Garden wetland ecological health comprehensive monitoring method and system based on hyperspectral imaging

Abstract

This invention discloses a comprehensive monitoring method and system for the ecological health of garden wetlands based on hyperspectral imaging, belonging to the field of ecological environment remote sensing monitoring technology. It acquires hyperspectral image data and corresponding climate data of the target garden wetland at multiple temporal phases, extracts ecological characterization features of vegetation, water bodies, and soil, and accumulates and converts the climate data into a climate-driven sequence. Based on the climate-driven sequence, it determines the boundary points of ecological response stages, constructs an ecological process scale, and maps multi-temporal observation data from calendar time to the ecological process scale. Then, it performs non-uniform rearrangement and collaborative correction on the initial ecological process sequence to obtain a stable ecological response feature sequence. Based on this, it extracts change trajectories, identifies abnormal change segments, and generates ecological health status evaluation results and early warning information. This invention can reduce misjudgments caused by phenological shifts.

Description

Technical Field

[0001] This invention relates to the field of remote sensing monitoring technology for ecological environment, specifically to a method and system for comprehensive monitoring of the ecological health of garden wetlands based on hyperspectral imaging. Background Technology

[0002] With the deepening application of hyperspectral imaging technology in ecological monitoring, the dynamic assessment of the ecological health of garden wetlands is gradually shifting from traditional manual surveys to quantitative analysis methods based on multi-temporal remote sensing data. Existing technologies typically construct multi-temporal hyperspectral time series to detect changes and analyze trends in ecological elements such as vegetation, water bodies, and soil, thereby achieving a comprehensive evaluation of wetland ecological status. However, in practical applications, due to the influence of climatic conditions such as temperature and precipitation in different years, wetland vegetation and related ecological processes often exhibit significant phenological differences. This leads to inconsistent ecological states at the same calendar time, resulting in biases in time-aligned sequence analysis results. Normal growth rhythm changes are easily misjudged as ecological degradation or abnormal fluctuations, seriously affecting the accuracy and reliability of monitoring results. Summary of the Invention

[0003] The purpose of this invention is to provide a comprehensive monitoring method and system for the ecological health of garden wetlands based on hyperspectral imaging, so as to overcome the shortcomings of the prior art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a comprehensive monitoring method for the ecological health of garden wetlands based on hyperspectral imaging, comprising: S1. Acquire hyperspectral image data of the target garden wetland at multiple time phases, and simultaneously acquire climate data of the corresponding time phases to establish a multi-temporal observation data set for the same spatial location; S2. Perform consistency processing on hyperspectral image data to obtain spatially aligned multi-temporal data, and extract ecological characteristics reflecting the state of vegetation, water bodies and soil. At the same time, perform cumulative transformation on climate data to form climate driving sequences. S3. Determine the stage boundary points of ecological response based on the climate-driven sequence, and construct a continuous ecological process scale accordingly. Map multi-temporal observation data from discrete calendar time to the ecological process scale to obtain the initial ecological process sequence. S4. The initial ecological process sequence is non-uniformly rearranged according to the ecological process scale to align data in the equivalent ecological response stage in different years, forming an ecological alignment sequence. S5. In the ecological alignment sequence, the ecological characterization features are collaboratively corrected to eliminate the system differences caused by phenological shifts and obtain a stable ecological response feature sequence. S6. Extract change trajectories based on ecological response feature sequences and identify anomalous change segments that deviate from the consistency of climate-driven processes; S7. Generate the ecological health status evaluation results of garden wetlands based on the change trajectory and abnormal change segments, and output the corresponding early warning information when abnormal change segments are detected.

[0005] Preferably, step S2 includes: The hyperspectral image data were sequentially processed for radiometric consistency, atmospheric and illumination consistency, and geometric consistency to obtain spatially aligned multi-temporal data. Based on spatially aligned multi-temporal data, ecological characterization features corresponding to vegetation, water bodies and soil are extracted respectively. Temperature, precipitation, and light data were processed at a diurnal scale and correlated with the ecological characteristics of the corresponding time phases.

[0006] Preferably, climate data is accumulated and transformed to form climate driving sequences, including: The baseline temperature is determined based on the dominant vegetation type of the target garden wetland, and the first day of the five consecutive days with an average daily temperature not lower than the baseline temperature and cumulative precipitation not less than 2 mm is taken as the ecological start date. From the ecological start date, the daily temperature contribution, effective precipitation and effective light intensity are accumulated daily. The daily accumulation results are matched with the ecological characteristics in the temporal sequence to form a climate driving sequence.

[0007] Preferably, step S3 includes: The ecological response initiation and decay boundaries are determined based on the cumulative change rate of the climate-driven sequence. Between the ecological response initiation and decay boundaries, a continuous ecological process scale from 0 to 1 is constructed based on the cumulative progress of the climate-driven quantities. The ecological characteristics corresponding to each time in the multi-temporal observation data are mapped to the continuous ecological process scale to obtain the initial ecological process sequence.

[0008] Preferably, step S4 includes: The initial ecological process sequence was divided into multiple continuous ecological segments according to the continuous ecological process scale, and the distribution range of ecological characteristics within each continuous ecological segment was statistically analyzed. Using the median value of ecological characteristics within a continuous ecological segment as a reference value, data that deviate from the reference value by more than 20% of the corresponding distribution range are locally shifted to adjacent continuous ecological segments. The initial ecological process sequence is rearranged based on the locally shifted ecological process scale, so that data in the equivalent ecological response stage in different years form an ecological alignment sequence.

[0009] Preferably, step S5 includes: In the ecological alignment sequence, multiple alignment units are divided according to the ecological process scale, and the central value and distribution range of the ecological characterization features within each alignment unit are statistically analyzed. When the deviation of an ecological characteristic from the center value exceeds 15% to 25% of the corresponding distribution range, the portion exceeding the allowable deviation range will be compressed towards the center value by a ratio of 50% to 70%. By combining climate-driven sequences, the direction of change of compressed ecological characteristics is verified, resulting in a stable ecological response characteristic sequence.

[0010] Preferably, the change direction verification includes: Under the same ecological process scale, the direction of change of climate driving forces and the direction of change of ecological characterization features between adjacent aligned units are read; When the direction of change of two consecutive aligned units of ecological characterization features is inconsistent with the direction of change of climate driving quantities, and the change is not determined to be a true anomalous change, the corresponding ecological characterization features will be marked as phenological offset data to be corrected. Based on the direction of change of ecological characterization features of adjacent aligned units, the phenological offset data to be corrected is directionally corrected to form a stable ecological response feature sequence.

[0011] Preferably, step S6 includes: Based on the stable ecological response feature sequence, the change in ecological characterization features between adjacent ecological process positions is calculated according to the ecological process scale to form a change trajectory. When the change in a continuous segment exceeds 1.5 times the overall average level for three consecutive ecological process intervals, the continuous segment is marked as a high-change segment. By comparing the direction of change of ecological characteristics in high-change sections with the climate-driven change trend, when the two are inconsistent and continue for more than two ecological process intervals, the corresponding section is identified as an anomalous change segment.

