Environment data monitoring method and system based on cloud computing and space-time characteristics

Abstract

The application provides an environment data monitoring method and system based on cloud computing and space-time characteristics, which comprises the following steps: receiving a set of environment data transmitted by environment monitoring devices distributed at different positions in a city block, extracting space-time characteristics of the set of environment data, and generating a set of space-time correlation characteristics containing time dimension change characteristics and space dimension distribution characteristics; performing pattern recognition on the set of space-time correlation characteristics by using a pre-constructed carbon flux analysis model, and generating city block carbon flux change characteristics; and identifying position characteristics and range characteristics of a carbon flux abnormal change area according to the city block carbon flux change characteristics. The application can improve the accuracy of city block carbon flux change characteristics, and based on accurate carbon flux change characteristics, the position and range of the carbon flux abnormal change area can be more accurately identified, thereby providing a reliable basis for fine management of city block carbon flux.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and system for monitoring environmental data based on cloud computing and spatiotemporal characteristics. Background Art

[0002] With the advancement of urban development, carbon flux monitoring in urban blocks has become an important step in achieving carbon neutrality goals. By continuously collecting and analyzing data such as gas concentrations and meteorological elements in the urban environment, the flow patterns and changing trends of carbon elements in different regions can be understood, providing a scientific basis for urban low-carbon management. Currently, carbon flux monitoring in urban blocks is typically achieved by deploying decentralized monitoring equipment to collect gas concentration data, followed by independent time series analysis or spatial distribution analysis. For example, the temporal dynamics of carbon flux are reflected by continuously recording the changes in gas concentrations at specific points over time, or the spatial distribution of carbon flux is described by the concentration differences at different locations at a fixed time. The overall changes in carbon flux are then inferred based on the analysis results of a single dimension. However, existing technologies often treat the change characteristics of the temporal dimension and the distribution characteristics of the spatial dimension as independent elements, without organically linking and integrating the two. This makes it difficult to fully capture the complex changes in carbon flux in the temporal and spatial dimensions. The resulting carbon flux change characteristics lack temporal and spatial correlation, which in turn affects the accurate identification of areas with abnormal carbon flux changes and makes it difficult to meet the actual needs of refined carbon flux monitoring in urban blocks. Summary of the Invention

[0003] The present invention provides an environmental data monitoring method and system based on cloud computing and spatiotemporal characteristics.

[0004] In a first aspect, an embodiment of the present invention provides an environmental data monitoring method based on cloud computing and spatiotemporal characteristics, the method comprising: receiving an environmental data set transmitted by environmental monitoring devices distributed at different locations in an urban block, the environmental data set comprising gas concentration data and meteorological element data with a time stamp, the time stamp being used to indicate the collection moment of the corresponding environmental data; performing spatiotemporal feature extraction on the environmental data set to generate a spatiotemporal correlation feature set comprising a time dimension change feature and a space dimension distribution feature; performing pattern recognition on the spatiotemporal correlation feature set through a pre-built carbon flux analysis model to generate carbon flux change features of an urban block; and identifying the location features and range features of an area with abnormal carbon flux changes based on the carbon flux change features of the urban block.

[0005] In a second aspect, an embodiment of the present invention provides a computer system, comprising: a memory storing a computer program; and a processor for loading the computer program to implement the environmental data monitoring method based on cloud computing and spatiotemporal characteristics as described above.

[0006] The environmental data monitoring method based on cloud computing and spatiotemporal characteristics provided by the present invention receives time-stamped gas concentration data and meteorological element data transmitted by environmental monitoring devices distributed at different locations in urban blocks, generates a spatiotemporal correlation feature set including time dimension change features and spatial dimension distribution features by performing spatiotemporal feature extraction processing on the environmental data set, performs pattern recognition processing on the spatiotemporal correlation feature set through a pre-built carbon flux analysis model to generate urban block carbon flux change features, and then identifies the location features and range features of the carbon flux abnormal change area based on the urban block carbon flux change features and generates a block carbon flux monitoring report. This method obtains a spatiotemporal correlation feature set by correlating the time dimension change features and the spatial dimension distribution features, so that the carbon flux analysis model can perform pattern recognition based on features that integrate time dynamics and spatial distribution information, thereby more comprehensively capturing the changing laws of urban block carbon fluxes and improving the accuracy of urban block carbon flux change features. Based on accurate carbon flux change features, the location and range of the carbon flux abnormal change area can be more accurately identified, providing a reliable basis for the refined management of urban block carbon fluxes. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] Figure 1 This is a flow chart of an environmental data monitoring method based on cloud computing and spatiotemporal characteristics provided by an embodiment of the present invention.

[0008] Figure 2 It is a schematic diagram of the composition of a computer system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0009] See also Figure 1 , Figure 1 This is a flowchart of a method for monitoring environmental data based on cloud computing and spatiotemporal characteristics, provided by an embodiment of the present invention. The method can be executed by a computer system deployed in the cloud, such as a cloud server, and includes the following steps:

[0010] Step S100: receiving an environmental data set transmitted by environmental monitoring devices distributed in different locations in a city block, the environmental data set including gas concentration data and meteorological element data with a time stamp, where the time stamp is used to indicate the collection time of the corresponding environmental data.

[0011] Environmental monitoring devices are deployed at key locations in urban areas, such as along streets and on top of buildings, to collect data related to the environment. Gas concentration data refers to the amount of gas in the environment, such as the concentration of carbon dioxide, carbon monoxide, and sulfur dioxide. Meteorological element data covers a variety of information related to meteorological conditions, such as temperature, humidity, wind speed, and wind direction. A time stamp is a time identifier assigned to each set of collected environmental data. This identifier clarifies when the data was collected and provides a time dimension reference for subsequent data analysis.

[0012] In practice, environmental monitoring devices collect data at preset intervals, such as measuring gas concentrations and meteorological factors at set intervals. The collected data is transmitted to a data receiving center via wireless communication technologies such as Wi-Fi, Bluetooth, or ZigBee. The data receiving center integrates all received data into a time-stamped environmental data set containing various gas concentration and meteorological data.

[0013] Step S200: extracting spatiotemporal features from the environmental data set to generate a spatiotemporal correlation feature set including temporal dimension change features and spatial dimension distribution features.

[0014] Spatiotemporal feature extraction is the process of extracting temporal and spatially relevant feature information from environmental data sets. Temporal variation features reflect how environmental data evolve over time, such as fluctuations in gas concentrations throughout the day or how meteorological factors vary over time. Spatial distribution features reflect the differences in the distribution of environmental data across different spatial locations in urban areas, such as the distribution of gas concentrations across different streets and regions, and the spatial variations of meteorological factors. The spatiotemporal correlation feature set integrates features from both the temporal and spatial dimensions to create a comprehensive feature set that more comprehensively describes the spatiotemporal characteristics of environmental data.

[0015] As an implementation manner, step S200 can be specifically implemented as the following steps S210 to S250:

[0016] Step S210: performing time series division on the gas concentration data in the environmental data set, dividing the continuously collected gas concentration data into multiple time segment data groups according to a preset time interval, each time segment data group contains a gas concentration data sequence within the same time length.

[0017] Time series segmentation is the process of segmenting continuous gas concentration data into segments according to preset rules. The preset time interval is a pre-set time length based on actual needs. This time interval can be adjusted to suit different analysis objectives. For example, a shorter time period can be set to analyze short-term gas concentration changes, or a longer time period can be set to observe long-term trends. A time segment data group is a collection of data segments obtained after segmentation. The gas concentration data within each data group has the same time length, which facilitates subsequent comparison and analysis of gas concentration data from different time periods.

[0018] For example, assuming the preset time interval is a set duration, the gas concentration data continuously collected by the environmental monitoring device will be divided according to this time interval. For example, the gas concentration data for a day can be divided into multiple time segment data groups according to the preset time interval, and each data group represents a sequence of gas concentration data within a time period.

[0019] Step S220: extracting time dimension features from each time segment data group, calculating the change amplitude and change direction of the gas concentration data between adjacent time segment data groups, and generating a time dimension change feature sequence describing the gas concentration over time.

[0020] Temporal feature extraction is the process of extracting characteristic information from time-segment data sets that reflects the temporal variation of gas concentrations. The magnitude of change refers to the difference in gas concentration between adjacent time-segment data sets, reflecting the severity of the change in gas concentration within those time periods. The direction of change indicates whether the gas concentration is increasing or decreasing, determined by determining the relative magnitude of gas concentrations within adjacent time-segment data sets. The temporal variation feature sequence is a sequence of features describing the temporal variation of gas concentrations, providing a visual representation of gas concentration changes over time.

[0021] When processing time-slice data sets for urban commercial districts, the gas concentration data at corresponding moments in each adjacent time-slice data set are first compared. The magnitude of change is calculated by calculating the difference between the two values. A positive difference indicates an increase in gas concentration; a negative difference indicates a decrease in gas concentration. By recording the magnitude and direction of change for each adjacent time-slice data set, a time-dimensional change feature sequence is formed. This sequence can help analysts understand the dynamic changes in gas concentration over different time periods, such as which times of day show a more rapid increase in gas concentration and which times show a more pronounced decrease in gas concentration.

[0022] As an implementation manner, step S220 can be specifically implemented as the following steps S221 to S224:

[0023] Step S221: Perform moving window sampling on the gas concentration data in each time segment data group, and extract continuous sub-time segment data from the time segment data group according to the preset window size and window moving step, each sub-time segment data contains a preset number of continuous sampling points.

[0024] Moving window sampling extracts data by sliding a fixed-size window across a time series. The preset window size refers to the number of time points included in the sampling window, while the window step size refers to the number of time points per window shift. Sub-time segment data is a continuous data segment obtained from a time segment data set through moving window sampling. Each sub-time segment contains a preset number of consecutive sampling points. The gas concentration data corresponding to these sampling points can reflect the local changes in gas concentration during that time period.

[0025] In the processing of gas concentration data in urban commercial blocks, it is assumed that the preset window size is a specific number and the window moving step is also a specific value. Starting from the starting position of each time segment data group, a continuous preset number of gas concentration data are selected as the first sub-time segment data. Then, the window moves to the right according to the preset moving step, and a continuous preset number of gas concentration data are selected again as the second sub-time segment data, and so on, until the entire time segment data group is traversed. For example, multiple sub-time segment data are obtained from a time segment data group through continuous moving window sampling. Each sub-time segment data can be regarded as a local representative of the time segment data group. By analyzing these sub-time segment data, we can have a more detailed understanding of the changing characteristics of gas concentration in different local time periods.

[0026] Step S222: Calculate the mean and fluctuation range of the gas concentration data in each sub-time segment data, and combine the mean and fluctuation range into a basic time feature vector. The dimension of the basic time feature vector is consistent with the number of sampling points.

[0027] The mean is the average of all gas concentration data within a time segment, reflecting the average gas concentration level during that time period. The mean is calculated by adding all gas concentration data within a time segment and dividing by the number of data points. The fluctuation range is the difference between the maximum and minimum gas concentration values ​​within a time segment, reflecting the fluctuation of gas concentration within that time segment. By calculating the fluctuation range, you can understand the magnitude of gas concentration fluctuations within a specific time segment.

[0028] The basic time feature vector is a vector formed by combining the mean and the fluctuation range, and its dimension is consistent with the number of sampling points in the sub-time segment data. The purpose of doing this is to accurately represent the feature information of each sub-time segment data and to facilitate subsequent fusion and analysis with other features. For example, for each sub-time segment data obtained by moving window sampling, its mean and fluctuation range are calculated separately, and then these two features are combined into a basic time feature vector. For example, a sub-time segment data contains gas concentration data of multiple sampling points. After calculating the mean and fluctuation range of the data, they are combined according to certain rules into a basic time feature vector with the same dimension as the number of sampling points. This vector can be used as a feature representation of the sub-time segment data.

