Method, device and equipment for generating abnormal clues based on multi-source data and medium
By performing spatiotemporal alignment and multi-level anomaly detection on multi-source heterogeneous data, anomaly clues are generated, solving the problems of data silos and limited anomaly identification, and achieving accurate anomaly push and efficient law enforcement response.
Patent Information
- Application Number
- CN202610298965.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-12
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2046-03-12
AI Technical Summary
Existing technologies suffer from severe data silos, limited anomaly identification methods, inaccurate anomaly alerts, and low efficiency in law enforcement response.
By performing spatiotemporal alignment processing on multi-source heterogeneous data and mapping it to a multi-layer coupled grid, feature extraction and multi-level anomaly detection are performed. Combined with static spatial matching, dynamic spatiotemporal backtracking and multi-source causal chain reasoning, anomaly clues are generated.
It has achieved cross-media, cross-scale, and cross-modal data fusion, which has improved the source identification and causal evidence support of abnormal events, and enhanced the accuracy of abnormal event notifications and the efficiency of law enforcement response.
Smart Images

Figure CN122196086B_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of environmental monitoring and management technology, specifically relating to methods, devices, equipment and media for generating anomaly clues based on multi-source data. Background Technology
[0002] With the deepening of ecological and environmental supervision, the sources of monitoring data are becoming increasingly diverse, including air quality monitoring, pollution source emission monitoring, video surveillance and other types of data. These data have significant differences in sampling frequency, spatial resolution and data format, and are scattered in different departmental systems, forming serious "data silos".
[0003] The shortcomings of existing methods include: severe data silos: data on water, air, space, and meteorology lack unified integration across media and modalities, making it difficult to construct a complete environmental situation map; limited anomaly identification methods: mainly relying on single-point threshold alarms or single-variable trend identification, lacking coupled anomaly mining of multi-variable and multi-source data; and inaccurate push notifications: existing anomaly events often remain at the level of indicator alarms, failing to link spatial location, causal chains, and responsible parties, resulting in low law enforcement response efficiency. Summary of the Invention
[0004] In view of the above analysis, the embodiments of the present invention aim to provide a method, apparatus, device and medium for generating anomaly clues based on multi-source data, so as to solve the technical problems of severe data silos, single anomaly identification methods and inaccurate push in the prior art.
[0005] The objective of this invention is achieved as follows: A first aspect of the present invention provides a method for generating anomaly clues based on multi-source data, comprising: Collect multi-source heterogeneous data, and map the multi-source heterogeneous data to a multi-layer coupled grid after spatiotemporal alignment processing to obtain gridded fused data. The multi-layer coupled grid includes macroscopic layer grid, mesoscopic layer grid and microscopic layer grid. Feature extraction is performed on the gridded fused data to obtain multi-level structured features. Multi-level anomaly detection is then performed based on the multi-level structured features to identify abnormal events. Based on the aforementioned abnormal events, static spatial matching, dynamic spatiotemporal backtracking, and multi-source causal chain reasoning are performed to generate abnormal clues including candidate pollution sources, causal probabilities, and evidence chains.
[0006] Furthermore, the multi-source heterogeneous data includes environmental monitoring data, pollution source data, meteorological data, geospatial data, and auxiliary data. The environmental monitoring data includes air quality monitoring data and water environment monitoring data. The pollution source data includes discharge permits, administrative penalties, concentration monitoring data, electricity consumption data, and video data. The meteorological data includes wind direction, wind speed, and temperature. The geospatial data includes administrative divisions and the distribution of sewage outlets. The auxiliary data includes the distribution of industrial parks and zoning control units.
[0007] Furthermore, the spatiotemporal alignment processing includes: dividing watershed units based on a digital elevation model, dividing air mass units based on wind field and topography, and dividing administrative grid units based on administrative divisions; establishing a composite spatial index for any point in space, composed of the corresponding watershed unit identifier, air mass unit identifier, and administrative grid unit identifier; aggregating high-frequency data from the multi-source heterogeneous data according to a first time granularity, filling low-frequency data from the multi-source heterogeneous data according to a second time granularity, and performing spatial interpolation on the meteorological data to establish a unified time index; mapping the spatial coordinates of the multi-source heterogeneous data to the composite spatial index to generate a spatial key; aligning the timestamps of the multi-source heterogeneous data to the unified time index to generate a standard timestamp; and concatenating the spatial key with the standard timestamp to generate a unified spatiotemporal primary key for the multi-source heterogeneous data.
[0008] Furthermore, the macro layer is used to integrate environmental monitoring data, meteorological data, and satellite inversion data; the meso layer is used to integrate pollution discharge permits, administrative penalties, and total electricity consumption; and the micro layer is used to integrate concentration monitoring data, video data, and minute-level electricity consumption data.
[0009] Furthermore, the feature extraction of the gridded fused data includes: standardizing the numerical monitoring data; performing trend analysis on the standardized data using a sliding window to extract trend features and fluctuation features; using a natural language processing model to perform semantic classification on discharge permits and penalty records to obtain pollution behavior labels; using a target detection model to perform event recognition on video data to obtain a visual anomaly event stream; and using frequency domain analysis to perform anomaly recognition on electricity consumption data to obtain a proxy variable for discharge intensity.
