An ecological green land intelligent monitoring system based on multi-modal data

Abstract

The present application relates to the technical field of ecological green land monitoring, in particular to an ecological green land intelligent monitoring system based on multi-modal data, which comprises a multi-source monitoring alignment module, an ecological state baseline construction module, an ecological state discrimination module, a key area identification module and a monitoring strategy updating module. The present application realizes efficient cooperation of remote sensing, unmanned aerial vehicle and ground node data through multi-modal sensing means, constructs multi-dimensional ecological state representation and periodic change characteristics, combines spatial continuity discrimination and historical trend induction, forms multi-scale identification of ecological types and change levels between regions, supports intelligent screening and dynamic monitoring of the concerned area, links data acquisition, processing and task allocation processes, optimizes monitoring task configuration in real time, captures ecological change signals in time, improves the data fusion capability and spatio-temporal response level of ecological monitoring, and provides dynamic data support for fine management and maintenance.

Description

Technical Field

[0001] This invention relates to the field of ecological green space monitoring technology, and in particular to an intelligent ecological green space monitoring system based on multimodal data. Background Technology

[0002] The field of ecological green space monitoring technology involves using information technology such as remote sensing, the Internet of Things, and artificial intelligence to dynamically perceive and collect data on the ecological status of urban green spaces and natural ecological spaces, supporting intelligent construction and refined management. Traditional ecological green space intelligent monitoring systems rely on a single modal data source to periodically monitor the ecological environment of green spaces. They typically use manual fixed-point sampling or single-type sensor deployment to collect limited environmental factors, such as temperature, humidity, and light intensity, and transmit the data to a local platform for viewing or recording via wired or short-range wireless communication.

[0003] Existing ecological green space monitoring relies solely on single data types for collection. Manual fixed-point sampling and single sensor deployment result in limited spatial and temporal coverage. The simplistic data structure makes it impossible to capture the complex interrelationships of ecological elements. Communication processes are easily affected by interference, information flow is prone to interruptions, data updates are slow, environmental change signals are difficult to capture in a timely manner, local anomalies are easily missed, and it is difficult to support dynamic analysis and detailed management of large-scale ecological conditions. There is a significant delay and disconnect between monitoring results and actual ecological changes. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing an intelligent monitoring system for ecological green spaces based on multimodal data.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: an intelligent monitoring system for ecological green spaces based on multimodal data, the system comprising:

[0006] The multi-source monitoring alignment module is based on the ecological green space monitoring area. It analyzes the hyperspectral coverage of UAVs and the time period of satellite monitoring, verifies the time stamps of ground sensors, compares the arrival order and overlap of multi-source data, checks the alignment status, and obtains multi-modal monitoring alignment parameters.

[0007] Based on the multimodal monitoring alignment parameters, the ecological status baseline construction module determines the overlapping areas of data, analyzes the historical remote sensing reflectance intervals and temperature and humidity change trajectories, integrates the monitoring data of each region into a unit dataset, and obtains the basic characterization set of the ecological unit.

[0008] The ecological status discrimination module, based on the basic characterization set of the ecological unit, compares the new cycle hyperspectral and baseline distributions, analyzes the degree of deviation between temperature and humidity records and historical fluctuations, determines the spatial distribution range of monitoring data differences, and obtains the ecological change type identifier.

[0009] Based on the ecological change type identifier, the key area identification module determines the continuous multi-period change space, analyzes the intersection of the UAV trajectory and the satellite monitoring area, identifies areas with concentrated state changes, and obtains a set of key monitoring area markers.

[0010] The monitoring strategy update module adjusts the coverage area of ​​the UAV's cruise path, optimizes the satellite remote sensing downlink range, and reallocates the ground sensor acquisition order based on the key monitoring area marker set to obtain the regional monitoring task configuration.

[0011] The present invention is improved in that the multimodal monitoring alignment parameters include a time synchronization factor, a spatial mapping index, and a source data consistency marker; the ecological unit basic characterization set includes a feature vector group, a dynamic state factor, and a unit classification number; the ecological change type identifier includes a type number, a change level, and an impact range identifier; the key monitoring area marker set includes a spatial clustering marker, a continuous monitoring label, and a priority allocation number; and the regional monitoring task configuration includes a task allocation table, a data acquisition scheduling code, and a priority execution sequence.

[0012] The present invention is improved in that the multi-source monitoring alignment module includes:

[0013] The spatial zoning extraction submodule is based on the ecological green space monitoring area. It analyzes the boundary characteristics of the monitoring area, determines the different ground cover types in the raster data acquired by remote sensing images, compares the spectral reflectance characteristics and spatial texture distribution, and classifies each area into grassland area or shrub area to obtain the land use zoning data of the monitoring area.

[0014] The coverage intersection analysis submodule compares the spatial correspondence with the remote sensing images based on the land use zoning data of the monitoring area, analyzes the spatial trajectory and temporal distribution of UAV and satellite remote sensing images, determines the overlapping area of ​​each image, calculates the spatial ratio of the coverage area, and obtains the multi-source remote sensing overlapping area ratio.

[0015] The timing parameter calculation submodule, based on the multi-source remote sensing overlap area ratio and combined with the spatial index and timestamp information of the overlap area, analyzes the temporal order of sensor data from each monitoring point, filters the timestamps of the spatially overlapping area, determines the arrival order and spatial distribution of each data packet, adjusts the data sorting, and obtains the multimodal monitoring alignment parameters.

[0016] The present invention is improved in that the ecological state baseline construction module includes:

[0017] The overlapping area determination submodule determines the overlap relationship between hyperspectral remote sensing image data and temperature and humidity sensor monitoring data in geographic coordinates based on the multimodal monitoring alignment parameters, identifies spatially overlapping monitoring units, and obtains a set of spatially overlapping unit indexes.

[0018] The periodic trajectory analysis submodule, based on the spatial coincidence unit index set, compares the remote sensing reflection parameter sequences and temperature and humidity record sequences of each monitoring unit in multiple periods, judges the temporal distribution differences of the change curves, determines the periodic joint situation between parameters, and obtains the periodic change joint parameter group.

[0019] The feature induction generation submodule, based on the aforementioned periodic change joint parameter set, filters periodic correlation parameters with stable patterns within each monitoring unit, determines the indicator characteristics of different monitoring units, and obtains the basic characterization set of ecological units.

[0020] The present invention is improved in that the ecological state discrimination module includes:

[0021] The remote sensing reflectance comparison submodule analyzes the new cycle of hyperspectral remote sensing reflectance data and baseline reflectance performance based on the ecological unit basic characterization set, compares the current reflectance distribution of each band in the corresponding spatial region with the baseline reflectance distribution, judges the changing trend and distribution difference of reflectance characteristics, and obtains the remote sensing reflectance difference sequence.

[0022] The temperature and humidity deviation quantification submodule analyzes the new cycle data of temperature and humidity in the area involved based on the remote sensing reflectance difference sequence, compares the data structure of historical segments, determines the position change of the current data in the historical segment, and obtains temperature and humidity change comparison data.