[0029] Step S223: Calculate the difference between the basic time feature vectors corresponding to adjacent sub-time segment data, extract the mean change and fluctuation range change rate of the two previous and next sub-time segment data, and generate a change rate feature vector representing the temporal dynamic characteristics.

[0030] Difference calculation involves comparing the differences between the underlying temporal feature vectors corresponding to adjacent sub-time segments. The mean change refers to the difference between the means of two consecutive sub-time segments, reflecting the variation in the average gas concentration level over adjacent local time segments. The fluctuation range change rate refers to the ratio of the fluctuation range between two consecutive sub-time segments, reflecting the degree of gas concentration fluctuation over adjacent local time segments. The change rate feature vector, a combination of the mean change and the fluctuation range change rate, represents the temporal dynamics of gas concentration between adjacent sub-time segments.

[0031] In analyzing gas concentration data in urban commercial districts, the corresponding basic time feature vectors are first obtained for each pair of adjacent sub-time segment data. The difference between the means in these two basic time feature vectors is then calculated to obtain the mean change, and the ratio of the change in the fluctuation range is calculated to obtain the fluctuation range change rate. Combining the mean change and the fluctuation range change rate generates a change rate feature vector representing the temporal dynamic characteristics. For example, by comparing the basic time feature vectors of two adjacent sub-time segment data, their mean change and fluctuation range change rate are obtained. These two features are combined into a change rate feature vector. This vector can reflect the changing dynamics of gas concentration in adjacent local time segments, providing a basis for further analysis of the temporal variation of gas concentration.

[0032] Step S224: concatenate the basic time feature vector and the change rate feature vector to obtain a time dimension change feature sequence including absolute features and relative change features, wherein each element in the time dimension change feature sequence includes corresponding time tag information.

[0033] The concatenation operation combines the basic time feature vector and the rate-of-change feature vector. The basic time feature vector contains absolute characteristics such as the mean and fluctuation range of the gas concentration within the sub-time segment data. These characteristics reflect the actual gas concentration during that time period. The rate-of-change feature vector contains relative change characteristics such as the change in the mean and the rate of change in the fluctuation range between adjacent sub-time segments, reflecting the dynamics of gas concentration changes in adjacent time periods. The time-dimensional change feature sequence is the result of concatenating these two vectors. It contains both absolute and relative characteristics of gas concentration, providing a more comprehensive description of how gas concentration changes over time.

[0034] Each element includes corresponding time stamp information to accurately associate the time-dimensional change feature sequence with actual time. For example, the basic time feature vector and rate-of-change feature vector corresponding to each sub-time segment are concatenated to produce a time-dimensional change feature sequence containing multiple elements. Each element corresponds to a time point or time period, and the time stamp information clearly indicates when the gas concentration feature represented by the element occurred.

[0035] Step S230: performing spatial grid division on the gas concentration data in the environmental data set, dividing the city blocks into multiple grid cells marked with geographic coordinates, each grid cell corresponding to a set of grid data sets containing gas concentration data and meteorological element data.

[0036] Spatial gridding is the process of dividing the geographic space of a city block into multiple small grid cells according to specific rules. A geographic coordinate tag assigns a unique geographic coordinate to each grid cell, which allows accurate location of the grid cell within the city block. A grid data set is a dataset containing gas concentration and meteorological element data for each grid cell, reflecting the environmental conditions within that grid cell.

[0037] In the application scenario of urban commercial blocks, the geographic boundary information of the urban blocks is first obtained, and then the entire block is evenly divided into multiple grid cells with geographic coordinate markers according to the preset grid size parameters. The boundary of each grid cell is defined by a longitude and latitude coordinate pair, so that its position can be accurately determined. Next, the environmental monitoring devices contained in each grid cell are identified, the gas concentration data and meteorological element data collected by these devices within a preset time period are obtained, and a mapping relationship table between grid cells and monitoring data is constructed. For example, a city commercial block is divided into multiple grid cells, and each grid cell may contain one or more environmental monitoring devices. The data collected by these devices are associated with the corresponding grid cells to form a grid data set for each grid cell, so that the environmental data at different locations can be analyzed in more detail.

[0038] As an implementation manner, step S230 may be specifically implemented as the following steps S231 to S234:

[0039] Step S231: Obtain the geographic boundary information of the city block, and evenly divide the regional space corresponding to the geographic boundary information into multiple grid cells with unique geographic coordinate markers according to preset grid size parameters. The boundary coordinates of each grid cell are defined by latitude and longitude coordinate pairs.

[0040] Geographic boundary information refers to the boundary data of the geographic scope of a city block, which can be obtained through a geographic information system (GIS) or other geographic mapping methods. The preset grid size parameters are the size of the grid cells pre-set according to actual needs, for example, they can be set to a specific length and width. Uniform division is to divide the geographic space of the city block into evenly spaced cells according to the preset grid size parameters, so that the size and shape of each grid cell are basically the same. A unique geographic coordinate marker is a unique geographic coordinate assigned to each grid cell. The location of the grid cell on the earth's surface can be accurately determined using a longitude and latitude coordinate pair. For example, the geographic boundary information of the block is first obtained through a GIS system, and then the regional space of the entire block is evenly divided according to the preset grid size parameters. For example, the geographic space of the block is divided into multiple square or rectangular grid cells according to a certain length and width, and each grid cell is assigned a unique longitude and latitude coordinate pair as its geographic coordinate marker.

[0041] Step S232: Identify the environmental monitoring device contained in each grid unit, obtain the gas concentration data and meteorological element data collected by the environmental monitoring device within a preset time period, and construct a mapping relationship table between the grid unit and the monitoring data.

[0042] Identifying the environmental monitoring devices contained in each grid unit is achieved by querying the installation location information of the environmental monitoring devices and matching them with the geographic coordinates of the grid units. The preset time period is a time range pre-set according to the analysis requirements, such as one day, one week, or one month. Constructing a mapping relationship table is the process of associating each grid unit with the gas concentration data and meteorological element data collected by the environmental monitoring devices in the unit. For example, by recording and managing the installation locations of the environmental monitoring devices, combined with the geographic coordinate information of the grid units, the environmental monitoring devices contained in each grid unit are identified. Then, the gas concentration data and meteorological element data collected by these devices within the preset time period are obtained from the data storage system. For example, there may be multiple environmental monitoring devices in a grid unit. The gas concentration data and meteorological element data collected by these devices on a certain day are associated with the grid unit to construct a mapping relationship table.

[0043] Step S233: Perform spatial interpolation on the gas concentration data within each grid cell. When there are multiple environmental monitoring devices in a grid cell, calculate the spatial impact weight of each device based on the installation height and monitoring range radius of the environmental monitoring device, and fuse the multi-source monitoring data within the same grid cell through weighted averaging.

[0044] Spatial interpolation is the process of supplementing and improving gas concentration data within a grid cell. When monitoring points within a grid cell don't fully cover the entire area, spatial interpolation can be used to estimate gas concentrations at other locations. The installation height and monitoring radius of an environmental monitoring device affect its ability to monitor the surrounding environment. A higher installation height potentially extends the monitoring range; a larger monitoring radius increases the area affected. Based on these factors, the spatial impact weight of each device is calculated. A higher weight indicates a greater impact on the gas concentration within the grid cell.

[0045] The weighted average method is to weight the data collected by multiple environmental monitoring devices in the same grid unit according to their spatial impact weights, and then divide it by the total weight to obtain the fused gas concentration data. For example, there may be multiple environmental monitoring devices installed at different locations and heights in a grid unit. For example, some devices are installed on the top of a building, and some are installed on the side of the street. The spatial impact weights of these devices are calculated based on their installation height and monitoring range radius, and then the gas concentration data collected by them are weighted averaged.

[0046] Step S234: performing spatiotemporal alignment on the fused gas concentration data and meteorological element data so that the gas concentration data and meteorological element data in the same grid cell have the same time tag sequence, and generating a grid data set containing dual indexes of time dimension and space dimension.

[0047] Spatiotemporal alignment is the process of matching and unifying the fused gas concentration data and meteorological element data in time and space. A time stamp sequence assigns a time identifier to each data point, allowing different types of data to be aligned in time. Generating a gridded dataset with dual indexes in both time and space facilitates querying and analyzing environmental data from different times and locations.

[0048] For example, a spatiotemporal alignment operation is performed on the fused gas concentration data and meteorological element data within each grid cell. First, the time stamp of the data is checked, and the gas concentration data and meteorological element data are matched according to the same time point. If there is only gas concentration data but no meteorological element data at a certain time point, or vice versa, corresponding processing is required, such as using interpolation or other methods to supplement it. Then, a dual index of the time dimension and the space dimension is added to each data point to form a grid data set containing time and space information. For example, after the gas concentration data and meteorological element data within a grid cell are spatiotemporally aligned, each data point has a corresponding time stamp and geographic coordinates. These two indexes can be used to quickly query and analyze the environmental conditions of the grid cell at different times.

[0049] Step S240: Perform spatial analysis on the gas concentration data within each grid cell, calculate the direction and magnitude of the gas concentration gradient between different grid cells, and generate a spatial dimension distribution feature matrix representing the spatial distribution difference.

[0050] Spatial analysis is the process of processing and mining gas concentration data within grid cells to discover patterns in their spatial distribution. The direction of the gas concentration gradient refers to the direction in which gas concentration changes most rapidly from low to high, reflecting the spatial trend of gas concentration. The gradient magnitude refers to the rate of change of gas concentration in that direction, reflecting the degree of variation in the spatial distribution of gas concentration. The spatial distribution feature matrix, consisting of the directions and magnitudes of gas concentration gradients between different grid cells, can intuitively demonstrate the spatial distribution differences of gas concentrations within urban blocks. For example, for the gas concentration data within each grid cell, the direction and magnitude of the gas concentration gradient between them are calculated by comparing the gas concentration values ​​of adjacent grid cells. For example, the gas concentration difference between a grid cell and its adjacent grid cells is calculated, and the gradient direction and magnitude are determined based on these differences. By organizing the gradient information between all grid cells into a matrix, we obtain the spatial distribution feature matrix. By analyzing this matrix, we can understand the spatial distribution of gas concentration within urban blocks, such as which areas have significant gas concentration fluctuations and which areas are relatively stable.

[0051] Step S250: perform spatiotemporal alignment on the time dimension change feature sequence and the space dimension distribution feature matrix to generate a spatiotemporal correlation feature set with temporal coherence and spatial correlation. Each feature element in the spatiotemporal correlation feature set contains both a time dimension tag and a spatial coordinate tag.

[0052] Spatiotemporal alignment is the process of matching and unifying the temporal variation feature sequence and the spatial distribution feature matrix in both time and space, enabling them to be analyzed and processed within the same spatiotemporal framework. Temporal coherence refers to the temporal continuity of the characteristic elements in the spatiotemporal correlation feature set, reflecting the temporal trend of gas concentration. Spatial correlation refers to the spatial correlation of characteristic elements, reflecting the relationship between gas concentrations at different locations.

[0053] A spatiotemporal correlation feature set is a comprehensive feature set obtained by fusing features from the time and space dimensions. Each feature element contains both a time dimension tag and a spatial coordinate tag, allowing accurate location of that feature element in time and space. For example, the time dimension change feature sequence and the spatial dimension distribution feature matrix are first aligned on both the time and space axes to align their temporal and spatial scales. These two features are then fused using a specific algorithm to generate a spatiotemporal correlation feature set. For example, the time dimension change feature sequence at each time point is combined with the spatial dimension distribution feature matrix at the corresponding moment, and each feature element is labeled with both the time dimension and spatial coordinates to form a complete spatiotemporal correlation feature set.

[0054] As an implementation manner, step S250 can be specifically implemented as the following steps S251 to S255:

[0055] Step S251: normalize the time axis of the time dimension change feature sequence, unify the time interval units of different time segments, and generate a standardized time feature sequence with equal time intervals.

