological data of the target area and perform preprocessing, and simultaneously acquire historical remote sensing images and corresponding historical meteorological data to construct a time-series dataset. The land cover identification module is used to identify the land cover type of a target area based on spectral reflectance characteristics and a classification algorithm. The reference area selection module is used to select fixed reference objects in historical remote sensing images, delineate a reference area with the reference object as the center, and use it as the benchmark area for learning the meteorological-spectral relationship. It is assumed that the land cover type of this area remains unchanged in time. The meteorological-spectral mapping model construction module is used to learn the dynamic relationship between meteorological data and spectral reflectance characteristics within a reference area using a deep learning model, and to establish a mapping function from meteorological data to spectral reflectance characteristics. The spectral anomaly detection and regional correction module is used to input the current remote sensing image and current meteorological data into the trained meteorological-spectral mapping model, predict the expected spectrum of the reference area under the current meteorological conditions, compare the actual spectrum of the target area in the current remote sensing image with the expected spectrum output by the model, and if the difference exceeds the set threshold, it is determined to be an area where the spectral-meteorological relationship does not correspond. The boundary of the area is then cropped and corrected to obtain a new target area after meteorological impact correction. The ecological parameter inversion and output module is used to invert and map ecological parameters based on the corrected new target area.

2. The ecological environment monitoring device based on remote sensing images according to claim 1, characterized in that, Current and historical meteorological data include at least precipitation, temperature, humidity, and cloud cover data.

3. The ecological environment monitoring device based on remote sensing images according to claim 2, characterized in that, Ecological parameters include NDVI, leaf area index, surface temperature, and vegetation cover.

4. The ecological environment monitoring device based on remote sensing images according to claim 3, characterized in that, The fixed reference objects are natural or man-made features that are spectrally and spatially stable in long-term remote sensing images.

5. The ecological environment monitoring device based on remote sensing images according to claim 4, characterized in that, It also includes a dynamic baseline construction module, which is used to establish a dynamic ecological baseline database that changes with the seasons for different geographical units based on long-term historical data.

6. The ecological environment monitoring device based on remote sensing images according to claim 5, characterized in that, It also includes an intelligent early warning module, which compares real-time ecological parameters with the dynamic baseline of the corresponding spatiotemporal location, and combines parameter change trend analysis to conduct graded assessment and early warning of ecological anomalies.

7. The ecological environment monitoring device based on remote sensing images according to claim 6, characterized in that, The intelligent early warning module performs state comparison based on the following logic: Obtain the parameter value P obtained from the inversion of the current spatiotemporal location; The normal value range extracted from the dynamic ecological baseline database, corresponding to the current pixel geographic location and the current time period; Determine whether the parameter value P falls within the normal range: If P ∈ the normal value range, it is marked as normal and will not proceed to the subsequent warning process; If P If the value falls within the normal range, it is marked as an abnormal state, and the deviation is calculated.

8. The ecological environment monitoring device based on remote sensing images according to claim 7, characterized in that, The intelligent early warning module performs trend analysis based on the following logic: Input a short time series parameter sequence: obtain the time series parameter values ​​of the corresponding anomalous pixels for the most recent N days, with the current parameter value P as the end point of the sequence; Trend fitting: Perform linear regression on short-time series data and calculate the trend slope S; Decision matrix construction: Construct a decision matrix with deviation as the horizontal axis and trend slope S as the vertical axis, and predefine the warning level and urgency of each cell in the matrix.

9. The ecological environment monitoring device based on remote sensing images according to claim 8, characterized in that, It also includes a model adaptive update module, which is used to continuously collect remote sensing images and synchronous meteorological data during long-term monitoring, and to trigger incremental learning or retraining of the meteorological-spectral mapping model periodically or when the prediction error of the meteorological-spectral mapping model is detected to continuously exceed the preset range.

10. The ecological environment monitoring device based on remote sensing images according to claim 9, characterized in that, The model adaptive update module determines whether the prediction error of the meteorological-spectral mapping model continuously exceeds a preset range based on the following logic: Within the reference area, the expected spectrum predicted by the model based on current meteorological data is compared with the actual spectrum extracted from the current remote sensing image. Calculate the root mean square error between the two; The root mean square error obtained from each processing step, along with its timestamp, is stored separately to form a continuous error time series. If a statistical test is performed on the error time series and an upward trend in the error is detected, and the current error value is already at the historical preset level, then it is determined that the performance is continuously degrading.

Images