[0023] The state classification and identification submodule extracts the current period's average reflection difference, temperature change segment number, and humidity change segment number based on the temperature and humidity change comparison data, calculates the multimodal joint discrimination quantity, maps the type number of each monitoring unit, and obtains the ecological change type identifier.

[0024] The present invention is improved in that the key area identification module includes:

[0025] The change area identification submodule analyzes the type and change of the spatial region corresponding to each cycle based on the ecological change type identifier, judges the changes in the type of adjacent cycles, calculates the type change frequency of each spatial region in all cycles, identifies the spatial region with concentrated type frequency, and obtains the cycle state variation region.

[0026] The monitoring area intersection analysis submodule obtains the geographical location of the spatial area based on the periodic state variation area, analyzes the spatial coverage of the UAV cruise trajectory and the satellite monitoring area, calculates the adjacency relationship between spatial areas, determines the area with concentrated spatial variation, and obtains the spatial continuous variation area.

[0027] The monitoring area marking submodule analyzes the adjacency relationships and type change frequency of spatial regions based on the continuously varying spatial regions, calculates the frequency of change of adjacent spatial region combinations, identifies spatial combinations with concentrated frequency of change, assigns continuous monitoring labels and numbers, and obtains a set of key monitoring area markings.

[0028] The present invention is improved in that the monitoring strategy update module includes:

[0029] The cruise path reconstruction submodule, based on the key monitoring area marker set, determines the boundary orientation and connection of high-frequency changing blocks, compares the UAV flight record trajectory of the previous cycle with the current spatial distribution of priority monitoring blocks, adjusts the path coverage area and node distribution, and obtains the UAV path update sequence.

[0030] The remote sensing region adjustment submodule, based on the UAV path update sequence, determines the distribution of spatial intersection and the bandwidth allocation of satellite imaging mission, compares the intersection coverage of different regions, filters the region mapping in the mission planning, and obtains the remote sensing imaging region coverage set.

[0031] The task configuration comparison submodule compares the regional changes in the distribution of monitoring tasks with the number of tasks based on the remote sensing imaging area coverage set, and adjusts the sensor order and period configuration for data acquisition to obtain the regional monitoring task configuration.

[0032] The present invention is improved in that the ecological green space monitoring area refers to the target green space area designated for ecological monitoring, which includes the ecological types of grassland area and shrub area, and the overlap refers to the common spatial coverage of UAV remote sensing, satellite remote sensing and ground monitoring data.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0034] In this invention, multimodal sensing methods are used to achieve efficient data collaboration among remote sensing, UAVs, and ground nodes, constructing multi-dimensional ecological state representations and periodic change characteristics. Combined with spatial continuity discrimination and historical trend induction, multi-scale identifiers of ecological types and change levels between regions are formed, supporting intelligent screening and dynamic monitoring of areas of interest. The data collection, processing, and task allocation processes are linked, monitoring task configuration is optimized in real time, and ecological change signals are captured in a timely manner, improving the data fusion capability and spatiotemporal response level of ecological monitoring, and providing dynamic data support for refined management and maintenance. Attached Figure Description

[0035] Figure 1 This is a system flowchart of the present invention;

[0036] Figure 2 This is a flowchart of the multi-source monitoring alignment module in this invention;

[0037] Figure 3 This is a flowchart of the ecological state baseline construction module in this invention;

[0038] Figure 4 This is a flowchart of the ecological state discrimination module in this invention;

[0039] Figure 5 This is a flowchart of the key area identification module in this invention;

[0040] Figure 6 This is a flowchart of the monitoring strategy update module in this invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0042] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and 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, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0043] All user-related information involved in this invention (including but not limited to biometric information, identity verification information, behavioral data, device information, and other data that can be used for identity verification and personalized services) is collected and processed with the user's full knowledge and voluntary consent. The use of data is limited to purposes necessary for providing the technical services of this invention, and reasonable technical and management measures will be taken to ensure the security and confidentiality of users' personal information in terms of information protection and privacy.

[0044] Example

[0045] Please see Figure 1 This invention provides a technical solution: an intelligent monitoring system for ecological green spaces based on multimodal data, comprising:

[0046] The multi-source monitoring alignment module is based on the ecological green space monitoring area, divides the grassland area and shrub area, analyzes the coverage of UAV hyperspectral remote sensing data acquisition, compares the satellite remote sensing monitoring time period, determines the time label of each monitoring point data uploaded by the ground temperature and humidity sensor, compares the arrival order and coverage area overlap of each source data, adjusts the data time sequence arrangement, and checks the multi-source data alignment status to obtain multi-modal monitoring alignment parameters.

[0047] The ecological status baseline construction module is based on multimodal monitoring alignment parameters to determine the overlapping area of ​​hyperspectral remote sensing and temperature and humidity data in spatial coordinates, analyze the remote sensing reflectance change range and temperature and humidity change trajectory of the region in historical multiple periods, integrate the monitoring data of each region into unit datasets, and summarize the characteristic changes of each period to obtain the basic characterization set of ecological units.

[0048] The ecological status discrimination module is based on the basic characterization set of ecological units. It compares the numerical distribution of the new cycle of hyperspectral remote sensing reflectance data with the baseline reflectance performance, analyzes the degree of deviation between the temperature and humidity records of this cycle and the historical fluctuation range, determines the spatial distribution range of the differences in monitoring data in each region, identifies the regional type after status discrimination, and obtains the ecological change type identifier.

[0049] The key area identification module identifies spatial areas within an ecological region that undergo continuous multi-period changes based on ecological change type identifiers. It analyzes the intersection of UAV cruise trajectories and satellite monitoring area distributions, identifies areas with continuous and concentrated state changes, and marks ecological zones that require continuous monitoring, thus obtaining a set of key monitoring area markers.

[0050] The monitoring strategy update module adjusts the coverage area of ​​the UAV cruise path in the next stage based on the key monitoring area marker set, optimizes the satellite remote sensing downlink area, reallocates the data acquisition order of the ground temperature and humidity sensors, and analyzes the changes in the updated regional task configuration compared with the previous cycle configuration to obtain the regional monitoring task configuration.

[0051] Multimodal monitoring alignment parameters include time synchronization factors, spatial mapping indexes, and source data consistency markers. The basic characterization set of ecological units includes feature vector groups, dynamic state factors, and unit classification numbers. Ecological change type identifiers include type numbers, change levels, and impact range identifiers. The key monitoring area marker set includes spatial clustering markers, continuous monitoring labels, and priority allocation numbers. Regional monitoring task configurations include task allocation tables, data collection scheduling codes, and priority execution sequences.