[0056] Time axis normalization is the process of uniformly adjusting the time intervals within a time dimension change feature sequence. The time interval units for different time segments may differ. For example, some time segments collect data in hours, while others collect data in minutes. Time axis normalization can unify these intervals into the same unit, facilitating subsequent analysis and processing.

[0057] Generating a standardized time feature sequence with equal time intervals is to rearrange and organize each element in the time dimension change feature sequence according to a uniform time interval. For example, when normalizing the time axis of the time dimension change feature sequence, first determine a unified time interval unit, such as unifying all time intervals into hours. Then, interpolate or sample the time dimension change feature sequence according to this time interval so that the time intervals between each element are equal. For example, for a time dimension change feature sequence containing different time intervals, it is converted into a standardized time feature sequence with equal time intervals through an interpolation algorithm.

[0058] Step S252: normalize the spatial scale of the spatial dimension distribution feature matrix, map the spatial coordinates of different grid cells to a unified spatial reference system through coordinate transformation, and generate a standardized spatial feature matrix with a unified spatial reference.

[0059] Spatial scale normalization is the process of processing the spatial dimension distribution matrix to make it consistent across spatial scales. The spatial coordinates of different grid cells may be based on different coordinate systems or reference frames. Through coordinate transformation, they can be mapped to a unified spatial reference frame, making the spatial coordinates of all grid cells comparable.

[0060] The standardized spatial feature matrix is ​​a matrix obtained after spatial scale normalization. It has a unified spatial reference and can more accurately reflect the spatial distribution of gas concentrations. For example, the spatial coordinate information of each grid cell is obtained, and the spatial coordinates of different grid cells are transformed according to a preset unified spatial reference system through a coordinate transformation algorithm. For example, the coordinates of all grid cells are converted from the local coordinate system to the global geographic coordinate system, and then the converted coordinates are associated with the spatial dimension distribution feature matrix to generate a standardized spatial feature matrix with a unified spatial reference. This can eliminate the errors caused by different coordinate systems and improve the accuracy of the analysis of the spatial distribution of gas concentrations.

[0061] Step S253: The standardized time feature sequence and the standardized space feature matrix are fused by the spatiotemporal alignment algorithm. The feature values ​​of the missing time points are supplemented by the linear interpolation method in the time dimension, and the feature gaps of the grid cells are filled by the bilinear interpolation method in the space dimension.

[0062] The spatiotemporal alignment algorithm fuses standardized time feature sequences with standardized spatial feature matrices to unify spatiotemporal information. Linear interpolation estimates the eigenvalues ​​of missing time points in the temporal dimension, inferring the value of the missing point using the known eigenvalues ​​of adjacent time points. Bilinear interpolation fills in missing grid cell features in the spatial dimension, estimating the value of the missing location using the eigenvalues ​​of adjacent grid cells.

[0063] Exemplarily, the standardized time feature sequence and the standardized spatial feature matrix are input into the spatiotemporal alignment algorithm. In the time dimension, if the eigenvalue of a certain time point is missing, it is supplemented by linear interpolation based on the eigenvalues ​​of its adjacent time points. For example, if the eigenvalues ​​before and after a time point are known, the eigenvalue of the missing time point is calculated by linear interpolation. In the spatial dimension, if the eigenvalue of a grid unit is vacant, it is filled by bilinear interpolation based on the eigenvalues ​​of its adjacent grid units. For example, a grid unit is surrounded by four adjacent grid units, and the eigenvalue of the missing grid unit is calculated by the eigenvalues ​​of these four units.

[0064] Step S254: Perform local spatiotemporal pattern mining on the fused feature data using the time moving window technology, extract spatiotemporal association rules within continuous time segments, and generate a set of association rules representing spatiotemporal dependencies.

[0065] Local spatiotemporal patterns refer to patterns or regularities exhibited by environmental data within continuous time segments and adjacent spatial regions. Spatiotemporal association rules describe the relationships between environmental data across time and space. By mining these rules, spatiotemporal dependencies can be discovered. An association rule set, consisting of a series of spatiotemporal association rules, helps analysts understand the interactions and patterns of change in environmental data across time and space. For example, a temporal moving window technique can be used to process fused feature data. A window size and moving step are set, and the window is slid across the feature data, each time selecting a continuous time segment and adjacent spatial region as a local spatiotemporal range. The feature data within this local range is then analyzed to discover spatiotemporal association rules. For example, a correlation rule may be found between changes in gas concentration in a certain area during a certain time period and changes in meteorological elements in adjacent areas. By organizing all mined association rules into a set, a set of association rules representing spatiotemporal dependencies is obtained. Analysis of this set can be used to predict future trends in environmental data.

[0066] Step S255: combining the association rule set with the standardized spatiotemporal features to generate a spatiotemporal association feature set containing time dependency and spatial topology. Each feature element in the spatiotemporal association feature set contains dual identifications of timestamp and spatial coordinates.

[0067] Feature combination is the process of integrating association rule sets and standardized spatiotemporal features to form a more comprehensive spatiotemporal feature set. Temporal dependencies refer to the temporal interrelationships and patterns of change in environmental data, while spatial topology refers to the spatial relationships and connections between environmental data.

[0068] The spatiotemporal correlation feature set is a set obtained after feature combination. It contains both time dependency and spatial topological relationship, and can more accurately describe the spatiotemporal characteristics of environmental data. Each feature element contains dual identification of timestamp and spatial coordinates, so that the position of the element in time and space can be accurately located. Exemplarily, each rule in the association rule set is matched and combined with the standardized spatiotemporal feature. For example, an association rule is associated with the corresponding standardized spatiotemporal feature, and their feature information is merged into a new feature element. Double identification of timestamp and spatial coordinate is added to each feature element to form a complete spatiotemporal correlation feature set. This set can provide richer and more accurate information for subsequent environmental data analysis and decision-making. For example, the change in gas concentration in a certain area at a certain time in the future can be predicted based on the spatiotemporal correlation feature set.

[0069] Step S300: Perform pattern recognition on the spatiotemporal correlation feature set using a pre-built carbon flux analysis model to generate carbon flux variation characteristics of urban blocks.

[0070] The pre-built carbon flux analysis model is a trained and optimized model used to process and analyze a collection of spatiotemporal correlation features to identify patterns and regularities. Pattern recognition involves identifying regular patterns within a collection of spatiotemporal correlation features that correlate with carbon flux variations within urban blocks. Carbon flux variation characteristics represent characteristic information reflecting temporal and spatial variations in carbon flux within urban blocks. Analyzing these characteristics can help us understand the carbon emissions and trends within urban blocks.

[0071] For example, a collection of spatiotemporal correlation features is input into a pre-built carbon flux analysis model. The model processes and analyzes the input features, identifying patterns associated with carbon flux changes. For example, the model may discover patterns in the spatiotemporal correlation features of certain areas over a certain time period that correlate with increases or decreases in carbon flux. By analyzing and extracting these patterns, carbon flux change characteristics for urban blocks are generated.

[0072] As an implementation manner, step S300 can be specifically implemented as the following steps S310 to S350:

[0073] Step S310: Input the spatiotemporal correlation feature set into the feature adaptation layer of the carbon flux analysis model, and convert the time dimension change features and the spatial dimension distribution features into standard feature vectors with the same feature dimensions. The dimension of the standard feature vector is consistent with the number of neurons in the input layer of the carbon flux analysis model.

[0074] The feature adaptation layer is a layer in the carbon flux analysis model that processes and transforms the input spatiotemporal feature set to meet the model input layer's requirements. Temporal variation features and spatial distribution features have different feature dimensions and data formats. The feature adaptation layer converts them into standard feature vectors with the same feature dimensions.

[0075] The dimension of the standard feature vector is consistent with the number of neurons in the input layer of the carbon flux analysis model to ensure that the input data can be correctly input into the model for processing. For example, the spatiotemporal correlation feature set is input into the feature adaptation layer of the carbon flux analysis model. This layer analyzes and processes the time dimension change characteristics and spatial dimension distribution characteristics, for example, through feature extraction and conversion algorithms, unifying features of different dimensions into the same dimension. These features are then combined into a standard feature vector with the same dimension as the number of neurons in the model input layer. For example, the input layer of a carbon flux analysis model has multiple neurons, and the spatiotemporal correlation feature set is converted into a standard feature vector with the same dimension, which can ensure that the model can accurately process and analyze the input data.

[0076] Step S320: Dynamically assign weights to the standard feature vectors through the spatiotemporal attention mechanism layer of the carbon flux analysis model to generate an attention-weighted feature map that enhances key spatiotemporal regions. The weight value of each feature point in the attention-weighted feature map is positively correlated with the contribution of the point to the carbon flux change.

[0077] The spatiotemporal attention mechanism layer dynamically assigns weights to different feature points based on the spatiotemporal characteristics of the input data. Dynamic weight assignment adjusts the weights of feature points in real time based on their importance and contribution to carbon flux changes. Key spatiotemporal regions are those temporal and spatial regions that have a significant impact on carbon flux changes. By enhancing the characteristics of these regions, the model's ability to identify and analyze carbon flux changes can be improved.

[0078] The attention-weighted feature map is a feature map obtained after processing by the spatiotemporal attention mechanism layer, in which the weight value of each feature point is positively correlated with the contribution of that point to carbon flux changes. For example, the standard feature vector is input into the spatiotemporal attention mechanism layer of the carbon flux analysis model. This layer analyzes the standard feature vector and assigns a weight value to each feature point based on its spatiotemporal information and its potential impact on carbon flux changes. For example, for a feature point in a certain area during a certain time period, if the change at that point is closely related to the change in carbon flux, it is assigned a higher weight value. These weight values ​​are multiplied by the standard feature vector to obtain the attention-weighted feature map. This map can highlight the characteristics of key spatiotemporal regions, improving the model's sensitivity and accuracy to carbon flux changes.

[0079] Step S330: performing deep feature extraction on the attention-weighted feature map through the multi-layer perceptron module of the carbon flux analysis model, and mapping the attention-weighted feature map into a carbon flux-related feature vector in a high-dimensional feature space through multi-layer nonlinear transformation.

[0080] The multilayer perceptron module, composed of multiple neuron layers, is used to perform deep processing and feature extraction on the input attention-weighted feature map. Deep feature extraction involves extracting more abstract and representative features from the attention-weighted feature map. These features can more accurately reflect changes in carbon flux.

[0081] Multi-layer nonlinear transformation refers to the application of multiple nonlinear transformations and processing to the attention-weighted feature map through multiple layers of neurons in the multi-layer perceptron module. These transformations map the attention-weighted feature map from a low-dimensional space to a high-dimensional feature space, resulting in a carbon flux-related feature vector. This vector contains richer feature information related to carbon flux changes, providing stronger support for subsequent analysis and decision-making. For example, the attention-weighted feature map is input into the multi-layer perceptron module of the carbon flux analysis model. The first layer of neurons in this module linearly combines and nonlinearly activates the input feature map to generate a new feature representation. This feature representation is then passed to the next layer of neurons for further processing and transformation. After multiple layers of nonlinear transformations, the attention-weighted feature map is mapped to a high-dimensional feature space, forming a carbon flux-related feature vector. For example, through processing by the multi-layer perceptron module, an attention-weighted feature map containing multiple feature points is transformed into a carbon flux-related feature vector with higher dimensionality and richer feature information.

[0082] As an implementation manner, step S330 may be specifically implemented as the following steps S331 to S335:

[0083] Step S331: Input the attention-weighted feature map into the first convolutional layer of the multi-layer perceptron module, perform local feature extraction on the attention-weighted feature map through a convolution kernel of a preset size, and generate a primary feature response map containing local spatial features.