[0012] Preferably, step S7 includes: The change trajectory was divided into multiple evaluation segments according to the ecological process scale, and the magnitude, direction, and proportion of abnormal change segments in each evaluation segment were statistically analyzed. Using the median value of the change range of each evaluation segment as the benchmark value, the evaluation segment with a change range that is consistently lower than 70% of the benchmark value and the proportion of abnormal change segments is less than 10% is judged as a stable state, and the evaluation segment with a change range that is higher than 130% of the benchmark value or the proportion of abnormal change segments is greater than 20% is judged as an abnormal state. When two or more consecutive evaluation segments are determined to be in an abnormal state, the corresponding spatial location is identified as being in an ecological risk state, and warning information of the corresponding level is output based on the proportion of abnormal change segments.

[0013] This invention also provides a comprehensive monitoring system for the ecological health of garden wetlands based on hyperspectral imaging, comprising: The data acquisition module acquires hyperspectral image data of the target garden wetland at multiple time phases and simultaneously acquires climate data of the corresponding time phases to establish a multi-temporal observation data set for the same spatial location. The feature extraction module performs consistency processing on hyperspectral image data to obtain spatially aligned multi-temporal data, and extracts ecological characteristics reflecting the state of vegetation, water bodies and soil. At the same time, it accumulates and transforms climate data to form climate-driven sequences. The ecological process mapping module determines the stage boundary points of ecological response based on the climate-driven sequence, and constructs a continuous ecological process scale accordingly. It maps multi-temporal observation data from discrete calendar time to the ecological process scale to obtain the initial ecological process sequence. The non-uniform rearrangement alignment module performs non-uniform rearrangement of the initial ecological process sequence according to the ecological process scale, so that data in the equivalent ecological response stage in different years are aligned to form an ecological alignment sequence. The collaborative correction module performs collaborative correction on the ecological characterization features in the ecological alignment sequence to eliminate system differences caused by phenological shifts and obtain a stable ecological response feature sequence. The anomaly identification module extracts change trajectories based on ecological response feature sequences and identifies anomalous change segments that deviate from the consistency of climate-driven processes. The early warning module generates an assessment of the ecological health status of the garden wetland based on the change trajectory and abnormal change segments, and outputs corresponding early warning information when abnormal change segments are detected.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows: 1. This invention transforms the benchmark for judging ecological changes in garden wetlands from calendar time to ecological process. Vegetation spectral response, changes in aquatic algae, and soil moisture are not simply controlled by date, but are driven by the cumulative effects of temperature, precipitation, and sunlight. Existing technologies compare hyperspectral data from multiple time phases on the same date, easily misinterpreting spectral differences caused by earlier or later phenological events as ecological degradation. This invention determines the boundary points of ecological response stages through climate-driven sequences and constructs a continuous ecological process scale, enabling comparisons of data from different years under equivalent ecological response stages, thereby reducing systematic bias caused by phenological phase misalignment at the source.

[0015] 2. This invention differs from existing technologies that only perform geometric registration or radiometric correction on images. This invention further performs non-uniform rearrangement and co-correction on the initial ecological process sequence, ensuring that ecological characteristics align with climate-driven processes. Its core approach is not simply data smoothing, but rather correcting the response stages of the same spatial location in different years based on ecological process scales, making the trajectories of vegetation, water bodies, and soil characteristics more consistent with actual ecological processes. This allows for a more accurate distinction between normal fluctuations driven by climate and abnormal changes caused by non-climate factors, reducing false alarms and improving the reliability of ecological health assessments and early warning information. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of the method of the present invention.

[0018] Figure 2 This is a flowchart of the system modules of the present invention. Detailed Implementation

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

[0020] Example 1, please refer to Figure 1 As shown in this embodiment, the comprehensive monitoring method for the ecological health of garden wetlands based on hyperspectral imaging includes: S1. Acquire hyperspectral image data of the target garden wetland at multiple time phases, and simultaneously acquire climate data of the corresponding time phases to establish a multi-temporal observation data set for the same spatial location; In one specific embodiment, step S1 is used to form the original observation basis required for subsequent ecological process correction and ecological health analysis. Specifically, the target garden wetland area to be monitored is first determined, and the target garden wetland is divided into several monitoring units according to the ecological composition of the area, such as water bodies, emergent plants, submerged plants, wetland grasslands, bare mudflats, and bank slope green spaces. Each monitoring unit can be determined according to a fixed grid, ecological function zoning, or typical sample plot method to ensure that the same spatial location can be repeatedly identified and continuously tracked at different monitoring time phases.

[0021] Hyperspectral imagery data was acquired at multiple time phases of the target garden wetland. These time phases preferably cover key ecological stages such as vegetation germination, vigorous growth, decline, and the non-growing season, but can also be acquired continuously on a monthly, ten-day, or preset monitoring cycle. The hyperspectral imagery data for each time phase includes at least continuous spectral reflectance information of the target garden wetland in the visible to near-infrared range, used to characterize differences in ecological states such as vegetation growth, water optical state, and soil moisture. To improve the comparability of data from different time phases, similar observation height, observation angle, spatial resolution, and acquisition time periods were preferred during the acquisition process, and the date, time, spatial coordinate range, and image coverage area corresponding to each acquisition were recorded.

[0022] While acquiring hyperspectral imagery data, corresponding temporal climate data is acquired simultaneously. The climate data includes at least temperature, precipitation, and illumination information. Temperature data may include daily average temperature, daily maximum temperature, and daily minimum temperature; precipitation data may include daily precipitation and cumulative precipitation; and illumination information may include sunshine duration or solar irradiance. The temporal range of the climate data preferably covers the continuous time period between two adjacent hyperspectral image acquisitions, not just the day of image acquisition, to facilitate subsequent characterization of the cumulative response characteristics of vegetation or wetland ecological processes. Climate data sources can include measured meteorological records from within or adjacent areas of the target garden wetland, or a combination of regional meteorological data and supplementary field observations.

[0023] The hyperspectral image data for each time phase is associated with the corresponding climate data using temporal and spatial markers. Specifically, the image acquisition date can be used as a reference time point, and climate data within a preset time window prior to that reference time point is bound to the image for that time phase, forming a linked record of time phase-spatial location-spectral information-climate information. For the same spatial location, pixels, grid units, or ecological monitoring units can be used as basic objects. Hyperspectral image information acquired on different dates is arranged sequentially according to acquisition time, and the corresponding temperature, precipitation, and illumination information are simultaneously written into the data record of that object.

[0024] Through the above processing, a multi-temporal observation dataset of the same spatial location is established. This multi-temporal observation dataset not only includes hyperspectral image data of each spatial location at different times, but also includes climate-driven information corresponding to the ecological state of each temporal phase. This allows subsequent steps to incorporate climate background constraints while analyzing spectral changes, thereby providing a data foundation for distinguishing between actual ecological degradation and apparent changes caused by phenological differences.

[0025] S2. Perform consistency processing on hyperspectral image data to obtain spatially aligned multi-temporal data, and extract ecological characteristics reflecting the state of vegetation, water bodies and soil. At the same time, perform cumulative transformation on climate data to form climate-driven sequences.