[0052] In the multi-source monitoring alignment module, the ecological green space monitoring area refers to the target green space area designated for ecological monitoring, including ecological types such as grassland and shrubland areas; the collection coverage area refers to the surface area that can be monitored by the UAV equipped with a hyperspectral remote sensing sensor during ground cruise scanning, i.e., the plot of land where data is actually acquired under the UAV's flight path; the data time stamp refers to the specific time of data acquisition automatically recorded by the ground temperature and humidity sensor or other monitoring equipment when collecting each set of data, used for subsequent time-series comparison and data synchronization; the data arrival order refers to the order in which data packets from different monitoring sources such as UAVs, satellites, and ground arrive at the data center or are processed. The order of nodes is used for multi-source data alignment; the overlap of coverage areas refers to the common spatial coverage of data from UAV remote sensing, satellite remote sensing, and ground monitoring, i.e., the spatial intersection of different monitoring methods acting on the same ecological unit or plot; the data temporal arrangement method refers to the processing method of sorting and aligning the collected multi-source data according to time labels and spatial partitioning rules to ensure the synchronization of multi-source data in the same area in both time and spatial dimensions; the multi-source data alignment status refers to whether different monitoring source data can achieve effective correspondence and synchronization after processing in spatial, temporal, and other dimensions, which is the basis for subsequent joint analysis.

[0053] In the ecological status baseline construction module, the overlapping area refers to the area covered by both hyperspectral remote sensing data and temperature and humidity data under the same spatial coordinates, and is the basic unit for conducting multimodal data fusion analysis; the remote sensing reflectance change range refers to the range of changes in remote sensing parameters such as reflectance at the same spatial location in hyperspectral remote sensing data over multiple historical periods, used to reflect the long-term dynamics of ecological elements such as vegetation or soil; the temperature and humidity change trajectory refers to the continuous change process of temperature and humidity data in a certain area over a historical period, and is a key parameter reflecting the changes in the microclimate and ecological environment of that area; the unit dataset refers to the collection of all monitoring data integrated for a specific ecological spatial area (such as a grassland or a shrub belt), which facilitates subsequent feature extraction and trend analysis; the feature change performance refers to the feature changes shown by remote sensing and multimodal data such as temperature and humidity in the same area over multiple monitoring periods, and is often used to identify ecological changes or anomalies.

[0054] In the ecological status discrimination module, baseline reflectance performance refers to the standard distribution of hyperspectral remote sensing reflectance parameters formed through multi-period data analysis in the previous period, which is used as a reference for ecological status comparison and discrimination; historical fluctuation range refers to the normal variation range of temperature and humidity monitoring data in a certain area within a historical period, which is an important basis for judging whether the data of the current period deviates from the normal state; deviation degree refers to the size of the difference between the current period data and the historical baseline or fluctuation range, which is used to measure the relative relationship between the current ecological status and the normal state; spatial distribution range of differences refers to the set of geographical areas that are detected to have large differences from the baseline performance in a monitoring period, which is the spatial location result of anomalies or changes; regional type after status discrimination refers to the classification result of the current status of the region after comparison and analysis, such as the classification of "normal", "stressed", "abnormal growth" and other types.

[0055] In the key area identification module, the changing spatial areas refer to ecological spatial areas whose status judgment results continuously change over multiple consecutive monitoring cycles, and are potential key targets of attention; the intersection area refers to the area where the drone's cruise trajectory and the satellite monitoring area overlap geographically, and is the core area for joint monitoring of multi-source data; the continuous and concentrated area refers to the spatial areas where the changing status is geographically continuous and forms clustered blocks, which is conducive to the formation of key monitoring belts or areas; the ecological zoning that requires continuous monitoring refers to the ecological spatial areas that have been judged and identified and marked as key areas to be focused on and tracked in the next cycle, and are the key targets for adjusting the smart monitoring strategy.

[0056] In the monitoring strategy update module, the downlink area refers to the geographic spatial range that the satellite remote sensing data is planned to transmit or focus on after the monitoring strategy is adjusted. It is the spatial configuration of the satellite monitoring plan. The area task configuration refers to the content and order of the collection tasks assigned to various monitoring devices (UAVs, ground sensors, satellites) for different areas, which is reflected in the task list of the monitoring system. The changes refer to the specific details of the changes in the spatial range, time arrangement, or collection frequency of each monitoring task in this period compared with the previous period. It is a direct manifestation of the dynamic adjustment of the monitoring strategy.

[0057] Please see Figure 2 The multi-source monitoring alignment module includes:

[0058] The spatial zoning extraction submodule is based on the ecological green space monitoring area. It analyzes the boundary characteristics of the monitoring area, determines the different ground cover types in the raster data acquired by remote sensing images, compares the spectral reflectance characteristics and spatial texture distribution, and classifies each area into grassland area or shrub area to obtain the land use zoning data of the monitoring area.

[0059] The boundary graphic information of the monitoring area is read and uniformly projected into the coordinate system of the remote sensing raster data. The remote sensing image is cropped to retain only the patches within the monitoring area. Then, the center position of each pixel is read sequentially, and the reflectance value corresponding to that position on the remote sensing image is obtained. At the same time, the texture value of its surrounding 3×3 neighborhood is extracted as auxiliary judgment information. The reflectance and texture values ​​are combined into feature terms and uniformly numbered. Then, each pixel is judged to see if it meets the grassland judgment criteria, that is, the set of pixels in the visible light green band with a reflectance greater than 0.3 and the red band with a reflectance greater than 0.45. If both conditions are met, it is regarded as a grassland area. If not, it proceeds to the next judgment, judging whether its near-infrared band reflectance is higher than 0.55, and calculating whether the uniformity of its neighborhood in the near-infrared grayscale image exceeds 0.6. If both conditions are met, it is classified as a shrub area. If neither condition is met, it is not judged for the time being. The data is categorized and recorded as other region types. For example, if a pixel in an image has a reflectance of 0.32 in the green band and 0.46 in the red band, meeting the grassland criteria, it is marked as a grassland region. Another pixel has a green band reflectance of 0.22, which does not meet the grassland criteria, so it enters the shrub category. Its near-infrared reflectance is 0.61 and its neighborhood uniformity is 0.72, also meeting the criteria, so it is marked as a shrub region. If there are other pixels with reflectances of 0.25, 0.42, and 0.51 and a neighborhood uniformity of 0.55, all below the criteria, they are marked as other types. After classification, adjacent pixels are merged. If adjacent pixels are both grassland regions, they are merged into one grassland block; if they are shrub regions, they are classified as shrub blocks. After processing the entire image in this way, land use zoning data is generated in units of regional blocks, covering the spatial distribution patterns of grassland, shrubs, and other regions.

[0060] The coverage intersection analysis submodule is based on the land use zoning data of the monitoring area, compares the spatial correspondence with the remote sensing images, analyzes the spatial trajectory and temporal distribution of UAV and satellite remote sensing images, determines the overlapping area of ​​each image, calculates the spatial proportion of the coverage area, and obtains the multi-source remote sensing overlapping area proportion.