[0084] The first convolutional layer, the first layer of the multilayer perceptron module, performs a convolution operation on the input attention-weighted feature map using a convolution kernel of a preset size. The convolution kernel is a small matrix that slides across the attention-weighted feature map, multiplies it element-wise with the local region in the map, and then calculates the sum to produce a new eigenvalue. Local feature extraction involves extracting feature information from the attention-weighted feature map through convolution operations. These features can reflect the spatial characteristics of that region.

[0085] The primary feature response map is the feature map obtained after processing by the first convolutional layer. It contains the local spatial features of the attention-weighted feature map. For example, the attention-weighted feature map is input to the first convolutional layer of the multilayer perceptron module. The convolution kernel of this layer slides across the attention-weighted feature map, performing a convolution operation on each local region. For example, a convolution kernel calculates a new feature value for a small region in the attention-weighted feature map. These feature values ​​are combined to form the primary feature response map. This map can display the feature information of the local region in the attention-weighted feature map, providing a basis for subsequent feature extraction and analysis.

[0086] Step S332: Input the primary feature response map into the batch normalization processing layer, standardize the eigenvalues ​​in the primary feature response map so that the mean of each feature channel is zero and the standard deviation is one, and generate a normalized feature map with numerical stability.

[0087] The batch normalization layer, a layer within the Multilayer Perceptron (MLP) module, normalizes the input primary feature response map. Normalization transforms the eigenvalues ​​in the primary feature response map to have a mean of zero and a standard deviation of one. This is done to improve model training efficiency and numerical stability, preventing problems such as vanishing or exploding gradients. A normalized feature map, obtained after processing by the batch normalization layer, is numerically stable and more effectively conveys feature information. For example, a primary feature response map is input to the batch normalization layer. This layer performs statistical analysis on each feature channel in the primary feature response map, calculating its mean and standard deviation. It then normalizes the eigenvalues ​​based on this statistical information. For example, for all eigenvalues ​​in a feature channel, the mean is subtracted and then divided by its standard deviation to obtain a new eigenvalue. This process of processing the eigenvalues ​​of all feature channels results in a normalized feature map. This map ensures that the eigenvalues ​​are within a reasonable range, improving model training and generalization.

[0088] Step S333: Perform nonlinear transformation on the normalized feature map through the activation function layer of the multi-layer perceptron module to enhance the expression ability of key change patterns in the feature map and generate an activation feature map containing nonlinear transformation features.

[0089] The activation function layer, a layer within the multilayer perceptron module, uses an activation function to perform a nonlinear transformation on the input normalized feature map. Nonlinear transformation involves transforming the eigenvalues ​​in the normalized feature map through the activation function, imparting nonlinear characteristics. Enhancing the ability to express key patterns of change involves using the activation function to highlight key patterns and features in the normalized feature map that are closely related to carbon flux changes.

[0090] The activated feature map is a feature map obtained after being processed by the activation function layer. It contains nonlinear transformation features and can more accurately reflect the key information of carbon flux changes. For example, the normalized feature map is input into the activation function layer of the multi-layer perceptron module. This layer uses an activation function, such as the ReLU function, to transform each eigenvalue in the normalized feature map. For example, for a eigenvalue, if it is greater than zero, it remains unchanged; if it is less than zero, it is set to zero. Through this nonlinear transformation, the expression ability of the key change pattern in the feature map is enhanced, and an activated feature map is obtained. This map can better display the characteristic information of carbon flux changes and provide support for subsequent feature analysis and decision-making.

[0091] Step S334: Input the activation feature map into the pooling operation layer, perform spatial dimension reduction on the activation feature map through the maximum pooling method, retain the significant feature information in the feature map, and generate a dimensionally compressed pooled feature map.

[0092] The pooling layer performs spatial dimensionality reduction on the input activation feature map. Max pooling slides a fixed-size window over the activation feature map, selecting the maximum value within the window as the output. Spatial dimensionality reduction reduces the spatial size of the activation feature map, lowering the feature dimensionality while preserving its salient features.

[0093] A pooled feature map is a feature map obtained after processing by the pooling layer. Its spatial dimensions are smaller than those of the activation feature map, but it retains its important feature information. For example, an activation feature map is input to a pooling layer. This layer slides a window of a preset size over the activation feature map, each time selecting the maximum value within the window as the output. This method reduces the spatial dimensions of the activation feature map, reducing computational effort while preserving significant feature information, improving model efficiency and generalization.

[0094] Step S335: Input the pooled feature map into the fully connected layer network, perform high-dimensional feature mapping on the pooled feature map through the multi-layer neuron connection weight matrix, and generate an abstract feature vector with carbon flux correlation characteristics. The dimension of the abstract feature vector increases exponentially with the increase of network depth.

[0095] The fully connected layer is the final layer of the multilayer perceptron module. It consists of multiple neurons, each connected to all feature points in the input pooled feature map. The multilayer neuron connection weight matrix is ​​a parameter in the fully connected layer network that determines how each neuron weights the input features.

[0096] A high-dimensional feature map is a process that maps a pooled feature map from a low-dimensional space to a high-dimensional feature space through a fully connected network layer, resulting in an abstract feature vector with richer feature information. The dimensionality of an abstract feature vector increases exponentially with network depth. This means that as the number of network layers increases, the dimensionality of the feature vector rapidly increases, allowing it to contain more feature information. For example, a pooled feature map is input into a fully connected network. The network performs a weighted summation of each feature point in the pooled feature map based on the multi-layer neuron connection weight matrix to obtain a new eigenvalue. These eigenvalues ​​are combined to form an abstract feature vector. As the network depth increases, the dimensionality of this vector continues to increase, containing more feature information related to carbon flux. For example, through processing in a fully connected network, a low-dimensional pooled feature map is converted into a high-dimensional abstract feature vector that can more accurately reflect changes in carbon flux within a city block.

[0097] Step S340: performing a spatiotemporal consistency check on the carbon flux associated eigenvectors to ensure that the eigenvector variation trends of adjacent time segments and the eigenvector distribution patterns of adjacent spatial regions conform to preset physical constraints.

[0098] Spatiotemporal consistency checking is the process of inspecting and adjusting the carbon flux-related eigenvectors to ensure that their temporal and spatial variations conform to physical laws. The eigenvector variation trend of adjacent time segments refers to the direction and magnitude of change in the carbon flux-related eigenvectors between adjacent time points. The eigenvector distribution pattern of adjacent spatial regions refers to the distribution pattern of the carbon flux-related eigenvectors at adjacent spatial locations. Pre-set physical constraints are rules and restrictions set based on actual physical principles and experience. For example, changes in carbon flux should be continuous, and differences in carbon flux between adjacent spatial regions should be within a certain range.

[0099] Exemplarily, the carbon flux associated eigenvector is checked for spatiotemporal consistency. First, the changing trend of the eigenvectors of adjacent time segments is checked to see whether it meets the continuity requirements of the carbon flux change. For example, if the carbon flux associated eigenvector of a time segment shows a sharp increase in carbon flux, while the adjacent time segment shows a sharp decrease, and this change does not conform to the actual physical process, the eigenvector needs to be adjusted. Then, the distribution pattern of the eigenvectors of adjacent spatial regions is checked to see whether it meets the continuity and correlation requirements of the carbon flux in space. For example, if the difference between the carbon flux associated eigenvectors of adjacent spatial regions is too large and cannot be explained by physical factors, the eigenvector needs to be corrected. Through spatiotemporal consistency verification, the accuracy and reliability of the carbon flux associated eigenvector can be improved.

[0100] As an implementation manner, step S340 can be specifically implemented as the following steps S341 to S345:

[0101] Step S341: extract the carbon flux-related feature vector sequence of continuous time segments, calculate the cosine similarity of the feature vectors between adjacent time segments, and generate a feature change continuity score in the time dimension. The continuity score ranges from [0, 1].

[0102] Extracting a sequence of carbon flux-related feature vectors for consecutive time segments involves arranging the carbon flux-related feature vectors in chronological order to form a sequence. Cosine similarity is a metric that measures the similarity between two vectors. By calculating the cosine similarity between feature vectors for adjacent time segments, we can understand their directional consistency.

[0103] The feature change continuity score in the time dimension is a score generated based on the cosine similarity of the feature vectors between adjacent time segments, reflecting the continuity of the change of the carbon flux-related feature vector in the time dimension. The continuity score ranges from [0,1]. The closer the value is to 1, the more continuous the change of the feature vector in time, and the closer the value is to 0, the more discontinuous the change. For example, a vector sequence of continuous time segments is extracted from the carbon flux-related feature vector. For the feature vectors of two adjacent time segments, their cosine similarity is calculated. For example, the two feature vectors are regarded as two vectors in space, and the cosine value of the angle between them is calculated. The cosine similarities of all adjacent time segments are sorted and analyzed to generate a feature change continuity score in the time dimension. This score can help determine whether the change of the carbon flux-related feature vector in time conforms to physical laws, and provide a basis for subsequent adjustments and corrections.

[0104] Step S342: When the continuity score is lower than the preset threshold, the carbon flux-related eigenvector is temporally smoothed, and the abnormally fluctuating eigenvector value is corrected by a weighted moving average algorithm so that the trend of the corrected eigenvector conforms to the preset temporal continuity constraint.

[0105] The preset threshold is a scoring standard set in advance based on actual conditions. When the continuity score is lower than this threshold, it indicates that the carbon flux correlation eigenvector is not changing continuously over time, and may be experiencing abnormal fluctuations. Temporal smoothing is the process of processing the carbon flux correlation eigenvector to make its changes over time more continuous and stable.

[0106] The weighted moving average algorithm obtains a new eigenvector value by weighted summing the eigenvector values ​​of adjacent time points, thereby correcting the eigenvector value of abnormal fluctuations. The preset time continuity constraints are some rules set according to physical principles and experience, which require that the changes in carbon flux-related eigenvectors over time should be continuous. For example, when the characteristic change continuity score in the time dimension is lower than the preset threshold, the weighted moving average algorithm is used to perform time smoothing on the carbon flux-related eigenvector. For example, for an abnormally fluctuating eigenvector value, a weight is assigned to each adjacent value based on the eigenvector values ​​of its adjacent time points, and then these adjacent values ​​are weighted and summed to obtain a new eigenvector value. In this way, the eigenvector value of abnormal fluctuations is corrected so that the trend of change of the corrected eigenvector meets the preset time continuity constraints. This can improve the quality of the carbon flux-related eigenvector.

[0107] Step S343: performing spatial gradient calculation on the carbon flux-related characteristic vectors of adjacent spatial regions to generate a spatial gradient matrix representing the spatial characteristic change rate, where each element in the spatial gradient matrix represents the directional derivative of the characteristic vector of the adjacent region.

[0108] Spatial gradient calculation is the process of processing the carbon flux-related eigenvectors of adjacent spatial regions to calculate the rate and direction of change between them. The spatial gradient matrix is ​​a matrix composed of the spatial gradients of the carbon flux-related eigenvectors of adjacent spatial regions, which can intuitively demonstrate the spatial variation of carbon flux.

[0109] Each element represents the directional derivative of the characteristic vector of the adjacent region, which means that each element in the spatial gradient matrix reflects the rate of change of the carbon flux-related characteristic vector of the adjacent spatial region in a certain direction. For example, the carbon flux-related characteristic vectors of adjacent spatial regions are compared and calculated. For example, for the carbon flux-related characteristic vectors of two adjacent grid cells, their differences in various dimensions are calculated to obtain a vector. This vector is normalized to obtain the directional derivative of the characteristic vector of the adjacent region. The directional derivatives of all adjacent spatial regions are organized into a matrix to obtain the spatial gradient matrix. This matrix can be used to analyze the spatial change trend and rate of change of the carbon flux, providing a basis for subsequent spatial consistency verification and correction.

[0110] Step S344: Analyze the gradient direction consistency in the spatial gradient matrix. When the gradient direction angle between adjacent regions exceeds a preset angle threshold, perform spatial interpolation correction on the feature vector of the corresponding region so that the feature changes in the spatial dimension meet the preset spatial continuity constraints.