[0026] First, radiometric consistency processing is performed on the hyperspectral image data obtained in step S1. Specifically, the original grayscale value, exposure time, and acquisition gain of each temporal image are read, dark current response is subtracted, and reflectivity is calculated using standard reflector data recorded synchronously during acquisition. When the signal-to-noise ratio of a certain band is less than 30, or the proportion of invalid pixels in that band to the effective coverage area of ​​that temporal phase exceeds 5%, that band is marked as a low-quality band and removed from subsequent processing. For a small number of missing bands between adjacent effective bands, the average reflectivity change trend of one effective band before and after the missing band is used to fill in the missing bands, ensuring that the hyperspectral image data from different acquisition time phases are consistent in radiometric dimensions and effective band range.

[0027] The hyperspectral image data, after radiometric consistency processing, underwent atmospheric and illumination consistency processing. Standard reflectors—whether stable hard paving, exposed dry soil, or artificially placed within the target garden wetland—were used as reference features. The reflectance values ​​of these reference features in images from different time phases were obtained and compared with the measured reflectance values. Based on the comparison results, proportional and offset corrections were performed on the reflectance of each band within the same time phase. Proportional correction ensures that the overall variation in the reflectance of the reference feature image matches the measured reflectance, while offset correction deducts the overall uplift caused by fog, water vapor, or scattering.

[0028] After processing, the reflectance difference of the same reference feature at different times is controlled within 10%. If it exceeds 10%, the reference feature is reselected for the time phase image or the time phase is marked as data to be reviewed.

[0029] Geometric consistency processing was performed on hyperspectral image data of each time phase to obtain spatially aligned multi-temporal data.

[0030] One time phase with complete coverage, minimal cloud and fog obstruction, and clear imagery is selected as the baseline time phase. Road intersections, shoreline turning points, bridge edges, fixed building corners, and long-term stable vegetation patch boundaries are selected as corresponding control points from the baseline time phase and other time phase images. The number of corresponding control points for each time phase is no less than 20 and they are evenly distributed within the target garden wetland area. Based on the positional differences between the corresponding control points, the time phase images to be processed are translated, rotated, scaled, and local deformation corrected. The images are then rearranged to the spatial grid of the baseline time phase using a bilinear resampling method.

[0031] After calibration, the residual deviation of the control points should not exceed 0.5 pixels. If it exceeds this range, control points should be added and recalibrated.

[0032] After obtaining spatially aligned multi-temporal data, ecological characterization features are extracted according to the same spatial location.

[0033] For vegetated areas, reflectance characteristics of green, red, red-edge, and near-infrared bands are extracted to form normalized vegetation difference characteristics, red-edge location characteristics, and red-edge slope characteristics. The normalized vegetation difference characteristics are obtained by the ratio of the difference between near-infrared and red reflectance to their combined value. The red-edge location characteristic is determined by the location with the fastest increase in reflectance from red to near-infrared. The red-edge slope characteristic is determined by the ratio of the increase in reflectance within the red-edge range to the corresponding band interval. For water areas, reflectance characteristics of blue, green, red, and near-infrared bands are extracted to form water turbidity-sensitive characteristics and algae growth-sensitive characteristics. For soil or tidal flat areas, reflectance variation characteristics of red, near-infrared, and short-wave infrared bands are extracted to form soil moisture characterization characteristics.

[0034] The climate data obtained in step S1 is processed for temporal consistency. Temperature data is processed into daily average temperature, daily maximum temperature, and daily minimum temperature; precipitation data is processed into daily precipitation; and sunshine information is processed into sunshine duration or daily solar irradiance. When climate data for a certain day is missing and the number of consecutive missing days does not exceed two, it is linearly filled in chronological order using similar climate data from adjacent dates before and after the missing date. When the number of consecutive missing days exceeds two, data from nearby meteorological observation points in the target garden wetland is used as a substitute, and the substitution marker is retained. After processing, each hyperspectral image corresponds to continuous diurnal climate data from the previous time phase to the current time phase.

[0035] The processed climate data is cumulatively transformed to form a climate-driven sequence. First, an ecological start date is determined, which is the first day of the year when the daily average temperature is not lower than the baseline temperature for five consecutive days, and the cumulative precipitation within those five days is not less than 2 mm. The baseline temperature is set according to the dominant vegetation type of the target garden wetland: 3℃ to 6℃ for hygrophytic herbaceous plants and 5℃ to 10℃ for emergent plants. If dominant vegetation data is unavailable, 5℃ is used. From the ecological start date, the daily temperature contribution, effective precipitation, and effective sunshine are accumulated daily. The daily temperature contribution is the portion of the daily average temperature exceeding the baseline temperature; if it does not exceed the baseline temperature, it is recorded as 0. Effective precipitation is the accumulated value of daily precipitation; when the daily precipitation exceeds 50 mm, the excess portion is counted as half. Effective sunshine is accumulated based on sunshine duration or daily solar irradiance. This yields a climate-driven sequence corresponding to the hyperspectral imagery of each time phase. This climate-driven sequence is used to characterize the ecological response process of the same spatial location in different years.

[0036] S3. Determine the stage boundary points of ecological response based on the climate-driven sequence, and construct a continuous ecological process scale accordingly. Map multi-temporal observation data from discrete calendar time to the ecological process scale to obtain the initial ecological process sequence.

[0037] The climate-driven sequence is a sequence formed by continuous daily accumulation from the ecological start date, with each day corresponding to one climate-driven quantity. The climate-driven sequence is first divided according to the hyperspectral image acquisition phase or a preset time interval, which can be 5 to 15 days, preferably 10 days.

[0038] For each time interval, the climate driving amount on the end date and the climate driving amount on the start date of the time interval are read, and the difference between the two is taken as the increase in climate driving amount within the time interval; then the increase in climate driving amount in the current time interval is compared with the increase in climate driving amount in the previous time interval to obtain the percentage change in the increase in climate driving amount between adjacent time intervals.

[0039] When the climate-driven increase in the current time interval is 0, the average of all non-zero climate-driven increases from the ecological start date to the current time interval is used as the comparison benchmark; if there is still no non-zero climate-driven increase, then the time interval is not included in the determination of the boundary point.

[0040] The percentage change in increase is retrieved in chronological order. When the percentage change in increase gradually increases from less than 10% to more than 25% over three consecutive time intervals, and the corresponding hyperspectral ecological characteristics do not show the opposite change, the end date of the third time interval is determined as the ecological response initiation boundary. When the percentage change in increase gradually decreases from more than 25% to less than 10% over three consecutive time intervals, and the changes in ecological characteristics tend to level off after the three time intervals, the end date of the third time interval is determined as the ecological response decay boundary.

[0041] The thresholds of 10% and 25% are the dividing line for determining the boundary. Specifically, they can be adjusted based on the historical climate driving sequence of the target garden wetland over the past 3 to 5 years. The adjustment rule is as follows: if the ecological response starts earlier in the historical sequence, the lower threshold will be set between 5% and 10%; if the ecological response starts later, the higher threshold will be set between 25% and 35%.

[0042] After determining the ecological response initiation boundary and the ecological response attenuation boundary, a continuous ecological process scale is constructed.