[0061] The actual flight trajectory and data coverage boundary area of ​​the UAV remote sensing images are read separately and overlaid into the land use zoning layer. Simultaneously, the satellite imagery boundaries for the corresponding time period are imported. The three are then visually registered in a unified map coordinate system. For each grassland and shrubland patch, it is checked whether it completely or partially falls within the coverage area of ​​any UAV image. If so, it is marked as an area observable by the UAV. Next, it is determined whether it also falls within the coverage area of ​​any satellite image. If both conditions are met, the area is considered a multi-source overlapping area. Spatial analysis tools are then used to extract the boundaries of the overlapping areas, and the total area is calculated. For example, if the total area of ​​a grassland patch is 3.0 square kilometers, and the portion falling within the UAV imagery coverage area is 2.4 square kilometers... The area is 1 square kilometers, and the area falling in the satellite image is 1.5 square kilometers. The intersection of the two is 1.2 square kilometers. Therefore, the overlap ratio of this area in the UAV image is calculated as 1.2 divided by 2.4, which is 50%, and the overlap ratio in the satellite image is 1.2 divided by 1.5, which is 80%. After recording the overlap ratio of each area, the total intersection area of ​​all areas and the total coverage area of ​​each area are summarized to form a complete multi-source remote sensing overlap area ratio data table. Areas with an intersection ratio of more than 25% are used as key analysis units for subsequent analysis. Data with an intersection area that is too low (below 15%) are excluded to ensure that the subsequent analysis data has a stable spatial overlap. After completing the intersection determination, a list of area numbers and their corresponding intersection ratios is generated.

[0062] The time series parameter calculation submodule is based on the multi-source remote sensing overlap area ratio, combined with the spatial index and timestamp information of the overlapping area, analyzes the time sequence of sensor data from each monitoring point, filters the time tags of the spatially overlapping area, determines the arrival order and spatial distribution of each data packet, adjusts the data sorting, and obtains the multimodal monitoring alignment parameters.

[0063] The time record field is read from the data collected from all monitoring points. This timestamp list is arranged in ascending order to form a sensor data time series. Then, the acquisition time lists of UAV and satellite images are read separately, and formatted uniformly as a five-segment time record format of year-month-day-hour-minute. The sensor time records are iterated through. For each record, the time difference between it and the most recent UAV image and the most recent satellite image are calculated. The smaller time difference is considered the primary alignment source for that sensor record, and its type is recorded as "UAV" or "Satellite". For example, if the data time at a monitoring point is 10:00 AM on August 10th, the UAV image time is 10:25 AM on August 10th (a time difference of 25 minutes), and the satellite image time is 12:00 PM on August 10th (a time difference of 120 minutes), then the UAV image is selected as the primary alignment source. After the main alignment source is recorded, for all monitoring points, their corresponding coordinates are extracted. In the GIS, it is determined whether they intersect with the spatial intersection area in the previous step. If they intersect, they are set as valid spatial overlap points, and their land use type number, such as grassland or shrub, is recorded. Then, the complete data of all monitoring points are compared in multiple ways, including data arrival order, time tag proximity, spatial matching accuracy with the main alignment source, and record field completeness. A comprehensive priority score is set for each record. For example, if a record has a time interval of 30 minutes with the main image, is marked as a perfect match in spatial matching accuracy, and has a field completeness of 85%, then the comprehensive score can be 0.87 after being weighted proportionally. Data with higher priority scores are arranged first. The data sorting list is regenerated from high to low priority, and an alignment mark is generated for each record to form multimodal monitoring alignment parameters.

[0064] Please see Figure 3 The ecological status baseline construction module includes:

[0065] The overlapping area determination submodule, based on multimodal monitoring alignment parameters, determines the overlap relationship between hyperspectral remote sensing image data and temperature and humidity sensor monitoring data in geographic coordinates, identifies spatially overlapping monitoring units, and obtains a set of spatially overlapping unit indexes.

[0066] Read the radiometrically corrected and georegistered image data from all hyperspectral remote sensing images and project it onto a coordinate system consistent with the recording locations of the temperature and humidity sensors. Extract the center coordinates of each remote sensing image record as a spatial reference. Then, extract the location information of each temperature and humidity monitoring point and establish a unique index. Traverse all remote sensing image coordinates and monitoring point coordinates, comparing them one by one by setting a maximum spatial error threshold of 500 meters. If the Euclidean distance between the remote sensing data point and the sensor point is less than or equal to 500 meters, it is judged as an overlapping point, and the match is recorded as a valid coincidence relationship. For example, remote sensing... The center coordinates of pixel A are (104.152, 30.733), and the coordinates of temperature and humidity sensor B are (104.150, 30.731). The distance between the two points is approximately 293 meters, which meets the judgment condition. After performing coordinate matching, the matching remote sensing data number and sensor number are combined into a one-to-one mapping table, and the matching timestamp, spatial location, land use attribute and data source are recorded. All monitoring unit indexes that meet the matching conditions are summarized. At the same time, remote sensing pixels that cannot form a coincidence relationship and isolated sensor records are removed. The index set is numbered and stored for management, forming a spatial coincidence unit index set.

[0067] The periodic trajectory analysis submodule, based on the spatial overlapping unit index set, compares the remote sensing reflectance parameter sequences and temperature and humidity record sequences of each monitoring unit in multiple periods, judges the temporal distribution differences of the change curves, summarizes the periodic joint situation of the parameters, and obtains the periodic change joint parameter group.

[0068] Extract a list of index numbers for all monitoring units that overlap in different time periods. For each monitoring unit, construct a corresponding multi-period remote sensing reflectance record sequence and a temperature and humidity record sequence. Divide each monitoring period into several standard time periods, for example, setting the period to one group every 10 days. For each group, sort the remote sensing data in ascending order by time and extract its principal reflectance value. Extract the daily median values ​​for temperature and humidity data, and merge them into a periodic sequence curve. For each monitoring unit, perform difference assessment on the two data sequences, calculating the range, mean difference, and frequency difference of the two sequences in each period. For example, the remote sensing reflectance range of a certain monitoring unit in three periods is 0.21, 0.24, and 0.19, and the temperature and humidity ranges are 3.5℃, ... After recording the difference between 4.2℃ and 3.8℃, a horizontal comparison is performed according to the cycle to determine whether there is a significant temporal misalignment between the peaks and troughs of the change curves within each cycle. If the peak values ​​of the remote sensing curves are concentrated in the first 3 days of the cycle while the peak values ​​of temperature and humidity are concentrated in the 6th to 8th days of the cycle, it is marked as a significant difference in temporal distribution. If the peak difference does not exceed 2 days, it is marked as a small difference. Remote sensing and temperature and humidity records with similar temporal sequences are grouped into the same cycle trajectory. Then, the data of each cycle are jointly judged to determine whether there is a positive correlation or an inverse trend fluctuation within the same cycle. For example, if the temperature also increases when the remote sensing reflectance increases, it is a positive correlation. If the temperature decreases when the remote sensing reflectance increases, it is an inverse trend. This kind of correlation information is coded and recorded according to the cycle. The cycle correlation types of all monitoring units are summarized to obtain the cycle change joint parameter group.

[0069] The feature induction generation submodule, based on the periodic variation joint parameter group, filters periodic correlation parameters with stable patterns within each monitoring unit, judges the index characteristics of different monitoring units, and obtains the basic characterization set of ecological units.