[0111] Gradient consistency refers to whether the gradient directions of adjacent regions in the spatial gradient matrix are consistent, reflecting whether the spatial variation of carbon flux is continuous and reasonable. The preset angle threshold is a pre-set angle standard based on actual conditions. When the angle between the gradient directions of adjacent regions exceeds this threshold, it indicates that the spatial variation of carbon flux is discontinuous and may be abnormal.

[0112] Spatial interpolation correction is to process the eigenvectors of the corresponding area, and estimate more reasonable eigenvector values ​​through the interpolation algorithm, so that the characteristic changes in the spatial dimension meet the preset spatial continuity constraints. The preset spatial continuity constraints are some rules set according to physical principles and experience, which require that the changes in carbon flux in space should be continuous. For example, the gradient direction in the spatial gradient matrix is ​​analyzed. For the gradient directions of adjacent areas, the angle between them is calculated. When the angle exceeds the preset angle threshold, the spatial interpolation algorithm is used to correct the eigenvector of the corresponding area. For example, based on the eigenvector values ​​of the adjacent areas, a more reasonable eigenvector value of the area is estimated through bilinear interpolation or other interpolation methods.

[0113] Step S345: Fusing the eigenvectors corrected for temporal continuity and the eigenvectors corrected for spatial continuity to generate a set of carbon flux-related eigenvectors that satisfies both temporal and spatial constraints.

[0114] Fusion is the process of combining the eigenvectors corrected for temporal continuity and those corrected for spatial continuity to form a new set of eigenvectors. A eigenvector corrected for temporal continuity is a eigenvector that, after temporal smoothing, complies with the preset continuity constraints in the temporal dimension. A eigenvector corrected for spatial continuity is a eigenvector that, after spatial interpolation, complies with the preset continuity constraints in the spatial dimension.

[0115] The set of carbon flux-related eigenvectors that simultaneously satisfy the dual constraints of time and space is a set of eigenvectors obtained after fusion. It meets the preset physical constraints in both time and space and can more accurately reflect the changes in carbon flux in urban blocks. For example, the eigenvectors corrected for time continuity and the eigenvectors corrected for space continuity are fused. For example, a simple weighted average method can be used to add the two eigenvectors according to certain weights to obtain a new eigenvector. By combining all the fused eigenvectors into a set, a set of carbon flux-related eigenvectors that simultaneously satisfy the dual constraints of time and space is obtained.

[0116] Step S350: Perform probability distribution conversion on the verified carbon flux-related feature vector through the output layer of the carbon flux analysis model to generate probability distribution curves representing different carbon flux states, and extract characteristic parameters corresponding to the peak values ​​in the probability distribution curves as the carbon flux change characteristics of urban blocks.

[0117] The output layer of the carbon flux analysis model is the final layer of the model, used to process and output the input verified carbon flux correlation feature vector. Probability distribution conversion is the process of converting the verified carbon flux correlation feature vector into a probability distribution. This conversion can be used to determine the probability of occurrence of different carbon flux states.

[0118] The probability distribution curves representing different carbon flux states are obtained after probability distribution conversion, demonstrating the probability distribution of different carbon flux states. Characteristic parameters corresponding to peaks in the probability distribution curves are extracted as characteristics of carbon flux variation in urban blocks because the carbon flux state corresponding to the peak is the most likely state, and its characteristic parameters can reflect the main characteristics of carbon flux variation in urban blocks. For example, the calibrated carbon flux-related feature vector is input into the output layer of the carbon flux analysis model. This layer processes the input feature vector and converts it into a probability distribution curve using a specific algorithm. For example, a softmax function is used to convert the feature vector into a probability distribution. The probability distribution curve is then analyzed to identify the peak and extract the characteristic parameters corresponding to the peak. These parameters can include information such as the magnitude and trend of carbon flux, serving as characteristics of carbon flux variation in urban blocks. Analysis of these characteristics can provide insights into the carbon emissions status and trends of urban blocks, providing a basis for carbon emission management and decision-making.

[0119] Step S400: Identify the location characteristics and range characteristics of the area with abnormal carbon flux changes based on the carbon flux change characteristics of the urban blocks.

[0120] Urban block carbon flux variation characteristics are derived from carbon flux analysis models and reflect the characteristics of urban block carbon flux changes, including the magnitude and trend of carbon flux. Areas of abnormal carbon flux variation are defined as areas within urban blocks where carbon flux changes significantly differ from normal conditions, indicating the presence of abnormal carbon emissions.

[0121] The location feature refers to the specific location information of the area with abnormal carbon flux changes in the urban block, and the range feature refers to the size and coverage information of the area. For example, based on the carbon flux change characteristics of the urban block, the carbon flux changes in each area are analyzed. By setting a range of normal carbon flux changes, when the carbon flux changes in a certain area exceed this range, it is identified as an area with abnormal carbon flux changes. Then, the location and range of the area are further analyzed to determine its location features and range features. For example, through geographic information system (GIS) technology, the geographical coordinates of the abnormal change area are located to determine its position on the block map; by measuring the boundary and area of ​​the area, its range features are obtained.

[0122] As an implementation manner, step S400 can be specifically implemented as the following steps S410 to S450:

[0123] Step S410: Perform eigenvalue decomposition on the carbon flux variation characteristics of urban blocks, extract the principal component components in the eigenvalues, and generate a principal component eigenvector set containing the main variation patterns.

[0124] Eigenvalue decomposition mathematically processes the characteristics of carbon flux variation in urban blocks, breaking them down into multiple eigenvalues ​​and corresponding eigenvectors. Principal components are the components with the largest contributions within the eigenvalues, representing the main patterns and trends in carbon flux variation.

[0125] The principal component eigenvector set is a set of eigenvectors corresponding to the extracted principal components, encompassing the key information about carbon flux changes in urban blocks. For example, eigenvalue decomposition is performed on the characteristics of carbon flux changes in urban blocks. Using a specific algorithm, eigenvalues ​​and corresponding eigenvectors are calculated. Then, based on the magnitude of the eigenvalues, the principal component components with the largest contributions are selected. For example, the first few principal component components with the largest eigenvalues ​​are selected. The eigenvectors corresponding to these principal component components are combined into a set, yielding a set of principal component eigenvectors encompassing the main patterns of change. Analysis of this set provides a clearer understanding of the key patterns and trends in carbon flux changes in urban blocks, providing a foundation for subsequent identification of abnormal areas.

[0126] Step S420: Input the principal component feature vector set into the abnormal pattern recognition function, and identify abnormal feature points that significantly deviate from the normal distribution range through deviation calculation. Each abnormal feature point contains a corresponding time tag and spatial coordinates.

[0127] The abnormal pattern recognition function is a predefined function used to identify abnormal patterns in a set of principal component eigenvectors. Deviation calculation uses a specific method to calculate the degree of difference between the principal component eigenvectors and the normal distribution range. When the difference exceeds a certain threshold, the point corresponding to the eigenvector is considered an abnormal feature point.

[0128] Abnormal feature points that deviate significantly from the normal distribution range refer to points in the principal component feature vector set that are significantly different from the normal situation. These points may represent abnormal changes in carbon flux. Each abnormal feature point contains a corresponding time stamp and spatial coordinates in order to accurately locate the position of the abnormal point in time and space. Exemplarily, the principal component feature vector set is input into the abnormal pattern recognition function. The function analyzes each principal component feature vector and calculates its deviation from the normal distribution range. For example, a statistical method is used to calculate the distance between the feature vector and the normal mean. When the deviation exceeds a preset threshold, the point corresponding to the feature vector is identified as an abnormal feature point. At the same time, a corresponding time stamp and spatial coordinate are added to each abnormal feature point. In this way, points with abnormal changes in carbon flux in urban blocks can be accurately identified, providing a basis for subsequent identification of abnormal areas.

[0129] Step S430: performing spatial clustering on the abnormal feature points, aggregating adjacent abnormal feature points into a plurality of abnormal feature regions according to the Euclidean distance of the spatial coordinates, each abnormal feature region containing at least three abnormal feature points.

[0130] Spatial clustering is the process of processing abnormal feature points and clustering adjacent abnormal feature points together to form regions. Euclidean distance is the straight-line distance between two spatial coordinates and can measure the spatial proximity between abnormal feature points.

[0131] An anomalous feature region is a cluster of adjacent anomalous feature points. Each region contains at least three anomalous feature points to ensure a certain scale and representativeness. For example, anomalous feature points can be spatially clustered. First, the Euclidean distance between each anomalous feature point and other anomalous feature points is calculated. A distance threshold is set; when the Euclidean distance between two anomalous feature points is less than the threshold, they are considered adjacent. Next, a clustering algorithm, such as DBSCAN (Density-Based Spatial Clustering Applications), is used to cluster adjacent anomalous feature points. Starting with an anomalous feature point, the algorithm searches for other adjacent points in its neighborhood, labeling these points as belonging to the same cluster until no more adjacent points are found, thus forming an anomalous feature region. This process continues until all anomalous feature points are assigned to their corresponding regions. Each anomalous feature region contains at least three anomalous feature points to ensure that it reflects the scale and characteristics of the anomalous carbon flux changes. For example, in an urban commercial district, if multiple adjacent anomalous feature points are concentrated in a certain area, they can be clustered into a single anomalous feature region, which helps to more accurately determine the spatial extent of the anomalous carbon flux changes.

[0132] As an implementation manner, step S430 may be specifically implemented as the following steps S431 to S435:

[0133] Step S431: Initialize the cluster center set, randomly select a preset number of initial cluster center points from the abnormal feature points, and each cluster center point has a unique spatial coordinate identifier.

[0134] Initializing the cluster center set is the starting step of the spatial clustering process, and its purpose is to provide an initial reference point for subsequent clustering operations. The preset number is the number of cluster centers set in advance according to actual needs. This number will affect the accuracy and rationality of the final clustering result. Randomly selecting initial cluster center points from abnormal feature points can ensure that the cluster centers have a certain degree of randomness and representativeness. Each cluster center point has a unique spatial coordinate identifier, so that the position of each cluster center in the urban block can be accurately located. For example, assuming that there are a large number of abnormal feature points distributed in different locations, according to the preset number of cluster centers, a corresponding number of points are randomly selected from these abnormal feature points as initial cluster center points.

[0135] Step S432: Calculate the Euclidean distance from each abnormal feature point to all cluster center points, assign each abnormal feature point to the cluster subset corresponding to the cluster center point closest to it, and obtain a preliminary cluster division result.

[0136] The Euclidean distance is calculated to measure the spatial proximity between each abnormal feature point and each cluster center point. The Euclidean distance is calculated based on the spatial coordinates of the abnormal feature point and the cluster center point, and is determined by calculating the straight-line distance between the two points. Each abnormal feature point is assigned to the cluster subset corresponding to the nearest cluster center point based on the principle of closest distance. This ensures that each abnormal feature point is reasonably classified into the most likely cluster. For example, for each abnormal feature point, its Euclidean distance to all initial cluster centers is calculated in turn. Then, these distances are compared and the abnormal feature point is assigned to the cluster subset corresponding to the nearest cluster center point. By performing this operation on all abnormal feature points, a preliminary clustering division result can be obtained, which divides the abnormal feature points into different groups.

[0137] Step S433: Calculate the geometric center coordinates of each cluster subset as the new cluster center point, and repeat the distance calculation and allocation steps until the moving distance of the cluster center point is less than the preset convergence threshold or reaches the preset number of iterations.