[0043] Using the ecological response initiation boundary as the starting point of the ecological process and the ecological response decay boundary as the ending point, the daily corresponding climate driving forces between the two are extracted, and the total cumulative amount of climate driving forces within this interval is determined. For any date within this interval, the cumulative value of climate driving forces from the ecological response initiation boundary to that date is read, and the proportion of this cumulative value to the total cumulative amount is used as the position of that date on the continuous ecological process scale. The continuous ecological process scale is a continuous interval from 0 to 1, where 0 corresponds to the ecological response initiation boundary, 1 corresponds to the ecological response decay boundary, and the values ​​between 0 and 1 represent the degree of completion of the ecological response process.

[0044] If a date is earlier than the ecological response initiation boundary, it is marked as pre-ecological response data and is not included in the 0-1 interval for phase alignment; if a date is later than the ecological response decay boundary, it is marked as post-ecological response data and is used for reference in subsequent decline states.

[0045] To avoid abrupt jumps in the scale due to extreme single-day precipitation or short-term high temperatures, daily climate drivers are smoothed before calculating the continuous ecological process scale. The smoothing process uses the average of the climate drivers over five days (two days before and after the current day, plus the current day itself) to replace the daily climate driver. When there are fewer than five days at the start or end of a sequence, the average of existing dates is used instead. After this processing, the continuous ecological process scale no longer uses calendar dates as the sole reference, but instead characterizes the wetland ecological response stage by the cumulative progress of climate drivers.

[0046] After obtaining the continuous ecological process scale, the acquired multi-temporal observation data is mapped from discrete calendar time to the ecological process scale to obtain the initial ecological process sequence.

[0047] For each phase of data in the multi-temporal observation data, the image acquisition date corresponding to that phase of data is read, and the ecological process position corresponding to that date is found in the continuous ecological process scale. If the image acquisition date is consistent with the date in the continuous ecological process scale, the ecological process position is directly assigned to that phase of data. If the image acquisition date is between two adjacent dates, or if the climate data is obtained using a non-daily recording method, the position of that phase of data in the continuous ecological process scale is determined according to the cumulative change ratio of climate driving factors between adjacent dates.

[0048] Using the same spatial location as the basic object, the hyperspectral ecological characterization features of this spatial location at different collection dates are bound to the corresponding ecological process locations, forming a recording unit of "spatial location, ecological process location, ecological characterization feature, and climate driving force". These recording units are then arranged in ascending order of ecological process location, transforming the multi-temporal observation data originally arranged according to calendar time into an initial ecological process sequence based on the ecological process scale. For multiple temporal data points within the same year with the same ecological process location or a difference less than 0.02, the average value of their ecological characterization features is taken as the representative value of that ecological process location. For data points in similar ecological process locations across different years, their year and spatial location labels are retained for subsequent cross-year alignment and change analysis.

[0049] Through the above mapping process, the data of each phase in the initial ecological process sequence have a unified ecological process reference, enabling subsequent processing to be carried out around the same ecological response stage, rather than just comparing based on the same calendar date.

[0050] S4. The initial ecological process sequence is non-uniformly rearranged according to the ecological process scale to align data in the equivalent ecological response stage in different years, forming an ecological alignment sequence.

[0051] Based on the ecological process scale, the continuous ecological process scale from 0 to 1 is divided into several continuous ecological segments. The scale interval of each ecological segment is set to 0.05 to 0.1, preferably 0.05. When the hyperspectral image acquisition time of the target garden wetland is less than 12, the scale interval of the ecological segment is preferably 0.1 to avoid too little data in a single segment.

[0052] For each year, the ecological process scale corresponding to each spatial location in the initial ecological process sequence is read and assigned to the corresponding ecological segment. If a data point is located at the boundary between two ecological segments, it is assigned to the segment following its ecological process scale. Subsequently, within each ecological segment, the ecological characteristics corresponding to vegetation, water bodies, and soil are statistically analyzed. These ecological characteristics include the obtained normalized difference vegetation characteristics, red-edge location characteristics, water turbidity sensitivity characteristics, algal growth sensitivity characteristics, and soil moisture characteristics. For each ecological characteristic, the characteristic values ​​for all years and all corresponding spatial locations within the ecological segment are read, the maximum and minimum characteristic values ​​are determined, and the difference between the maximum and minimum characteristic values ​​is taken as the characteristic distribution range of that ecological characteristic within the ecological segment.

[0053] After completing the ecological segmentation and characteristic distribution range statistics, local deviation judgments were made on the data within each ecological segment. Within each ecological segment, all feature values ​​of the same ecological characteristic were arranged from low to high according to their numerical values; when the number of feature values ​​was odd, the feature value in the middle of the sorted position was taken as the reference value for that ecological segment; when the number of feature values ​​was even, the average of the two middle feature values ​​was taken as the reference value for that ecological segment.

[0054] For any data point in any year within the same ecological segment, calculate the degree of deviation of its ecological characterization characteristics relative to the reference value. The degree of deviation is the proportion of the absolute difference between the ecological characterization characteristics of the data and the reference value to the characteristic distribution range of the ecological segment. When the characteristic distribution range is 0, it indicates that there is no discrete difference in the corresponding ecological characterization characteristics within the ecological segment, and the data is not subject to local adjustment.

[0055] When the deviation exceeds 20%, further compare the closeness of the data with the reference value of the previous ecological segment and the reference value of the next ecological segment. If its ecological characteristics are closer to the reference value of the previous ecological segment, shift the ecological process scale corresponding to the data forward by no more than one ecological segment interval. If its ecological characteristics are closer to the reference value of the next ecological segment, shift the ecological process scale corresponding to the data backward by no more than one ecological segment interval. If the difference in closeness between adjacent ecological segments is less than 5%, keep the original ecological process scale unchanged to avoid over-adjustment.

[0056] After obtaining the adjusted ecological process scale, the initial ecological process sequence is reordered. Taking the same spatial location as the processing object, the ecological characteristics of that spatial location in different years and their adjusted ecological process scales are read and rearranged in ascending order of the adjusted ecological process scales.

[0057] During reordering, the changing trend of ecological characteristics between adjacent data is first determined. The changing trend is determined by whether the ecological characteristics of the later data increase, decrease, or remain stable relative to the earlier data. When the changing trends of data from different years within the same ecological segment are consistent, and the adjusted ecological process scale is within the same ecological segment, this part of the data is aggregated as data of the equivalent ecological response stage. When a certain data has a changing trend opposite to that of most years within the same ecological segment, its year label is retained and it is not included in the aggregation result for subsequent abnormal change judgment.

[0058] By reordering as described above, data from different years that are in the equivalent ecological response stage are arranged in a unified ecological process order, forming a consistent ecological alignment sequence across years. The ecological alignment sequence includes spatial location, year marker, adjusted ecological process scale, and corresponding ecological characterization features.

[0059] After forming the ecological alignment sequence, its continuity is checked and missing data is filled in. For the same spatial location and the same year, adjacent data in the ecological alignment sequence are read in the order of ecological process scale, and the ecological process scale interval between adjacent data is determined; when the interval exceeds 0.1, the interval is determined as a missing segment.

[0060] For missing segments, first read the ecological process scale and ecological characterization features of the preceding and following data for the missing segment. Then, determine the imputation value according to the proportion of the distance between the missing position and the preceding and following ecological process scales. The process of determining the imputation value is as follows: use the ecological characterization features of the preceding data as the starting value and the ecological characterization features of the following data as the ending value. Weight the starting and ending values ​​according to the distance of the missing position from the preceding and following data, with closer endpoints receiving higher weights. If the missing segment is located at the beginning or end of the ecological alignment sequence, extrapolation imputation is performed using the change directions of two adjacent existing data. Extrapolation imputation should not exceed one ecological segment interval.