[0070] All monitoring units were identified as having periodically correlated parameter combinations across multiple periods. The stability of the variation pattern of each parameter group across all periods was calculated. The criteria for judgment were whether the correlation trend consistency rate between parameters exceeded 70% and whether the range variation was controlled within a set threshold range. The correlation trend consistency rate threshold was set at 0.7, and the range variation threshold at 0.15. In a given monitoring unit, if remote sensing reflectance and temperature / humidity showed the same trend in 5 out of 6 periods, and the maximum range variation of each did not exceed 0.14, it was considered a stable periodic correlation, and this parameter group was selected as the stable model parameters. The system categorizes and summarizes all parameter groups that meet the conditions, marks them with the corresponding monitoring unit numbers, and then compares the frequency and value range distribution of stable parameter groups in different monitoring units. For example, if a certain group of reflectance values ​​is concentrated between 0.35 and 0.48, temperature values ​​are between 20 and 28℃, and humidity is concentrated between 55% and 65%, its value range and dominant characteristics are recorded. The system analyzes whether there are similar characteristic patterns in several monitoring units. If three or more monitoring units have this set of parameter combinations and are in similar value ranges, they are classified into the same ecological characteristic type, and their characteristic patterns are numbered and archived to obtain the basic characterization set of ecological units.

[0071] Please see Figure 4 The ecological status determination module includes:

[0072] The remote sensing reflectance comparison submodule is based on the ecological unit basic characterization set. It analyzes the new cycle of hyperspectral remote sensing reflectance data and baseline reflectance performance, compares the current reflectance distribution of each band in the corresponding spatial region with the baseline reflectance distribution, judges the changing trend and distribution difference of reflectance characteristics, and obtains the remote sensing reflectance difference sequence.

[0073] Hyperspectral remote sensing image data is acquired from the current period and mapped to monitoring units in the ecological unit basic characterization set according to their coordinate locations. A two-dimensional reflectance matrix is ​​constructed using the remote sensing reflectance values ​​of all bands within the current period, categorized by spatial unit number and band number. Simultaneously, historical reflectance baseline data for the same spatial location is retrieved from the basic characterization set. The two are compared item by item according to band number, calculating the reflectance difference between the current period and the baseline for each band and forming a difference sequence. For example, for monitoring unit number 1201, its green band reflectance is 0.38 in the current period, the baseline value is 0.33, and the difference is 0.05; the red edge band reflectance is currently 0.52, the baseline value is 0.45, and the difference is 0.07; the near-infrared band reflectance is currently 0.58, the baseline value is 0.62, and the difference is −0.04. The differences are then sorted sequentially by band number. The reflectance change sequence is formed, and then the trend of the change sequence is judged. The difference judgment range for each band is set as -0.03 to 0.03 as stable change, more than positive 0.03 as an upward trend, and less than negative 0.03 as a downward trend. The difference sequence is mapped to the change sign sequence accordingly. For example, the reflectance difference sequence of the above monitoring unit is 0.05, 0.07, and -0.04, corresponding to the trends of rising, rising, and falling. After recording the change trend, its spatial distribution characteristics are further compared. If more than 60% of the monitoring points in multiple monitoring units in a certain land area show the same trend in the same band, then the area is judged to have uniform reflectance characteristics in that band. If there is no consistent trend or the distribution ratio of trend signs is less than 30%, it is recorded as a spatial difference area, and a remote sensing reflectance difference sequence is obtained.

[0074] The temperature and humidity deviation quantification submodule analyzes the new cycle data of temperature and humidity in the area based on the remote sensing reflectance difference sequence, compares the data structure of historical segments, determines the position change of the current data in the historical segment, summarizes the combination of temperature and humidity characteristics, and obtains temperature and humidity change comparison data.

[0075] The monitoring unit numbers marked with significant change trends are extracted from the reflectance difference sequence. Temperature and humidity data records for each unit within the current cycle are read and compared side-by-side with historical temperature and humidity records at the same position. Historical data is divided into time-segment windows, each containing daily average temperature and humidity values, forming a historical temperature and humidity data structure. The current cycle data is matched with historical time segments to determine its ranking position within the historical structure. For example, if a monitoring point has a daily average temperature of 26.5℃ in the current cycle, and its historical daily average range for six cycles is 23.1℃ to 27.0℃, the current value is ranked 5th, corresponding to a high deviation level. If it were ranked first... As an extreme value shift, the current humidity is 58%, the historical range is 55% to 68%, and the current value is ranked 3rd, which is judged as a median shift. The temperature and humidity shift levels are then combined to form a temperature and humidity feature combination label. This type of shift judgment is performed on all monitoring units and their combination types are recorded, such as "high temperature-medium humidity" and "extreme high temperature-low humidity". At the same time, a shift boundary threshold is set. If the current value is more than 2% outside the historical maximum and minimum values, it is considered an extreme shift. If it is within 2% of the extreme edge within the range, it is considered an edge shift. Other values ​​are classified as neutral shifts. This refines the relationship between the current value and the historical range, generates the temperature and humidity deviation structure of each monitoring unit and the corresponding comparison combination type label, and obtains temperature and humidity change comparison data.

[0076] The status classification and identification submodule extracts the average reflection difference, temperature change segment number, and humidity change segment number based on temperature and humidity change comparison data, using the following formula:

[0077] ;

[0078] Calculate the multimodal joint discriminant, map it to the type number of each monitoring unit, and obtain the ecological change type identifier, where... This represents a multimodal joint discriminant, used to quantify the overall difference level of a spatial region under three parameters: hyperspectral reflectance, temperature variation, and humidity variation. This represents the average hyperspectral reflectance difference in the spatial region during the current period, specifically the absolute difference between the reflectance data of each band in the current period and the mean value of the baseline period. This represents the average hyperspectral reflectance data for the spatial region of the baseline period, i.e., the statistical mean of reflectance data in the same band during historical periods. This indicates the segment number of the temperature change in the current period's spatial region, corresponding to the segment number of the change in current temperature relative to the historical baseline periodic temperature within the preset change segmentation structure. This indicates the total number of segments representing the temperature change, used to normalize the segment numbering of the temperature change. This indicates the segment number of the humidity change in the current period's spatial region, corresponding to the segment number in the preset change segmentation structure for the amount of change in current humidity relative to the historical baseline periodic humidity. This indicates the total number of segments in the humidity change segment division, used to normalize the humidity change segment numbering.