[0138] The geometric center coordinates of each cluster subset are calculated to update the position of the cluster center point so that the cluster center can better represent the distribution of abnormal feature points within the cluster subset. The geometric center coordinates are calculated by averaging the spatial coordinates of all abnormal feature points within the cluster subset. Repeating the distance calculation and assignment steps is to continuously optimize the clustering results so that abnormal feature points can be more accurately assigned to appropriate clusters. The preset convergence threshold is a pre-set distance value. When the moving distance of the cluster center point is less than this threshold, it means that the clustering result has stabilized and no further iteration is required. The preset number of iterations is to prevent the clustering process from falling into an infinite loop. If the preset number of iterations is reached, the iteration is stopped even if the moving distance of the cluster center point is not less than the convergence threshold. For example, the geometric center coordinates of each cluster subset are first calculated to obtain a new cluster center point. Then, the Euclidean distance of each abnormal feature point to the new cluster center point is recalculated, and the abnormal feature point is reallocated to the cluster subset with the closest distance. This process continues, constantly updating the cluster centers and redistributing abnormal feature points, until the convergence condition is met or the number of iterations is reached. For example, after multiple iterations, the movement distance of the cluster centers gradually decreases. When it is less than the preset convergence threshold, the clustering process is considered to have converged and a relatively stable clustering result has been obtained.

[0139] Step S434: Calculate the silhouette coefficient of each cluster subset. The silhouette coefficient is used to evaluate the compactness and separation of the clustering results. When the silhouette coefficient exceeds a preset threshold, it is determined to be a valid abnormal feature area.

[0140] The silhouette coefficient is a metric used to evaluate the quality of clustering results, comprehensively considering the compactness and separation of clusters. Compactness refers to the closeness between abnormal feature points within a cluster subset, while separation refers to the degree of difference between different cluster subsets. The silhouette coefficient is calculated based on the average distance of each abnormal feature point to other points in its cluster subset and the average closest distance to other cluster subsets. When the silhouette coefficient exceeds a preset threshold, it indicates that the cluster subset has good compactness and separation, and the region corresponding to the cluster subset is considered a valid abnormal feature region. For example, the silhouette coefficient is calculated for each cluster subset. For example, if the abnormal feature points in a cluster subset are close to each other and far from other cluster subsets, its silhouette coefficient will be high. The calculated silhouette coefficient is compared with the preset threshold. If it exceeds the threshold, the cluster subset is designated as a valid abnormal feature region, indicating that the abnormal feature points in this region are clearly clustered and different from other regions.

[0141] Step S435: Merge adjacent valid abnormal feature regions. When the minimum circumscribed rectangles of two regions have overlapping areas and the overlapping ratio exceeds a preset ratio threshold, merge them into a single abnormal feature region.

[0142] The purpose of merging adjacent valid abnormal feature areas is to further optimize the division of abnormal areas and avoid the appearance of overly dispersed or unreasonable small areas. The minimum enclosing rectangle is the smallest rectangle that can completely contain an abnormal feature area. By calculating the minimum enclosing rectangle, it is more convenient to determine whether two areas are adjacent and overlapping. The overlap ratio refers to the ratio of the area of ​​the overlapping part of the two minimum enclosing rectangles to the total area of ​​the two rectangles. When the overlap ratio exceeds the preset ratio threshold, it means that the correlation between the two areas is strong, and merging them into a single abnormal feature area can more accurately reflect the range of abnormal changes in carbon flux. Exemplarily, for all abnormal feature areas determined to be valid, their minimum enclosing rectangles are checked. For example, the minimum enclosing rectangles of two adjacent valid abnormal feature areas may have a certain overlap. Calculate the ratio of the area of ​​the overlapping part to the total area of ​​the two rectangles. If the ratio exceeds the preset ratio threshold, the two areas are merged into a larger abnormal feature area.

[0143] Step S440: Calculate the minimum circumscribed polygon boundary of each abnormal feature area, and extract the vertex coordinate sequence of the polygon boundary as the position feature descriptor of the abnormal change area.

[0144] The minimum bounding polygon (MBP) boundary is the smallest polygon that completely encloses the abnormal feature region, accurately defining the region's extent. The process of calculating the MBP boundary involves processing all abnormal feature points within the region to find a minimum polygon that encompasses all points. Extracting the vertex coordinate sequence of the polygon boundary involves arranging the spatial coordinates of each vertex of the MBP in a specific order to form a sequence. This sequence can serve as a positional feature descriptor for the abnormal change region, accurately locating and describing the location of the abnormal change region. For example, for each abnormal feature region obtained after clustering and merging, its minimum bounding polygon boundary is calculated. For example, if an abnormal feature region contains multiple abnormal feature points distributed at different locations, an algorithm is used to find a minimum polygon that encompasses all of these points. Then, the vertex coordinates of the polygon boundary are extracted in a clockwise or counterclockwise order to form a vertex coordinate sequence. This sequence can serve as the positional feature descriptor for the abnormal change region for subsequent analysis and presentation.

[0145] As an implementation manner, step S440 may be specifically implemented as the following steps S441 to S445:

[0146] Step S441: performing convex hull calculation on all abnormal feature points in the abnormal feature area to generate a minimum convex polygon boundary that can completely contain all abnormal feature points.

[0147] Convex hull calculation is an algorithm for finding the smallest convex polygon on a plane that can completely contain a set of points. A convex polygon is a polygon in which all internal angles are less than 180 degrees and the line between any two points is inside the polygon. By performing convex hull calculation on all abnormal feature points in the abnormal feature area, a minimum convex polygon boundary can be obtained that can most compactly surround all abnormal feature points. Exemplarily, for multiple abnormal feature points in an abnormal feature area, a convex hull calculation algorithm is used. For example, the Graham scanning algorithm or the Jarvis stepping algorithm is used to find the vertices that can form the smallest convex polygon from these abnormal feature points. The polygon formed by these vertices is the minimum convex polygon boundary that can completely contain all abnormal feature points, and can accurately define the approximate range of the abnormal feature area.

[0148] Step S442: extracting the vertex coordinate sequence of the minimum convex polygon boundary, and sorting the vertex coordinates in a clockwise direction to generate an ordered vertex coordinate sequence.

[0149] Extracting the vertex coordinate sequence of the minimum convex polygon boundary involves extracting the spatial coordinates of each vertex of the minimum convex polygon to form a sequence. Sorting the vertex coordinates in a clockwise direction ensures a uniform order in the vertex coordinate sequence, facilitating subsequent processing and comparison. Generating an ordered vertex coordinate sequence can more clearly represent the shape and boundary of the minimum convex polygon. For example, for a calculated minimum convex polygon boundary, extract the spatial coordinates of its vertices. For example, record the coordinates of each vertex in sequence. Then, sort these coordinates in a clockwise direction based on the spatial positional relationships of the vertices. This results in an ordered vertex coordinate sequence that accurately describes the boundary shape and position of the abnormally changing region.

[0150] Step S443: performing boundary smoothing on the vertex coordinate sequence, performing smooth transition on boundary segments between adjacent vertices by Bezier curve fitting, and generating a smooth boundary description including curve parameters.

[0151] Boundary smoothing is the process of processing the polygon boundary described by a vertex coordinate sequence to make it smoother and more natural. Bezier curve fitting is a commonly used curve fitting method that can generate a smooth curve using given control points. Smoothing the boundary segments between adjacent vertices is to use Bezier curves to connect adjacent vertices so that the boundary is no longer a broken line but a smooth curve. The smooth boundary description containing curve parameters is a set of curve parameters obtained by Bezier curve fitting, and these parameters can accurately describe the shape of the smooth boundary. For example, for an ordered sequence of vertex coordinates, the boundary segments between adjacent vertices are processed. For example, appropriate control points are selected and the adjacent vertices are connected using the Bezier curve fitting algorithm to form a smooth curve. The parameters of the Bezier curve are recorded, and these parameters, together with the vertex coordinates, constitute the smooth boundary description containing curve parameters.

[0152] Step S444: Calculate the geometric center coordinates of the smoothed polygon boundary, divide the polygon boundary into four quadrant areas based on the geometric center coordinates, and generate an area division result with directional characteristics.

[0153] The geometric center coordinates are the centroid position of the smoothed polygon boundary, which can be calculated by the average value of the polygon vertex coordinates. The polygon boundary is divided into four quadrants by the geometric center coordinates in order to give directional characteristics to the abnormal change area, so as to facilitate more detailed analysis and description. Generating a region division result with directional characteristics can make the description of the abnormal change area more comprehensive and accurate. For example, for the polygon boundary after boundary smoothing, calculate its geometric center coordinates. For example, add the spatial coordinates of all the vertices of the polygon and then divide it by the number of vertices to obtain the geometric center coordinates. With the geometric center as the origin, divide the polygon boundary into four quadrants, each marked with a different direction. In this way, a region division result with directional characteristics is obtained, through which the characteristics of the abnormal change area in different directions can be more clearly understood.

[0154] Step S445: Fusing the vertex coordinate sequence and the region division result to generate a position feature descriptor containing a spatial topological relationship. Each vertex coordinate in the position feature descriptor contains a quadrant identifier and boundary curve parameters.

[0155] The fusion of vertex coordinate sequence and region division result is to integrate the boundary information described by the vertex coordinate sequence with the direction information obtained by region division to form a more comprehensive description. The position feature descriptor containing spatial topological relationship can accurately reflect the position, shape and relationship with the surrounding space of the abnormal change area. Each vertex coordinate contains quadrant identification and boundary curve parameters so that the vertex coordinate has richer information and can describe the boundary characteristics of the abnormal change area in more detail. For example, the ordered vertex coordinate sequence is fused with the region division result with directional characteristics. For example, for each vertex coordinate, the quadrant identification and the corresponding boundary curve parameters are added. In this way, a position feature descriptor containing spatial topological relationship is generated, through which the position and boundary characteristics of the abnormal change area can be accurately located and described, providing detailed information for subsequent analysis and decision-making.

[0156] Step S450: measuring the geometric center coordinates of the minimum circumscribed polygon boundary and the distance set from each vertex to the center coordinates, and generating a range feature descriptor including the center coordinates and maximum distance parameters.

[0157] Measuring the geometric center coordinates of the minimum circumscribed polygon boundary is to determine the center position of the abnormal change region. The geometric center coordinates are calculated by averaging the spatial coordinates of all vertices of the minimum circumscribed polygon. The set of distances from each vertex to the center coordinates refers to the distance from each vertex of the minimum circumscribed polygon to its geometric center. By calculating these distances, the size and shape characteristics of the abnormal change region can be understood. The maximum distance parameter is the maximum value in the distance set and can reflect the maximum extension range of the abnormal change region. Generating a range feature descriptor containing the center coordinates and the maximum distance parameter can accurately describe the range size and location of the abnormal change region. For example, for each abnormal feature region's minimum circumscribed polygon boundary, its geometric center coordinates are first calculated. Then, the distance from each vertex to the center coordinate is calculated in sequence to obtain a distance set. The maximum value in the distance set is found as the maximum distance parameter. The geometric center coordinates and the maximum distance parameter are combined to generate a range feature descriptor.

[0158] As an implementation manner, the method provided in the embodiment of the present invention may further include the following steps S500 to S900:

[0159] Step S500: parsing the polygon boundary vertex coordinate sequence in the location feature, and converting the polygon boundary vertex coordinate sequence into a geographic coordinate set in the city block map coordinate system through a coordinate conversion algorithm.

[0160] Parsing the polygon boundary vertex coordinate sequence in the location feature is to process the vertex coordinate sequence contained in the location feature descriptor and extract the coordinate information of each vertex. The coordinate conversion algorithm is an algorithm that converts the original vertex coordinates into geographic coordinates in the city block map coordinate system. This is because different coordinate systems may cause differences in coordinate representation. Through coordinate conversion, the vertex coordinates can be matched with the coordinate system of the city block map. For example, assuming that the vertex coordinate sequence in the location feature descriptor is based on a certain local coordinate system, in order to accurately locate the abnormal change area on the city block map, coordinate conversion is required. For example, a projection transformation algorithm is used to convert the vertex coordinates from the local coordinate system to the geographic coordinate system used by the city block map to obtain a corresponding geographic coordinate set. This ensures that the abnormal change area can be accurately displayed on the city block map.