[0061] After the completion is completed, the ecological process scale interval of adjacent data is checked again to ensure that it does not exceed 0.1, thereby obtaining a complete ecological alignment sequence that can be used for subsequent collaborative correction.

[0062] S5. In the ecological alignment sequence, ecological characterization features are collaboratively corrected to eliminate system differences caused by phenological shifts and obtain a stable ecological response feature sequence.

[0063] Based on the ecological process scale, data belonging to the same spatial location and at the same or similar ecological process positions in the ecological alignment sequence are aggregated; the similar ecological process positions refer to the range where the difference in ecological process scale does not exceed 0.025. Multiple alignment units are divided according to the scale interval of 0.05 to 0.1, preferably 0.05; when there are fewer than 3 valid data in a single alignment unit, it is merged with the adjacent alignment units with closer ecological process scales.

[0064] For each alignment unit, the ecological characteristics of vegetation, water bodies and soil are statistically analyzed. The maximum, minimum and all valid values ​​of the same ecological characteristic within the alignment unit are read, and the difference between the maximum and minimum values ​​is determined as the distribution range. At the same time, all valid values ​​are arranged in numerical order, and the value in the middle position is taken as the center value. If the number of valid values ​​is even, the average of the two middle values ​​is taken as the center value.

[0065] Within each alignment unit, using the center value obtained in step 51 as a reference, the offset of the corresponding ecological characterization feature for each year is calculated. The ecological characterization feature for a given year within the alignment unit is read, and the absolute difference between this ecological characterization feature and the center value is taken as the offset. The proportion of this offset to the distribution range of the alignment unit is then determined. When the distribution range is 0, it indicates that the corresponding ecological characterization feature within the alignment unit does not exhibit discrete differences, and amplitude compression is not performed. The offset judgment limit is set to 15% to 25% of the distribution range, preferably 20%. When there are fewer than 5 valid data points within the alignment unit, 25% is used; when there are at least 5 valid data points and the dispersion is low, 15% to 20% is used. When the offset of a certain ecological characterization feature exceeds the judgment limit, the portion exceeding the allowable offset range between the ecological characterization feature and the center value is subtracted towards the center value by 50% to 70%, retaining the portion not exceeding the allowable offset range, so that the ecological characterization feature converges towards the center value while maintaining its original direction of change.

[0066] Consistency verification was performed on the amplitude-compressed ecological characterization features by combining climate-driven sequences. Under the same ecological process scale, the direction of change of climate-driven quantities between adjacent alignment units was read, as well as the direction of change of ecological characterization features between adjacent alignment units within the same spatial location and year. The direction of change included increase, decrease, and essentially unchanged, where essentially unchanged meant that the change amplitude between adjacent alignment units did not exceed 5% of the distribution range of the ecological characterization feature. When the climate-driven quantity showed an increase, and two consecutive alignment units of the ecological characterization feature showed changes in opposite directions, and these opposite changes were not identified as anomalous segments, the ecological characterization feature was marked as phenological shift data requiring correction. When the climate-driven quantity showed a decrease, the same method was used to determine whether there was an opposite change.

[0067] For phenological offset data to be corrected, the direction of change of the same ecological characteristic within two adjacent alignment units is used as a reference to correct the direction of change of adjacent values ​​so that the direction of change of adjacent values ​​is consistent with the direction of climate-driven change; the corrected values ​​shall not exceed the distribution range boundary of the alignment unit.

[0068] Using the same spatial location as the basic object, the ecological characteristics of each alignment unit are arranged in ascending order of ecological process scale, and the year marker, ecological process scale, climate driving quantity, and corresponding vegetation, water body, and soil ecological characteristics are retained. In the case where there are multiple corrected feature values ​​within the same alignment unit, the average result is taken as the representative feature value of the alignment unit, while the original year markers are retained for subsequent traceability.

[0069] After sorting, check for gaps between adjacent aligned units. If gaps exist, fill them in based on the changing trends of ecological characteristics of adjacent aligned units, with the filler value located between adjacent feature values. The resulting stable ecological response feature sequence is used for subsequent change trajectory extraction and anomalous change fragment identification.

[0070] S6. Extract change trajectories based on ecological response feature sequences and identify anomalous change segments that deviate from the consistency of climate-driven processes.

[0071] Using the same spatial location as the basic object, consecutive adjacent data in the stable ecological response feature sequence are read in ascending order of ecological process scale, and the changes in the ecological characterization features corresponding to vegetation, water bodies, and soil are extracted respectively. The change is the difference between the ecological characterization feature of the later ecological process location and the ecological characterization feature of the previous ecological process location; when the difference is positive, it is recorded as an increase; when the difference is negative, it is recorded as a decrease; when the absolute value of the difference does not exceed 5% of the distribution range of the ecological characterization feature within the entire ecological process scale, it is recorded as a basically stable change.

[0072] The changes, directions, and corresponding spatial locations between adjacent positions are cumulatively recorded according to the ecological process scale sequence to form a change trajectory sequence. The change trajectory sequence includes at least the starting position of the ecological process, the ending position of the ecological process, the change amount of ecological characteristics, the direction of change, and the cumulative change state, which is used to characterize the continuous change process of the ecological state of garden wetlands in the ecological process dimension.

[0073] The change trajectory sequence is divided into several continuous segments according to the ecological process scale. Each continuous segment contains no less than 3 adjacent ecological process intervals. When the number of intervals in a single segment is less than 3, it is merged into the segment where the adjacent ecological processes are closer.

[0074] For each continuous segment, the absolute value of the change in each ecological characteristic is read, and the average of all absolute values ​​of change within the segment is taken as the average level of the segment. The difference between the maximum and minimum absolute values ​​of change is taken as the fluctuation range of the segment. Simultaneously, the absolute values ​​of change in all continuous segments within the change trajectory sequence are read, and their average is taken as the overall average level. If the absolute value of change in three consecutive ecological process intervals within a continuous segment exceeds 1.5 times the overall average level, and the direction of change in these three ecological process intervals is consistent, then the continuous segment is marked as a high-change segment; if the direction of change changes repeatedly, it is only recorded as a fluctuating segment and not directly designated as a high-change segment.

[0075] After identifying high-variability segments, anomalous change fragments are identified by combining climate-driven sequences. First, based on the ecological process scale, the climate-driven quantities in the climate-driven sequence are correlated with the changes in ecological characteristics in the change trajectory sequence, ensuring that the same ecological process location has a corresponding climate-driven change trend and ecological characteristic change direction. The climate-driven change trend is determined by the increase, decrease, or near-stability of climate-driven quantities between adjacent ecological process locations, where near-stability means that the change in climate-driven quantities between adjacent ecological process locations does not exceed 5% of the annual climate-driven quantity distribution range. Subsequently, the direction of ecological characteristic change under the same ecological process scale is compared with the climate-driven change trend. When the climate-driven change trend is increasing, but the ecological characteristics continuously show a decrease contrary to the normal ecological response, or when the climate-driven change trend is decreasing, but the ecological characteristics continuously show an increase inconsistent with the normal ecological response, and this inconsistency lasts for more than two ecological process intervals, the corresponding segment is identified as an anomalous change fragment. For cases where different ecological characteristics reflect different directions, a positive or negative response relationship is pre-determined according to the ecological meaning of the ecological characteristics, and then a consistency judgment is made based on this relationship.