[0079] Multimodal joint discriminant This refers to a single discriminant parameter obtained by jointly calculating the three types of monitoring data—hyperspectral remote sensing reflectance, temperature, and humidity changes—within each spatial region using normalization and weighting methods. This parameter reflects the overall deviation of the spatial region from its historical baseline state across multiple data dimensions; it is used to quantify the comprehensive change intensity of a spatial region within the current monitoring period; and it obtains the mean hyperspectral remote sensing reflectance difference of the corresponding spatial unit in the current period. Historical baseline reflectance mean Temperature change segment numbering Total number of temperature change segments Humidity change segment numbering Total number of humidity change segments The current period's hyperspectral reflectance data are 0.61 for the red band, 0.58 for the green band, and 0.63 for the blue band. In the historical baseline period, the reflectance values ​​for the red, green, and blue bands were 0.51, 0.52, and 0.55, respectively. The calculated differences in reflectance among the three bands are 0.1, 0.06, and 0.08, respectively. The arithmetic mean is then calculated as follows: The baseline average reflectance is The current average temperature in the periodic region is 26.3℃, the historical baseline average temperature is 24.5℃, and the temperature change is +1.8℃. The structure is segmented according to the preset temperature change (total number of segments). Each segment corresponds to a 1°C variation range, with segment numbers ranging from 1 to 5. This is mapped to the segment number. The current humidity measurement is 72%, the historical baseline average humidity is 65%, and the humidity change is +7%. The humidity is segmented according to the preset humidity change (total number of segments). Each segment corresponds to a 3% variation range, with segment numbers ranging from 1 to 5, mapped to segment numbers. Data of different dimensions were processed using proportional normalization. The normalized parameters are as follows: ; ; Substitute the above parameters into the formula to calculate the result:

[0080] ;

[0081] The preset interval division criteria are as follows: when The time marker is type A, indicating that the state is stable and has not changed significantly; when The time marker is type B, indicating that the state is in a transitional range; when The time marker is designated as type C, indicating that the state has undergone a drastic change or anomaly; this result demonstrates the effectiveness of the multimodal joint discriminant. Falling into the preset discrimination interval The state type number corresponding to this interval is B, which is used to identify the current space unit as having a moderate disturbance state in which there is a significant deviation in the three monitoring indicators of hyperspectral reflectance, temperature and humidity, but has not yet reached a drastic change.

[0082] Please see Figure 5 The key area identification module includes:

[0083] The change area identification submodule analyzes the type and change of spatial regions corresponding to each cycle based on the ecological change type identifier, judges the changes in type between adjacent cycles, calculates the frequency of type change of each spatial region in all cycles, identifies spatial regions with concentrated type frequencies, and obtains the cycle state variation region.

[0084] The ecological green space monitoring area is divided into several fixed grid units according to the existing spatial division. For example, the grassland area is divided into 12 units and the shrub area into 8 units. Then, the type number and change level records of each unit in 6 consecutive monitoring cycles are read and arranged in time-label order to form a cycle sequence. Subsequently, each unit is compared with adjacent cycles one by one. First, it is compared whether the type numbers of the previous and next cycles are consistent. If they are different, a change is recorded. Then, it is compared whether the change level crosses the preset interval. The change level is divided into four intervals: 0 to 0.2, 0.21 to 0.5, 0.51 to 0.8, and 0.81 to 1. If it crosses the interval, another change is recorded. Finally, it is compared whether the influence range identification is consistent. If they are inconsistent, another record is added. For example, a grassland unit has a type number of 1, 1, ... in 6 cycles. 2, 2, 3, 3, resulting in 2 type changes, with change levels of 0.18, 0.25, 0.45, 0.48, 0.62, and 0.70. These changes occur across the first and second cycles and the fourth and fifth cycles, resulting in 2 changes in total. The affected area changes once in the third and fourth cycles, resulting in a total of 5 changes. Dividing the number of changes for this unit by the logarithm of the cycle (5) yields a frequency of 1.0. This process is then repeated for all units, and the results are divided into intervals of 0 to 0.2, 0.21 to 0.4, 0.41 to 0.7, and 0.71 to 1. If three or more consecutive adjacent units have frequencies greater than 0.41, they are marked as the same spatial variation zone. For example, G03, G04, and G05 have frequencies of 1.0, 0.8, and 0.6 respectively and are adjacent in position, so they are collectively marked as a periodic state variation region.

[0085] The monitoring area intersection analysis submodule is based on the periodic state variation region. It obtains the geographical location of the spatial region, analyzes the spatial coverage of the UAV cruise trajectory and the satellite monitoring area, calculates the adjacency relationship between spatial regions, obtains the degree of spatial variation, determines the region with concentrated spatial variation, and obtains the spatially continuous variation region.

[0086] The center and boundary coordinates of each spatial unit are obtained. For example, the center point of a unit is at longitude 121.47° and latitude 31.23°, and the boundary consists of four vertices. Then, the UAV's cruise trajectory is connected into a path line by connecting consecutive waypoints, and a coverage zone is set on both sides of the trajectory, 10 meters apart. The satellite monitoring range is then converted into a closed area. Subsequently, it is determined whether each spatial unit falls within the UAV coverage zone. First, it is determined whether the center point is located within the zone. If not, it is determined whether the boundary overlaps with the coverage zone. If the overlap area is greater than 10 square meters, it is considered a coverage. Then, the satellite range is determined for the same unit. If the center point is within the area or the boundary intersects, it is recorded as a coverage. If both types of coverage are satisfied, it is recorded as an intersection unit. For example, if the boundary of a unit overlaps with the UAV coverage zone by 12 square meters and the center point is located within the satellite area, then this unit is considered an intersection unit. The next step involves adjacency checks between the intersecting units. By comparing whether two units share a common edge, if the ratio of the common edge length to the unit's edge length is greater than 0.15, they are considered strong adjacencies; between 0.05 and 0.15, they are considered weak adjacencies; and less than 0.05, they are not considered adjacencies. For example, if two units share a side length of 4 meters and the other side is 20 meters, the ratio is 0.2, indicating a strong adjacency. Then, combining the previously obtained frequency variation, it is determined whether the unit's own frequency and the average frequency of all its strong adjacent units are simultaneously above 0.41. If they are satisfied, it is marked as a concentrated unit. For example, if a unit's frequency is 0.9 and the average frequency of its adjacent units is 0.7, it is marked as a concentrated unit. Finally, concentrated units with continuous strong adjacency relationships are connected into a chain. If the chain length is not less than 2 units, a spatially continuous variation region is formed.

[0087] The monitoring area labeling submodule analyzes the adjacency relationships and type change frequency of spatial regions based on spatially continuous variation regions, calculates the frequency of change of adjacent spatial region combinations, identifies spatial combinations with concentrated frequency of change, assigns continuous monitoring labels and numbers, and obtains a set of key monitoring area labels.