[0161] Step S600: Mark the spatial coverage of the abnormal change area on the digital map according to the geographic coordinate set, and generate a visual spatial distribution map including the area contour line and the center identification point.

[0162] Marking the spatial coverage of the abnormal change region on a digital map based on a set of geographic coordinates involves applying the geographic coordinates obtained through coordinate transformation to the digital map to display the actual location and extent of the abnormal change region. A regional outline is the boundary line of the abnormal change region drawn based on the geographic coordinates, clearly defining the extent of the abnormal change region. A central marker is the geometric center of the abnormal change region marked on the digital map, providing a more intuitive understanding of the central location of the abnormal change region. For example, a set of geographic coordinates can be input into digital mapping software, which will draw a regional outline of the abnormal change region based on these coordinates. Simultaneously, a central marker is marked on the digital map based on the geometric center coordinates of the abnormal change region. This creates a visual spatial distribution map containing the regional outline and central marker, allowing for a clear visualization of the specific location and extent of the abnormal carbon flux change region within a city block.

[0163] Step S700: extract the center coordinates and maximum distance parameters from the range feature, calculate the area coverage value and perimeter parameters of the abnormal change area, and generate a numerical feature set that quantitatively describes the area scale.

[0164] Extracting the center coordinates and maximum distance parameters from the range feature descriptor is key information for calculating region size. The center coordinates are used to determine the center location of the abnormal change region, while the maximum distance parameter can be used to estimate the approximate size of the abnormal change region. The area coverage value and perimeter parameters of the abnormal change region are calculated based on the center coordinates, maximum distance parameters, and the shape of the abnormal change region. The area coverage value reflects the size of the abnormal change region, while the perimeter parameter reflects the length of the abnormal change region's boundary. Generating a numerical feature set that quantitatively describes the region's size involves combining the calculated area coverage value and perimeter parameters to form a set that quantitatively describes the size of the abnormal change region. For example, the center coordinates and maximum distance parameters are extracted from the range feature descriptor. For example, based on the maximum distance parameter and the approximate circular or polygonal shape of the abnormal change region, the area coverage value and perimeter parameters are calculated using the corresponding area and perimeter calculation formulas. Combining these two parameters into a numerical feature set provides a more accurate understanding of the size of the abnormal change region.

[0165] Step S800: Fusing the visualized spatial distribution map and the numerical feature set to generate a comprehensive monitoring report framework including spatial positioning information and quantitative indicators.

[0166] Fusion of the visual spatial distribution map and the numerical feature set integrates visual spatial distribution information and quantitative regional scale information, so that the monitoring report has both intuitive spatial presentation and accurate quantitative data. The comprehensive monitoring report framework, which includes spatial positioning information and quantitative indicators, is a report template that includes the spatial location, range, and quantitative scale data of the abnormal change area, which can provide a basis for the subsequent generation of a complete monitoring report. For example, the generated visual spatial distribution map and the numerical feature set are fused. For example, the area coverage value and perimeter parameters in the numerical feature set are added to the corresponding positions of the visual spatial distribution map, or the visual spatial distribution map is included as part of the report and laid out according to a certain layout together with the numerical feature set. This generates a comprehensive monitoring report framework that can more comprehensively display information on the areas of abnormal carbon flux changes in urban blocks, making it easier for relevant personnel to view and analyze.

[0167] As an implementation manner, step S800 can be specifically implemented as the following steps S810 to S850:

[0168] Step S810: Overlay the visualized spatial distribution map with the city block basic map data to ensure accurate correspondence between spatial positions and generate a spatial distribution layer with geographic reference.

[0169] Overlaying the visualized spatial distribution map with the city block base map data integrates the visualization of the abnormal change areas with the basic geographic information of the city blocks. The city block base map data contains basic geographic information such as roads, buildings, and terrain. This overlay allows the abnormal change areas to be accurately placed at the actual geographic locations of the city blocks. Accurate spatial location alignment ensures that the abnormal change areas in the visualized spatial distribution map match the actual locations on the city block base map, providing a more intuitive understanding of the relationship between the abnormal change areas and the surrounding environment. Generating a georeferenced spatial distribution layer combines the overlay results into a new layer that not only contains the visualization information of the abnormal change areas but also has an accurate georeference. For example, geographic information system (GIS) software is used to overlay the visualized spatial distribution map with the city block base map data. For example, the visualized spatial distribution map is added as a new layer to the city block base map. By adjusting the layer's position and transparency, the locations of the abnormal change areas are precisely aligned with the actual locations on the base map. This generates a georeferenced spatial distribution layer that allows for a more accurate visualization of the abnormal carbon flux change areas within the geographic context of the city blocks.

[0170] Step S820: normalize the area coverage values ​​and perimeter parameters in the numerical feature set, convert them into standardized scores, and generate relative index values ​​that are convenient for cross-region comparison.

[0171] Magnitude normalization processes the area coverage values ​​and perimeter parameters in a numerical feature set to bring them to a uniform magnitude and range. Since the areas and perimeters of different abnormal change regions can vary significantly, magnitude normalization can eliminate this discrepancy and make the data more comparable. Converting to a standardized score maps the normalized data to a specific score range, such as [0, 1] or [0, 100]. This allows for more intuitive comparison of indicator values ​​across regions. Generating relative indicator values ​​that facilitate cross-regional comparison involves converting the area coverage values ​​and perimeter parameters into relative indicators that can be compared across different abnormal change regions. For example, for the area coverage values ​​and perimeter parameters in a numerical feature set, first calculate their maximum and minimum values. Then, use a magnitude normalization algorithm, such as linear normalization, to map each parameter value to a uniform range. For example, both the area coverage value and perimeter parameter can be normalized to the range [0, 1]. The normalized values ​​are then converted to standardized scores, for example, by multiplying them by 100 to obtain a score within the range [0, 100]. This generates relative index values ​​that are easy to compare across regions, and these index values ​​can be used to more easily compare the sizes of different abnormal change areas.

[0172] Step S830: Using a report template generator, the spatial distribution layers and relative index values ​​are laid out according to preset layout rules to generate a report main body framework including a graphic and text combined area.

[0173] The report template generator is a pre-designed tool for generating report templates based on input information. Laying out the spatial distribution layer and relative index values ​​according to preset layout rules means arranging the visual spatial distribution information and the quantitative relative index values ​​in a certain format and layout, making the report readable and professional. The report main body framework containing the combined text and graphics area is a report template containing the visual spatial distribution layer and relative index values, which can be used as the main part of the subsequent generation of a complete report. For example, using the report template generator, the spatial distribution layer with a geographic reference and the relative index values ​​that facilitate cross-regional comparison are input into the template generator. According to the preset layout rules, such as placing the spatial distribution layer at the top of the report and the relative index values ​​at the bottom in the form of a table or chart, etc., the report is laid out. In this way, a report main body framework containing the combined text and graphics area is generated, through which the spatial distribution and quantitative index information of the abnormal carbon flux change areas in the urban blocks can be clearly displayed.

[0174] Step S840: embed time tag information in the report body frame, associate monitoring data of different time segments through the time axis control, and realize the dynamic display function of the time dimension of the monitoring results.

[0175] Embedding time stamp information in the report's main frame combines monitoring data for abnormal change areas with corresponding time information, enabling the report to include not only spatial and quantitative information but also temporal information. The timeline control is a tool for displaying time series data, allowing for convenient switching between monitoring data from different time segments. Linking monitoring data from different time segments involves connecting monitoring data from abnormal change areas collected at different time points to the timeline control for dynamic display. Implementing dynamic temporal display of monitoring results allows users to intuitively understand the development and changes of abnormal change areas over time. For example, by adding the time stamp information corresponding to the monitoring data for each abnormal change area to the report's main frame, a timeline control can be set up in the report to associate visual spatial distribution maps and relative indicator values ​​at different time points with the corresponding time points on the timeline. When users operate the timeline control, they can switch between monitoring data from different time segments, such as the range, area, and perimeter of the abnormal change area at different times. This enables dynamic temporal display of monitoring results, providing richer information for analyzing trends in abnormal carbon flux changes.

[0176] Step S851: Add an abnormal change trend analysis module, use a curve fitting algorithm to perform trend prediction on the carbon flux change characteristics at consecutive time points, and generate an extended report module containing prediction curves for future time periods.

[0177] Adding an Abnormal Change Trend Analysis module adds a dedicated module to the comprehensive monitoring report framework for analyzing abnormal change trends. A curve fitting algorithm is used to fit a curve based on known data points to predict future data trends. Trend prediction for carbon flux variation characteristics at consecutive time points utilizes a curve fitting algorithm to predict carbon flux changes for future time periods based on historical carbon flux variation characteristics. The extended report module, which generates a forecast curve for future time periods, displays the trend prediction results in the form of a curve in the report, forming an extended report section that provides users with information on future trends in abnormal carbon flux changes. For example, data on carbon flux variation characteristics at consecutive time points is collected, such as carbon flux values ​​at different times and the area of ​​abnormal change regions. An appropriate curve fitting algorithm, such as linear regression, polynomial regression, or exponential regression, is then selected to fit this data. For example, a polynomial regression algorithm can be used to fit a curve that reflects the carbon flux trend. Based on the fitted curve, carbon flux changes for future time periods are predicted and displayed as a curve in a new report module.

[0178] Step S900: embed time stamp information and carbon flux change feature description in the comprehensive monitoring report framework to generate a block carbon flux monitoring report containing spatiotemporal distribution characteristics.

[0179] Embedding time stamp information within the comprehensive monitoring report framework adds the time information corresponding to each piece of monitoring data to the report, allowing the report to reflect the temporal dimension of the monitoring data. The carbon flux variation characteristic description provides a detailed textual description of the changes in carbon flux over time and space, including trends and anomalies. Generating a block carbon flux monitoring report that includes spatiotemporal distribution characteristics integrates time stamp information, carbon flux variation characteristic descriptions, and the comprehensive monitoring report framework to form a complete report that comprehensively demonstrates the spatiotemporal distribution characteristics of carbon flux in urban blocks.

[0180] It is understandable that the various algorithms involved in the above-mentioned introductions of the embodiments of the present invention, such as the Euclidean distance algorithm, the cosine distance algorithm, the spatiotemporal alignment algorithm, etc., can all be learned from the relevant content in the prior art. In order to save space, they will not be expanded too much in the embodiments of the present invention. In addition, when implementing the scheme of the present invention, those skilled in the art can supplement the details according to the common knowledge in this field. For example, according to the common knowledge in this field, normalization can be used to eliminate dimensional conflicts before feature fusion, interpolation can be used to eliminate dimensional differences, and thresholds can be reasonably set based on historical data, experience or business scenario requirements. The model can be trained based on a general model training method, and the number of layers in the model structure can be set based on actual needs, the activation function can be selected, etc. The present invention will no longer provide redundant introductions to the overly detailed implementation process.

[0181] See also Figure 2 , Figure 2 This is a schematic diagram of the structure of a computer system provided in an embodiment of the present invention. The computer system includes at least a processor 101, a communication interface 102, and a memory 103. The processor 101, communication interface 102, and memory 103 may be connected via a bus or other means. The processor 101 (also known as the Central Processing Unit (CPU)) is the computing and control core of the computer system, capable of parsing various instructions within the computer system and processing various data within the computer system. The communication interface 102 may optionally include a standard wired interface or a wireless interface (such as Wi-Fi, a mobile communication interface, etc.), which can be used to send and receive data under the control of the processor 101. The communication interface 102 may also be used for data transmission and interaction within the computer system. The memory 103 is a storage device in the computer system for storing programs and data. It is understood that the memory 103 herein may include both the built-in memory of the computer system and, of course, the extended memory supported by the computer system. The memory 103 provides storage space, which stores the computer system's operating system, but this is not limited to this in the present invention.