[0076] After identifying anomalous change segments, boundary correction is performed. The starting and ending ecological process positions of each anomalous change segment are read, and the direction and magnitude of change in adjacent segments are examined. If the direction of change in the preceding ecological process interval is consistent with that of the anomalous change segment, and the absolute value of its change exceeds 60% of the average absolute value of change within the anomalous change segment, the starting boundary of the anomalous change segment is extended forward by one ecological process interval. If the following ecological process interval meets the same condition, the ending boundary of the anomalous change segment is extended backward by one ecological process interval.

[0077] If the ecological process scale interval between two adjacent anomalous change segments is less than 0.05, and the corresponding ecological characteristic types and change directions are the same, then the two segments will be merged into the same anomalous change segment; if the change directions are different, then they will be retained separately.

[0078] After boundary correction is completed, the output includes the spatial location, the location of the start of the abnormal ecological process, the location of the end of the abnormal ecological process, the corresponding ecological characteristics, and the direction of the abnormal change.

[0079] S7. Generate the ecological health status evaluation results of garden wetlands based on the change trajectory and abnormal change segments, and output the corresponding early warning information when abnormal change segments are detected.

[0080] Using the same spatial location as the basic object, the change trajectory is read in order of increasing ecological process scale, and the ecological process scale is divided into several evaluation segments. The scale interval of each evaluation segment is set to 0.1 to 0.2, preferably 0.1. When the effective change trajectory in a certain evaluation segment is less than 3 ecological process intervals, the evaluation segment is merged with the adjacent evaluation segment.

[0081] For each evaluation segment, the direction and magnitude of changes in the corresponding ecological characteristics of vegetation, water bodies, and soil are statistically analyzed. The proportion of anomalous change segments is obtained by reading the length of anomalous change segments falling within the evaluation segment and comparing it with the ecological process scale length of the evaluation segment; when an anomalous change segment only partially falls within the evaluation segment, only the overlapping part is counted.

[0082] After obtaining the direction, magnitude, and proportion of abnormal changes in each evaluation segment, the state of the evaluation segment is determined. First, the magnitude of changes at the same spatial location across all evaluation segments is arranged numerically, and the magnitude of the change at the middle position is taken as the baseline value. When the number of evaluation segments is even, the average of the two middle magnitudes is taken as the baseline value. If the magnitude of changes in consecutive adjacent ecological process intervals within an evaluation segment is all below 70% of the baseline value, and the proportion of abnormal changes in that evaluation segment is less than 10%, then the evaluation segment is determined to be in a stable state. If the magnitude of changes within an evaluation segment is above 130% of the baseline value, or the proportion of abnormal changes in that evaluation segment is greater than 20%, then the evaluation segment is determined to be in an abnormal state. Otherwise, the evaluation segment is determined to be in a fluctuating state.

[0083] It should be noted that the 70%, 130%, 10%, and 20% mentioned are evaluation judgment limits, which can be calibrated based on historical monitoring data of more than 3 consecutive years. The calibrated values ​​are 60% to 80%, 120% to 150%, 5% to 15%, and 15% to 25%, respectively.

[0084] For the same spatial location, the stable state, fluctuating state, or abnormal state of each evaluation segment is read sequentially, and it is determined whether the abnormal state occurs continuously. When two or more consecutive evaluation segments are determined to be in an abnormal state, the spatial location is determined to be in an ecological risk state. The starting position of the ecological process of the first abnormal state evaluation segment is taken as the starting position of the ecological risk state, and the ending position of the ecological process of the last consecutive abnormal state evaluation segment is taken as the ending position of the ecological risk state.

[0085] If the interval between two adjacent ecological risk states does not exceed 0.1 ecological process scales, and the corresponding ecological characteristic types are the same and the direction of change is consistent, then the two ecological risk states will be merged and recorded; if the corresponding ecological characteristic types are different, then they will be recorded separately to preserve the differences in vegetation, water body and soil conditions.

[0086] When an ecological risk state exists, early warning information is generated based on the abnormal change segments corresponding to the ecological risk state and the scope of their ecological processes. The spatial location covered by the ecological risk state, the starting and ending locations of the ecological process, the proportion of abnormal change segments, the corresponding ecological characteristic types, and the direction of change are read and written into the early warning information. Simultaneously, the early warning level is determined based on the proportion of abnormal change segments: when the proportion is greater than 30%, it is marked as a high-level warning; when the proportion is between 20% and 30%, it is marked as a medium-level warning; and when the proportion is less than 20% and the conditions for determining an ecological risk state are met, it is marked as a low-level warning.

[0087] The warning information also includes the corresponding calendar time range, which is obtained by reverse lookup of the correspondence between the ecological process scale and the continuous ecological process scale formed in step 3, so as to convert the ecological process position into an executable monitoring time period.

[0088] Example 2: To verify the technical effectiveness of the comprehensive monitoring method for the ecological health of garden wetlands based on hyperspectral imaging described in this application, this example selects a garden wetland with an area of ​​approximately 20 hectares as the verification object. This garden wetland includes emergent plant areas, wetland grasslands, water areas, exposed tidal flat areas, and green slope areas. The target garden wetland is divided into 30 fixed monitoring units, including 18 vegetation monitoring units, 7 water monitoring units, and 5 soil and tidal flat monitoring units. Hyperspectral image data of multiple time phases are acquired over three consecutive years, and temperature, precipitation, and illumination data for each year are recorded simultaneously. After radiometric consistency processing, atmospheric and illumination consistency processing, and geometric consistency processing, normalized vegetation difference characteristics, red edge location characteristics, water turbidity sensitivity characteristics, algal growth sensitivity characteristics, and soil moisture characterization characteristics are extracted from the hyperspectral image data. Climate data is converted into climate-driven sequences according to the aforementioned methods, and an ecological process scale is further constructed.

[0089] For ease of comparison, Comparative Example 1 is provided in this embodiment. Comparative Example 1 uses a conventional calendar time alignment method, that is, directly comparing the hyperspectral ecological characterization features of the same or similar dates in different years, and extracting change trajectories and identifying anomalous changes accordingly. Example 2 adopts the method of this application, that is, firstly, the ecological response initiation boundary point and the ecological response attenuation boundary point are determined based on the climate driving sequence, then the observation data of each time phase are mapped to the ecological process scale, and after non-uniform rearrangement and collaborative correction, a stable ecological response feature sequence is formed, followed by change trajectory extraction and anomalous change segment identification.

[0090] In this embodiment, the spring warming and precipitation processes differed across the three years, resulting in inconsistent initiation times of vegetation ecological responses. When observed according to calendar time, some years had already entered their rapid growth phase by early April, while others were still in the early germination stages. If directly aligned by date, the hyperspectral ecological characterization features on the same date would show significant differences. Using the method described in this application, the data for each year are mapped to an ecological process scale of 0 to 1, enabling the alignment of data from different years at equivalent ecological response stages. Examples of the stage boundaries and ecological process scales are shown in Table 1.