[0088] For each spatial unit in a continuous chain, extract the periodic change records. For example, a chain contains 3 units, with change periods of the 2nd to 3rd period, the 4th to 5th period, etc. Then, combine adjacent units within the chain, such as forming two-unit combinations and three-unit combinations, and check whether each combination changes simultaneously within the same period pair. If all units within the same period pair change, record one sustained change. For example, if two units change simultaneously in the 2nd to 3rd period and the 4th to 5th period, the sustained change count is 2. Divide the sustained change count by the period pair number 5 to obtain a ratio of 0.4. Then, divide the range according to the ratio: 0 to 0.2 is normal, 0.21 to 0.4 is sustained, 0.41 to 0.6 is high, and 0.61 to 1 is high sustained. If a group If the ratio is 0.4, it is classified into the continuous interval. Then, combinations that meet the continuous interval are merged. If two combinations share a unit and the ratio difference does not exceed 0.15, they are merged into the same monitoring area. For example, combination A with a ratio of 0.4 and combination B with a ratio of 0.35 have a difference of 0.05 and are merged. Finally, priority is given according to the spatial area. Areas of 0 to 400 square meters are classified as Level 3, 401 to 900 square meters as Level 2, and areas exceeding 900 square meters as Level 1. The priority is adjusted according to the continuous ratio. Each increase of 0.1 raises the priority by one level, but not by more than one level. For example, an area of ​​800 square meters is initially classified as Level 2. If the continuous ratio of 0.4 is higher than the baseline of 0.2 by two intervals, it is raised to Level 1 and a continuous monitoring label and number are assigned to it, thus forming a set of key monitoring area markers.

[0089] Please see Figure 6 The monitoring strategy update module includes:

[0090] The cruise path reconstruction submodule determines the boundary orientation and connectivity of high-frequency changing blocks based on the key monitoring area marker set, compares the UAV flight record trajectory of the previous cycle with the current spatial distribution of priority monitoring blocks, adjusts the path coverage area and node distribution, and obtains the UAV path update sequence.

[0091] Extract the boundary polygon information of all areas designated for continuous monitoring according to the marker number, obtain the vertex coordinate sequence of each area's contour, and generate a boundary direction vector according to the coordinate direction. Then, calculate the minimum connection distance between the boundary endpoints of each area. If the connection distance is less than 100 meters, it is recorded as a connected area pair. For example, if the minimum distance between areas A and B is 85 meters, it is marked as "boundary connected" in the record. Then, read the UAV flight trajectory record of the previous cycle. This trajectory record consists of consecutive waypoint numbers, waypoint coordinates, and corresponding flight segment times. Project and superimpose this trajectory path onto the current marker set area layer in space, and determine whether it covers at least 60% of the vertices of the current area boundary region. If the coverage is insufficient, it is considered that the path has not completely passed through the target monitoring area. Then, adjust the spatial distribution density of the current priority monitoring block. Statistical analysis is performed. If there are more than two monitoring area markers within each square kilometer, it is designated as a "high-density monitoring area". The path node density in this area is set to be no less than one node per 0.3 square kilometers. Then, the node spacing distribution of the previous cycle is compared with the current number of nodes to be set. The newly set nodes are inserted at the center of the area not covered by the trajectory of the previous cycle, and the original flight segments are shifted and adjusted according to their spacing and direction. For example, if the spacing between two nodes in the previous cycle was 900 meters, the new cycle plan should shorten it to 600 meters. Then, a new node is inserted in the middle and the path polyline direction is adjusted. The reconstructed path after such adjustment must meet the following requirements: all continuous monitoring marked areas are covered by waypoints, and no flight segment is greater than 1000 meters. The new UAV flight path number, node coordinate table and path polyline set are output to obtain the UAV path update sequence.

[0092] The remote sensing region adjustment submodule determines the distribution of spatial intersection and the bandwidth allocation of satellite imaging mission based on the UAV path update sequence, compares the intersection coverage of different regions, filters the region mapping in the mission planning, and obtains the remote sensing imaging region coverage set.

[0093] Extract the coverage buffer corresponding to each updated UAV path, setting the path buffer radius to 50 meters. After generating the path coverage layer, overlay it with the existing satellite remote sensing plan layer for analysis. Extract all satellite imaging mission blocks that spatially intersect with the path buffers, and read the map sheet number, center coordinates, boundary coordinates, and imaging bandwidth allocation value for each mission block. The imaging bandwidth is allocated based on the number of images captured within the mission day, with a value range of 1 to 10. If the current mission bandwidth is 7 and the corresponding path intersection area is greater than 40% of the total area, it is considered a valid intersection; if it is less than 20%, it is marked as insufficient intersection. After calculating the intersection area of ​​all path buffers and satellite mission map sheets, generate an intersection ratio table for comparison. Areas with a ratio greater than 0.4 are marked as "strong overlap areas," those between 0.2 and 0.4 are "medium overlap areas," and those below 0.2 are not included in the imaging area selection. The map sheet numbers in the strong overlap areas are uniformly output as candidates for remote sensing imaging area mapping. The candidate map sheets are then sorted from high to low according to the task bandwidth value. Areas with a bandwidth of not less than 6 are preferentially selected for inclusion in the new round of remote sensing plan. For example, the intersection ratios of map sheet numbers X01, X03, and X07 are 0.48, 0.43, and 0.52, respectively, corresponding to bandwidths of 6, 7, and 5. Therefore, X01 and X03 are retained, and X07 is removed. The selected map sheet numbers, center coordinates, and spatial coverage boundaries are uniformly summarized to form the remote sensing imaging area coverage set for the new cycle.

[0094] The task configuration comparison submodule is based on the remote sensing imaging area coverage set. It compares the regional changes in the distribution of monitoring tasks with the number of tasks, adjusts the sensor order and periodic configuration for data acquisition, and summarizes the changes in the periodic task structure to obtain the regional monitoring task configuration.

[0095] Read all region numbers, task assignment types, sensor usage order, and data acquisition frequency from the monitoring task configuration table of the previous cycle. Then, read the region numbers within the remote sensing imaging area coverage set of the new cycle. Perform union and difference operations on the region number sets of the two cycles to extract the numbers of newly added and withdrawn regions. Compare the task configuration content of each retained region to determine if there is a change in sensor usage order. If the usage order of a certain region in the previous cycle was "ground-UAV-satellite," while in the current cycle it is "satellite-ground-UAV," then record it as an order adjustment. At the same time, extract the acquisition cycle setting for each region. If the previous cycle was once every 7 days, while the current cycle is once every 5 days, then record it as a cycle. The system summarizes and statistically analyzes all changes in area number, change type, and change content to generate a structural change record table. It then compares the total number of tasks, comparing the number of tasks in the previous period (N1) with the number of tasks in the current period (N2). If N2 > N1, it is recorded as "task increase"; otherwise, it is recorded as "task decrease." Simultaneously, it calculates the percentage change in task quantity by area type (grassland, shrubland, junction zone). For example, if the number of tasks in the grassland area increases from 40 to 52 (a 30% increase), and the number of tasks in the shrubland area decreases from 32 to 28 (a 12.5% ​​decrease), it categorizes all configuration changes by area number and outputs a sensor sequence adjustment table, a periodic frequency update table, and task increase / decrease details to obtain the area monitoring task configuration.