[0182] In one embodiment, the processor 101 executes the environmental data monitoring method based on cloud computing and spatiotemporal characteristics provided in the above embodiment of the present invention by running the computer program in the memory 103 .

Claims

1. A method for monitoring environmental data based on cloud computing and spatiotemporal characteristics, characterized in that: The method comprises: Receiving an environmental data set transmitted by environmental monitoring devices distributed at different locations in a city block, the environmental data set including gas concentration data and meteorological element data with a time stamp, the time stamp being used to indicate a time when the corresponding environmental data was collected; Extracting spatiotemporal features from the environmental data set to generate a spatiotemporal correlation feature set including temporal dimension change features and spatial dimension distribution features; The pre-built carbon flux analysis model is used to perform pattern recognition on the spatiotemporal correlation feature set to generate carbon flux change characteristics of urban blocks; specifically, the method includes: inputting the spatiotemporal correlation feature set into the feature adaptation layer of the carbon flux analysis model, converting the time dimension change characteristics and the space dimension distribution characteristics into standard feature vectors with the same feature dimensions, and the dimension of the standard feature vector is consistent with the number of neurons in the input layer of the carbon flux analysis model; dynamically assigning weights to the standard feature vectors through the spatiotemporal attention mechanism layer of the carbon flux analysis model to generate an attention weighted feature map that strengthens key spatiotemporal areas, and the weight value of each feature point in the attention weighted feature map is proportional to the contribution of the point to the carbon flux change. positive correlation; performing deep feature extraction on the attention-weighted feature map through the multi-layer perceptron module of the carbon flux analysis model, and mapping the attention-weighted feature map into a carbon flux-related feature vector in a high-dimensional feature space through multi-layer nonlinear transformation; performing spatiotemporal consistency check on the carbon flux-related feature vector so that the change trend of the feature vectors of adjacent time segments and the distribution law of the feature vectors of adjacent spatial regions meet the preset physical constraint conditions; performing probability distribution conversion on the checked carbon flux-related feature vector through the output layer of the carbon flux analysis model, generating a probability distribution curve representing different carbon flux states, and extracting the characteristic parameters corresponding to the peak values ​​in the probability distribution curve as the carbon flux change characteristics of the urban block; The location characteristics and range characteristics of the area with abnormal carbon flux changes are identified based on the carbon flux change characteristics of the urban blocks.

2. The method according to claim 1, wherein The step of extracting spatiotemporal features from the environmental data set to generate a spatiotemporal correlation feature set including time dimension change features and space dimension distribution features includes: Performing time series division on the gas concentration data in the environmental data set, dividing the continuously collected gas concentration data into a plurality of time segment data groups according to a preset time interval, each time segment data group containing a gas concentration data sequence within the same time length; Extract time dimension features from each time segment data group, calculate the change amplitude and change direction of gas concentration data between adjacent time segment data groups, and generate a time dimension change feature sequence describing the gas concentration over time; Performing spatial grid division on the gas concentration data in the environmental data set, dividing the city block into a plurality of grid cells marked with geographic coordinates, each grid cell corresponding to a set of grid data sets containing gas concentration data and meteorological element data; Perform spatial analysis on the gas concentration data within each grid cell, calculate the direction and magnitude of the gas concentration gradient between different grid cells, and generate a spatial dimension distribution feature matrix representing the spatial distribution differences; The time dimension change feature sequence and the space dimension distribution feature matrix are aligned in time and space to generate a time and space correlation feature set with time coherence and space correlation, and each feature element in the time and space correlation feature set includes both a time dimension mark and a space coordinate mark.

3. The method according to claim 2, wherein The method of extracting time dimension features from each time segment data group, calculating the change amplitude and change direction of gas concentration data between adjacent time segment data groups, and generating a time dimension change feature sequence describing the gas concentration over time includes: Performing moving window sampling on the gas concentration data in each time segment data group, extracting continuous sub-time segment data from the time segment data group according to a preset window size and window moving step, each sub-time segment data including a preset number of continuous sampling points; Calculate the mean and fluctuation range of the gas concentration data in each sub-time segment data, and combine the mean and the fluctuation range into a basic time feature vector, where the dimension of the basic time feature vector is consistent with the number of sampling points; Calculate the difference between the basic time feature vectors corresponding to adjacent sub-time segment data, extract the mean change and fluctuation range change rate of the two previous and next sub-time segment data, and generate a change rate feature vector representing the time dynamic characteristics; The basic time feature vector and the change rate feature vector are concatenated to obtain a time dimension change feature sequence including absolute features and relative change features, wherein each element in the time dimension change feature sequence includes corresponding time tag information.

4. The method according to claim 2, wherein The gas concentration data in the environmental data set are spatially gridded, and the city blocks are divided into a plurality of grid cells marked with geographic coordinates, each grid cell corresponding to a set of grid data sets containing gas concentration data and meteorological element data, including: Obtaining geographic boundary information of a city block, and evenly dividing the area space corresponding to the geographic boundary information into a plurality of grid cells with unique geographic coordinate markers according to preset grid size parameters; Identify the environmental monitoring device contained in each grid unit, obtain the gas concentration data and meteorological element data collected by the environmental monitoring device within a preset time period, and construct a mapping relationship table between the grid unit and the monitoring data; Perform spatial interpolation on the gas concentration data within each grid cell. When there are multiple environmental monitoring devices within a grid cell, calculate the spatial impact weight of each device based on the installation height and monitoring range radius of the environmental monitoring device, and fuse the multi-source monitoring data within the same grid cell through weighted averaging. The fused gas concentration data and meteorological element data are temporally and spatially aligned so that the gas concentration data and meteorological element data in the same grid cell have the same time tag sequence, generating a grid data set containing dual indexes of time dimension and space dimension.

5. The method according to claim 4, wherein The method of performing deep feature extraction on the attention-weighted feature map through the multi-layer perceptron module of the carbon flux analysis model, and mapping the attention-weighted feature map into a carbon flux-related feature vector in a high-dimensional feature space through multi-layer nonlinear transformation, includes: Inputting the attention-weighted feature map into the first convolutional layer of the multi-layer perceptron module, performing local feature extraction on the attention-weighted feature map using a convolution kernel of a preset size, and generating a primary feature response map containing local spatial features; Inputting the primary feature response map into a batch normalization processing layer, normalizing the feature values ​​in the primary feature response map so that the mean of each feature channel is zero and the standard deviation is one, thereby generating a normalized feature map with numerical stability; Performing a nonlinear transformation on the normalized feature map through the activation function layer of the multi-layer perceptron module to generate an activation feature map containing nonlinear transformation features; Inputting the activation feature map into the pooling operation layer, performing spatial dimension reduction on the activation feature map by the maximum pooling method, retaining the significant feature information in the feature map, and generating a dimensionally compressed pooled feature map; The pooled feature map is input into a fully connected layer network, and high-dimensional feature mapping is performed on the pooled feature map through a multi-layer neuron connection weight matrix to generate an abstract feature vector with carbon flux correlation characteristics. The dimension of the abstract feature vector increases exponentially with the increase of network depth.

6. The method according to claim 4, wherein The spatiotemporal consistency check of the carbon flux associated characteristic vector is performed to ensure that the characteristic vector change trend of adjacent time segments and the characteristic vector distribution law of adjacent spatial regions meet the preset physical constraint conditions, including: Extract the carbon flux-related feature vector sequence of continuous time segments, calculate the cosine similarity of the feature vectors between adjacent time segments, and generate the feature change continuity score in the time dimension; When the continuity score is lower than a preset threshold, the carbon flux-related eigenvector is temporally smoothed, and the abnormally fluctuating eigenvector value is corrected by a weighted moving average algorithm so that the trend of the corrected eigenvector conforms to the preset temporal continuity constraint; Performing spatial gradient calculation on carbon flux-related characteristic vectors of adjacent spatial regions to generate a spatial gradient matrix representing a spatial characteristic change rate, wherein each element in the spatial gradient matrix represents a directional derivative of the characteristic vector of the adjacent region; Analyze the gradient direction consistency in the spatial gradient matrix. When the gradient direction angle of adjacent regions exceeds a preset angle threshold, perform spatial interpolation correction on the feature vectors of the corresponding regions so that the feature changes in the spatial dimension meet the preset spatial continuity constraint conditions. The eigenvectors corrected for temporal continuity and the eigenvectors corrected for spatial continuity are fused to generate a set of carbon flux-related eigenvectors that satisfies both temporal and spatial constraints.

7. The method according to claim 1, wherein The identifying the location characteristics and range characteristics of the area with abnormal carbon flux changes based on the carbon flux change characteristics of the urban blocks includes: Performing eigenvalue decomposition on the carbon flux change characteristics of the urban blocks, extracting principal component components from the eigenvalues, and generating a set of principal component eigenvectors containing the main change patterns; Input the principal component feature vector set into an abnormal pattern recognition function, and identify abnormal feature points that significantly deviate from the normal distribution range by deviation calculation, where each abnormal feature point includes a corresponding time stamp and spatial coordinates; Performing spatial clustering on the abnormal feature points, aggregating adjacent abnormal feature points into a plurality of abnormal feature regions according to the Euclidean distance of spatial coordinates, each abnormal feature region containing at least three abnormal feature points; Calculate the minimum circumscribed polygon boundary of each abnormal feature area and extract the vertex coordinate sequence of the polygon boundary as the position feature descriptor of the abnormal change area; The geometric center coordinates of the minimum circumscribed polygon boundary and the distance set from each vertex to the center coordinates are measured to generate a range feature descriptor including the center coordinates and maximum distance parameters.

8. The method according to claim 7, wherein The spatial clustering of the abnormal feature points, aggregating adjacent abnormal feature points into multiple abnormal feature areas according to the Euclidean distance of spatial coordinates, includes: Initializing a cluster center set, randomly selecting a preset number of initial cluster center points from the abnormal feature points, each cluster center point having a unique spatial coordinate identifier; Calculate the Euclidean distance from each abnormal feature point to all cluster center points, assign each abnormal feature point to the cluster subset corresponding to the nearest cluster center point, and obtain the preliminary cluster division result; Calculate the geometric center coordinates of each cluster subset as the new cluster center point, and repeat the distance calculation and assignment steps until the moving distance of the cluster center point is less than the preset convergence threshold or reaches the preset number of iterations; Calculate the silhouette coefficient of each cluster subset, which is used to evaluate the compactness and separation of the clustering results. When the silhouette coefficient exceeds a preset threshold, it is determined to be a valid abnormal feature area; Merge adjacent valid abnormal feature regions. When the minimum circumscribed rectangles of two regions overlap and the overlap ratio exceeds a preset ratio threshold, merge them into a single abnormal feature region. The method further comprises: Parsing the polygon boundary vertex coordinate sequence in the location feature, and converting the polygon boundary vertex coordinate sequence into a geographic coordinate set in a city block map coordinate system through a coordinate conversion algorithm; Marking the spatial coverage of the abnormal change area on the digital map according to the geographic coordinate set, and generating a visual spatial distribution map including the area contour line and the center identification point; Extracting the center coordinates and maximum distance parameters from the range features, calculating the area coverage value and perimeter parameters of the abnormal change area, and generating a numerical feature set that quantitatively describes the area scale; Fusing the visual spatial distribution map and the numerical feature set to generate a comprehensive monitoring report framework including spatial positioning information and quantitative indicators; Time stamp information and carbon flux change characteristic description are embedded in the comprehensive monitoring report framework to generate a block carbon flux monitoring report containing spatiotemporal distribution characteristics.

9. A computer system, characterized in that: include: a memory storing a computer program; A processor, configured to load the computer program to implement the environmental data monitoring method based on cloud computing and spatiotemporal characteristics as described in any one of claims 1 to 8.

Images