[0091] Table 1. Ecological Process Data Table

[0092] As shown in Table 1, the ecological progress scale corresponding to the same calendar date varies across different years. For example, on April 10th, the ecological progress scale corresponds to 0.16 for year 1, only 0.05 for year 2, and 0.20 for year 3. Directly comparing the data for April 10th from the three years could lead to a misinterpretation of the lower vegetation characteristics in year 2 (due to a later ecological progress stage) as vegetation degradation. The method in this application repositions the ecological progress scale, shifting the data comparison from "same date" to "same ecological response stage."

[0093] Further comparisons were made between 12 of the 18 vegetation monitoring units that exhibited stable growth and had not undergone artificial pruning or replanting. The dispersion of normalized vegetation differences between Comparative Example 1 and Example 2 at the same stage was statistically analyzed. The statistical results are shown in Table 2.

[0094] Table 2 Project Comparison Data Table

[0095] As shown in Table 2, among the stable vegetation monitoring units that did not show significant degradation, Comparative Example 1 identified 7 monitoring units as having degraded vegetation due to inconsistent phenological stages in different years. Of these, only 2 were confirmed as truly degraded upon on-site verification, resulting in 5 misclassifications. In Example 2, after alignment under the ecological process scale, only 2 monitoring units were identified as having degraded vegetation, and both were consistent with the on-site verification results. This demonstrates that the method of this application can reduce the impact of apparent differences caused by earlier or later phenological stages on the assessment of ecological health.

[0096] For the water body area and the tidal flat area, this embodiment further compares the results of abnormal change segment identification. On-site verification includes water transparency measurement, chlorophyll content detection, tidal flat water content recording, and bank slope vegetation survey. The statistical results are shown in Table 3.

[0097] Table 3 Comparison of Monitoring Data

[0098] Table 3 shows that Comparative Example 1 identified 14 anomalies, of which 5 were confirmed through on-site verification and 9 were false alarms. Example 2 identified 6 anomalies, of which 5 were confirmed through on-site verification and 1 was a false alarm. The newly added false alarm in Example 2 was a localized turbidity change in the water monitoring unit after a short period of heavy rainfall; this change returned to the normal range in subsequent retests. The above results indicate that, given the annual climate differences and misalignment of ecological response stages in garden wetlands, Example 2 can reduce false alarms caused by improper time alignment.

[0099] To further quantify the consistency of ecological health assessment results, this embodiment uses the on-site verification results as a reference to statistically analyze the determination results of stable, fluctuating, and abnormal states. During the statistical analysis, if the determination result of a certain assessment section is consistent with the on-site verification result, it is recorded as consistent; if a stable or fluctuating state is determined to be an abnormal state, it is recorded as a false alarm; if an abnormal state is determined to be a stable or fluctuating state, it is recorded as a missed alarm. The statistical results are shown in Table 4.

[0100] Table 4 Data Analysis and Evaluation Table

[0101] As shown in Table 4, the consistency rate of judgment in Comparative Example 1 was 74.4%, and the consistency rate of judgment in Example 2 was 93.3%. The number of false alarm segments in Example 2 decreased from 18 to 4, and the number of missed alarm segments decreased from 5 to 2. These results demonstrate that this application, through climate-driven sequences, ecological process scaling, non-uniform rearrangement, and synergistic correction, enables the comparison of hyperspectral ecological characterization features from different years at equivalent ecological response stages, thereby improving the consistency between ecological health status assessment results and on-site verification results.

[0102] Furthermore, this embodiment also verified the output results of the early warning information. Comparative Example 1 issued 11 early warnings over 3 years, including 3 high-level warnings, 5 medium-level warnings, and 3 low-level warnings; after on-site verification, 4 warnings were confirmed as valid. Example 2 issued 5 early warnings, including 1 high-level warning, 3 medium-level warnings, and 1 low-level warning; after on-site verification, 4 warnings were confirmed as valid. In other words, Example 2 reduced the number of invalid warnings while maintaining a relatively stable number of valid warnings, making the early warning information more focused on the actual ecological risk status. The results are shown in Table 5.

[0103] Table 5. Comparison of Early Warning Data

[0104] In summary, this embodiment validates the beneficial effects of the proposed method using data from three consecutive years, multiple monitoring units, and on-site verification. The results show that in landscape wetland monitoring scenarios where annual phenological events are either advanced or delayed, the proposed method can convert multi-temporal hyperspectral data from discrete calendar time to an ecological process scale. Within this scale, it performs non-uniform rearrangement, collaborative correction, change trajectory extraction, and identification of anomalous change segments, thereby reducing misjudgments caused by phenological phase shifts, improving the consistency between ecological health status assessment results and on-site verification results, and reducing invalid early warning outputs.

[0105] Example 3, please refer to Figure 2 As shown in this embodiment, the comprehensive monitoring system for the ecological health of garden wetlands based on hyperspectral imaging includes: The data acquisition module acquires hyperspectral image data of the target garden wetland at multiple time phases and simultaneously acquires climate data of the corresponding time phases to establish a multi-temporal observation data set for the same spatial location. The feature extraction module performs consistency processing on hyperspectral image data to obtain spatially aligned multi-temporal data, and extracts ecological characteristics reflecting the state of vegetation, water bodies and soil. At the same time, it accumulates and transforms climate data to form climate-driven sequences. The ecological process mapping module determines the stage boundary points of ecological response based on the climate-driven sequence, and constructs a continuous ecological process scale accordingly. It maps multi-temporal observation data from discrete calendar time to the ecological process scale to obtain the initial ecological process sequence. The non-uniform rearrangement alignment module performs non-uniform rearrangement of the initial ecological process sequence according to the ecological process scale, so that data in the equivalent ecological response stage in different years are aligned to form an ecological alignment sequence. The collaborative correction module performs collaborative correction on the ecological characterization features in the ecological alignment sequence to eliminate system differences caused by phenological shifts and obtain a stable ecological response feature sequence. The anomaly identification module extracts change trajectories based on ecological response feature sequences and identifies anomalous change segments that deviate from the consistency of climate-driven processes. The early warning module generates an assessment of the ecological health status of the garden wetland based on the change trajectory and abnormal change segments, and outputs corresponding early warning information when abnormal change segments are detected.

[0106] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A comprehensive monitoring method for the ecological health of garden wetlands based on hyperspectral imaging, characterized in that, include: S1. Acquire hyperspectral image data of the target garden wetland at multiple time phases, and simultaneously acquire climate data of the corresponding time phases to establish a multi-temporal observation data set for the same spatial location; S2. Perform consistency processing on hyperspectral image data to obtain spatially aligned multi-temporal data, and extract ecological characteristics reflecting the state of vegetation, water bodies and soil. At the same time, perform cumulative transformation on climate data to form climate driving sequences. S3. Determine the stage boundary points of ecological response based on the climate-driven sequence, and construct a continuous ecological process scale accordingly. Map multi-temporal observation data from discrete calendar time to the ecological process scale to obtain the initial ecological process sequence. S4. The initial ecological process sequence is non-uniformly rearranged according to the ecological process scale to align data in the equivalent ecological response stage in different years, forming an ecological alignment sequence. S5. In the ecological alignment sequence, the ecological characterization features are collaboratively corrected to eliminate the system differences caused by phenological shifts and obtain a stable ecological response feature sequence. S6. Extract change trajectories based on ecological response feature sequences and identify anomalous change segments that deviate from the consistency of climate-driven processes; S7. Generate the ecological health status evaluation results of garden wetlands based on the change trajectory and abnormal change segments, and output the corresponding early warning information when abnormal change segments are detected.