[0096] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A smart monitoring system for ecological green spaces based on multimodal data, characterized in that, The system includes: The multi-source monitoring alignment module is based on the ecological green space monitoring area. It analyzes the hyperspectral coverage of UAVs and the time period of satellite monitoring, verifies the time stamps of ground sensors, compares the arrival order and overlap of multi-source data, checks the alignment status, and obtains multi-modal monitoring alignment parameters. Based on the multimodal monitoring alignment parameters, the ecological status baseline construction module determines the overlapping areas of data, analyzes the historical remote sensing reflectance intervals and temperature and humidity change trajectories, integrates the monitoring data of each region into a unit dataset, and obtains the basic characterization set of the ecological unit. The ecological status discrimination module, based on the basic characterization set of the ecological unit, compares the new cycle hyperspectral and baseline distributions, analyzes the degree of deviation between temperature and humidity records and historical fluctuations, determines the spatial distribution range of monitoring data differences, and obtains the ecological change type identifier. Based on the ecological change type identifier, the key area identification module determines the continuous multi-period change space, analyzes the intersection of the UAV trajectory and the satellite monitoring area, identifies areas with concentrated state changes, and obtains a set of key monitoring area markers. The monitoring strategy update module adjusts the coverage area of ​​the UAV's cruise path, optimizes the satellite remote sensing downlink range, and reallocates the ground sensor acquisition order based on the key monitoring area marker set to obtain the regional monitoring task configuration.

2. The intelligent monitoring system for ecological green spaces based on multimodal data according to claim 1, characterized in that, The multimodal monitoring alignment parameters include time synchronization factors, spatial mapping indexes, and source data consistency markers. The ecological unit basic characterization set includes feature vector groups, dynamic state factors, and unit classification numbers. The ecological change type identifier includes type number, change level, and impact range identifier. The key monitoring area marker set includes spatial clustering markers, continuous monitoring labels, and priority allocation numbers. The regional monitoring task configuration includes a task allocation table, acquisition scheduling code, and priority execution sequence.

3. The intelligent monitoring system for ecological green spaces based on multimodal data according to claim 1, characterized in that, The multi-source monitoring alignment module includes: The spatial zoning extraction submodule is based on the ecological green space monitoring area. It analyzes the boundary characteristics of the monitoring area, determines the different ground cover types in the raster data acquired by remote sensing images, compares the spectral reflectance characteristics and spatial texture distribution, and classifies each area into grassland area or shrub area to obtain the land use zoning data of the monitoring area. The coverage intersection analysis submodule compares the spatial correspondence with the remote sensing images based on the land use zoning data of the monitoring area, analyzes the spatial trajectory and temporal distribution of UAV and satellite remote sensing images, determines the overlapping area of ​​each image, calculates the spatial ratio of the coverage area, and obtains the multi-source remote sensing overlapping area ratio. The timing parameter calculation submodule, based on the multi-source remote sensing overlap area ratio and combined with the spatial index and timestamp information of the overlap area, analyzes the temporal order of sensor data from each monitoring point, filters the timestamps of the spatially overlapping area, determines the arrival order and spatial distribution of each data packet, adjusts the data sorting, and obtains the multimodal monitoring alignment parameters.

4. The intelligent monitoring system for ecological green spaces based on multimodal data according to claim 1, characterized in that, The ecological status baseline construction module includes: The overlapping area determination submodule determines the overlap relationship between hyperspectral remote sensing image data and temperature and humidity sensor monitoring data in geographic coordinates based on the multimodal monitoring alignment parameters, identifies spatially overlapping monitoring units, and obtains a set of spatially overlapping unit indexes. The periodic trajectory analysis submodule, based on the spatial coincidence unit index set, compares the remote sensing reflection parameter sequences and temperature and humidity record sequences of each monitoring unit in multiple periods, judges the temporal distribution differences of the change curves, determines the periodic joint situation between parameters, and obtains the periodic change joint parameter group. The feature induction generation submodule, based on the aforementioned periodic change joint parameter set, filters periodic correlation parameters with stable patterns within each monitoring unit, determines the indicator characteristics of different monitoring units, and obtains the basic characterization set of ecological units.

5. The intelligent monitoring system for ecological green spaces based on multimodal data according to claim 1, characterized in that, The ecological state discrimination module includes: The remote sensing reflectance comparison submodule analyzes the new cycle of hyperspectral remote sensing reflectance data and baseline reflectance performance based on the ecological unit basic characterization set, compares the current reflectance distribution of each band in the corresponding spatial region with the baseline reflectance distribution, judges the changing trend and distribution difference of reflectance characteristics, and obtains the remote sensing reflectance difference sequence. The temperature and humidity deviation quantification submodule analyzes the new cycle data of temperature and humidity in the area involved based on the remote sensing reflectance difference sequence, compares the data structure of historical segments, determines the position change of the current data in the historical segment, and obtains temperature and humidity change comparison data. The state classification and identification submodule extracts the current period's average reflection difference, temperature change segment number, and humidity change segment number based on the temperature and humidity change comparison data, calculates the multimodal joint discrimination quantity, maps the type number of each monitoring unit, and obtains the ecological change type identifier.

6. The intelligent monitoring system for ecological green spaces based on multimodal data according to claim 1, characterized in that, The key area identification module includes: The change area identification submodule analyzes the type and change of the spatial region corresponding to each cycle based on the ecological change type identifier, judges the changes in the type of adjacent cycles, calculates the type change frequency of each spatial region in all cycles, identifies the spatial region with concentrated type frequency, and obtains the cycle state variation region. The monitoring area intersection analysis submodule obtains the geographical location of the spatial area based on the periodic state variation area, analyzes the spatial coverage of the UAV cruise trajectory and the satellite monitoring area, calculates the adjacency relationship between spatial areas, determines the area with concentrated spatial variation, and obtains the spatial continuous variation area. The monitoring area marking submodule analyzes the adjacency relationships and type change frequency of spatial regions based on the continuously varying spatial regions, calculates the frequency of change of adjacent spatial region combinations, identifies spatial combinations with concentrated frequency of change, assigns continuous monitoring labels and numbers, and obtains a set of key monitoring area markings.

7. The intelligent monitoring system for ecological green spaces based on multimodal data according to claim 1, characterized in that, The monitoring strategy update module includes: The cruise path reconstruction submodule, based on the key monitoring area marker set, determines the boundary orientation and connection of high-frequency changing blocks, compares the UAV flight record trajectory of the previous cycle with the current spatial distribution of priority monitoring blocks, adjusts the path coverage area and node distribution, and obtains the UAV path update sequence. The remote sensing region adjustment submodule, based on the UAV path update sequence, determines the distribution of spatial intersection and the bandwidth allocation of satellite imaging mission, compares the intersection coverage of different regions, filters the region mapping in the mission planning, and obtains the remote sensing imaging region coverage set. The task configuration comparison submodule compares the regional changes in the distribution of monitoring tasks with the number of tasks based on the remote sensing imaging area coverage set, and adjusts the sensor order and period configuration for data acquisition to obtain the regional monitoring task configuration.

8. The intelligent monitoring system for ecological green spaces based on multimodal data according to claim 1, characterized in that, The ecological green space monitoring area refers to the target green space area designated for ecological monitoring, including the ecological types of grassland areas and shrub areas. The overlap refers to the spatial coverage of the common area of ​​UAV remote sensing, satellite remote sensing and ground monitoring data.

Images