2. The method for comprehensive monitoring of the ecological health of garden wetlands based on hyperspectral imaging according to claim 1, characterized in that, Step S2 includes: The hyperspectral image data were sequentially processed for radiometric consistency, atmospheric and illumination consistency, and geometric consistency to obtain spatially aligned multi-temporal data. Based on spatially aligned multi-temporal data, ecological characterization features corresponding to vegetation, water bodies and soil are extracted respectively. Temperature, precipitation, and light data were processed at a diurnal scale and correlated with the ecological characteristics of the corresponding time phases.

3. The method for comprehensive monitoring of the ecological health of garden wetlands based on hyperspectral imaging according to claim 1, characterized in that, Climate data is cumulatively transformed to form climate driving sequences, including: The baseline temperature is determined based on the dominant vegetation type of the target garden wetland, and the first day of the five consecutive days with a daily average temperature not lower than the baseline temperature and a cumulative precipitation of not less than 2 mm is taken as the ecological start date. From the ecological start date, the daily temperature contribution, effective precipitation and effective light intensity are accumulated daily. The daily accumulation results are matched with the ecological characteristics in the temporal sequence to form a climate driving sequence.

4. The method for comprehensive monitoring of the ecological health of garden wetlands based on hyperspectral imaging according to claim 1, characterized in that, Step S3 includes: The ecological response initiation and decay boundaries are determined based on the cumulative change rate of the climate-driven sequence. Between the ecological response initiation and decay boundaries, a continuous ecological process scale from 0 to 1 is constructed based on the cumulative progress of the climate-driven quantities. The ecological characteristics corresponding to each time in the multi-temporal observation data are mapped to the continuous ecological process scale to obtain the initial ecological process sequence.

5. The method for comprehensive monitoring of the ecological health of garden wetlands based on hyperspectral imaging according to claim 1, characterized in that, Step S4 includes: The initial ecological process sequence was divided into multiple continuous ecological segments according to the continuous ecological process scale, and the distribution range of ecological characteristics within each continuous ecological segment was statistically analyzed. Using the median value of ecological characteristics within a continuous ecological segment as a reference value, data that deviate from the reference value by more than 20% of the corresponding distribution range are locally shifted to adjacent continuous ecological segments. The initial ecological process sequence is rearranged based on the locally shifted ecological process scale, so that data in the equivalent ecological response stage in different years form an ecological alignment sequence.

6. The method for comprehensive monitoring of the ecological health of garden wetlands based on hyperspectral imaging according to claim 1, characterized in that, Step S5 includes: In the ecological alignment sequence, multiple alignment units are divided according to the ecological process scale, and the central value and distribution range of the ecological characterization features within each alignment unit are statistically analyzed. When the deviation of an ecological characteristic from the center value exceeds 15% to 25% of the corresponding distribution range, the portion exceeding the allowable deviation range will be compressed towards the center value by a ratio of 50% to 70%. By combining climate-driven sequences, the direction of change of compressed ecological characteristics is verified, resulting in a stable ecological response characteristic sequence.

7. The method for comprehensive monitoring of the ecological health of garden wetlands based on hyperspectral imaging according to claim 1, characterized in that, The change direction verification includes: Under the same ecological process scale, the direction of change of climate driving forces and the direction of change of ecological characterization features between adjacent aligned units are read; When the direction of change of two consecutive aligned units of ecological characterization features is inconsistent with the direction of change of climate driving quantities, and the change is not determined to be a true anomalous change, the corresponding ecological characterization features will be marked as phenological offset data to be corrected. Based on the direction of change of ecological characterization features of adjacent aligned units, the phenological offset data to be corrected is directionally corrected to form a stable ecological response feature sequence.

8. The method for comprehensive monitoring of the ecological health of garden wetlands based on hyperspectral imaging according to claim 1, characterized in that, Step S6 includes: Based on the stable ecological response feature sequence, the change in ecological characterization features between adjacent ecological process positions is calculated according to the ecological process scale to form a change trajectory. When the change in a continuous segment exceeds 1.5 times the overall average level for three consecutive ecological process intervals, the continuous segment is marked as a high-change segment. By comparing the direction of change of ecological characteristics in high-change sections with the climate-driven change trend, when the two are inconsistent and continue for more than two ecological process intervals, the corresponding section is identified as an anomalous change segment.

9. The method for comprehensive monitoring of the ecological health of garden wetlands based on hyperspectral imaging according to claim 1, characterized in that, Step S7 includes: The change trajectory was divided into multiple evaluation segments according to the ecological process scale, and the magnitude, direction, and proportion of abnormal change segments in each evaluation segment were statistically analyzed. Using the median value of the change range of each evaluation segment as the benchmark value, the evaluation segment with a change range that is consistently lower than 70% of the benchmark value and the proportion of abnormal change segments is less than 10% is judged as a stable state, and the evaluation segment with a change range that is higher than 130% of the benchmark value or the proportion of abnormal change segments is greater than 20% is judged as an abnormal state. When two or more consecutive evaluation segments are determined to be in an abnormal state, the corresponding spatial location is identified as being in an ecological risk state, and warning information of the corresponding level is output based on the proportion of abnormal change segments.

10. A comprehensive monitoring system for the ecological health of garden wetlands based on hyperspectral imaging, used to implement the comprehensive monitoring method for the ecological health of garden wetlands based on hyperspectral imaging as described in any one of claims 1-9, characterized in that, include: The data acquisition module acquires hyperspectral image data of the target garden wetland at multiple time phases and simultaneously acquires climate data of the corresponding time phases to establish a multi-temporal observation data set for the same spatial location. The feature extraction module performs consistency processing on hyperspectral image data to obtain spatially aligned multi-temporal data, and extracts ecological characteristics reflecting the state of vegetation, water bodies and soil. At the same time, it accumulates and transforms climate data to form climate-driven sequences. The ecological process mapping module determines the stage boundary points of ecological response based on the climate-driven sequence, and constructs a continuous ecological process scale accordingly. It maps multi-temporal observation data from discrete calendar time to the ecological process scale to obtain the initial ecological process sequence. The non-uniform rearrangement alignment module performs non-uniform rearrangement of the initial ecological process sequence according to the ecological process scale, so that data in the equivalent ecological response stage in different years are aligned to form an ecological alignment sequence. The collaborative correction module performs collaborative correction on the ecological characterization features in the ecological alignment sequence to eliminate system differences caused by phenological shifts and obtain a stable ecological response feature sequence. The anomaly identification module extracts change trajectories based on ecological response feature sequences and identifies anomalous change segments that deviate from the consistency of climate-driven processes. The early warning module generates an assessment of the ecological health status of the garden wetland based on the change trajectory and abnormal change segments, and outputs corresponding early warning information when abnormal change segments are detected.

Images