[0010] Furthermore, the multi-level anomaly detection includes: performing gradual trend detection based on the trend characteristics to identify slow-rising anomalies; performing multi-indicator coupling detection based on the meteorological data, concentration monitoring data, and pollution intensity proxy variables to identify abnormal coupling between meteorology, emissions, and energy consumption; and performing cross-media collaborative detection based on the environmental monitoring data and meteorological data to identify compound pollution; the anomaly detection result is an anomaly event that includes abnormal trends, abnormal coupling, and compound pollution.
[0011] Furthermore, the static spatial matching, dynamic spatiotemporal backtracking, and multi-source causal chain inference based on the anomalous event include: calculating the spatial correlation between the grid where the anomalous event is located and the administrative region, industrial park, and sewage outlet; determining the first candidate pollution source based on the spatial correlation; calculating the theoretical concentration and arrival time from the first candidate pollution source to the grid where the anomalous event is located based on the Gaussian plume diffusion model; calculating the transport path and peak arrival time of pollutants along the river based on the one-dimensional river diffusion model; performing spatiotemporal matching of the calculation results with the anomalous event to screen the second candidate pollution source; constructing a multi-source evidence node including pollution behavior tags, video anomalous event streams, and sewage discharge intensity proxy variables; calculating the causal probability of the second candidate pollution source under various evidence conditions based on the Bayesian inference model; weighted fusion of the causal probabilities of the second candidate pollution source under various evidence conditions; and generating anomaly clues including a list of candidate pollution sources, corresponding causal probabilities, and evidence chains.
[0012] A second aspect of the present invention provides an apparatus for generating anomaly clues based on multi-source data, comprising: The data collection and processing module is used to collect multi-source heterogeneous data, and after the multi-source heterogeneous data undergoes spatiotemporal alignment processing, it maps the multi-source heterogeneous data onto a multi-layer coupled grid to obtain gridded fused data. An abnormal event identification module is used to extract features from the gridded fused data to obtain multi-level structured features, and to perform multi-level anomaly detection based on the multi-level structured features to identify abnormal events. The anomaly clue generation module is used to perform static spatial matching, dynamic spatiotemporal backtracking, and multi-source causal chain reasoning based on the anomaly event, and generate anomaly clues including candidate pollution sources, causal probabilities, and evidence chains.
[0013] A third aspect of the present invention provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and the computer program, when executed by the processor, implements the method for generating anomaly clues based on multi-source data as described in any embodiment.
[0014] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method for generating anomaly clues based on multi-source data as described in any embodiment.
[0015] Compared with the prior art, the present invention can achieve at least one of the following beneficial effects: The method for generating anomaly clues based on multi-source data provided by this invention achieves cross-media, cross-scale, and cross-modal data fusion by uniformly aligning multi-source heterogeneous data in space and time and mapping it to a multi-layered coupled grid at the macro, meso, and micro levels, thus solving the "data silo" problem. By generating multi-level structured features from the fused data and performing multi-level anomaly detection, it overcomes the problem of insufficient anomaly identification capability caused by traditional methods relying on single-point thresholds or single-variable trends. Based on anomaly events, it performs static spatial matching, dynamic spatiotemporal backtracking, and multi-source causal chain reasoning, unifying multi-modal information into a causal analysis framework to generate anomaly clues containing candidate pollution sources, causal probabilities, and evidence chains. This gives anomaly events clear source orientation and causal evidence support, significantly improving the accuracy of anomaly alerts and the efficiency of law enforcement response. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of this specification or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the embodiments of this specification. For those skilled in the art, other drawings can be obtained based on these drawings.
[0017] Figure 1 This is a flowchart of the method for generating anomaly clues based on multi-source data provided in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the device for generating anomaly clues based on multi-source data provided in Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the electronic device architecture provided in Embodiment 3 of the present invention. Detailed Implementation
[0018] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. It should be noted that, unless otherwise specified, the implementation methods and features in the implementation methods in this disclosure can be combined, separated, interchanged, and / or rearranged. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0019] Example 1 A specific embodiment of the present invention, such as Figure 1 As shown, a method for generating anomaly clues based on multi-source data is disclosed, including the following steps: S1. Collect multi-source heterogeneous data, and map the multi-source heterogeneous data to a multi-layer coupled grid after spatiotemporal alignment processing to obtain gridded fused data. The multi-layer coupled grid includes a macroscopic layer grid, a mesoscopic layer grid, and a microscopic layer grid.
[0020] In this embodiment, the multi-source heterogeneous data includes environmental monitoring data, pollution source data, meteorological data, geospatial data, and auxiliary data. The environmental monitoring data includes air quality monitoring data and water environment monitoring data. The pollution source data includes discharge permits, administrative penalties, concentration monitoring data, electricity consumption data, and video data. The meteorological data includes wind direction, wind speed, and temperature. The geospatial data includes administrative divisions and the distribution of sewage outlets. The auxiliary data includes the distribution of industrial parks and zoning control units.
[0021] Because multi-source heterogeneous data includes structured, semi-structured and unstructured data, and the sampling time scale and spatial precision differ significantly, spatiotemporal alignment processing is required.
[0022] The spatiotemporal alignment process provided in this embodiment includes the following steps: (1) Divide watershed units based on digital elevation model, air mass units based on wind field and topography, and administrative grid units based on administrative divisions, and establish a composite spatial index for any point in space, consisting of the corresponding watershed unit identifier, air mass unit identifier and administrative grid unit identifier. (2) Aggregate the high-frequency data in the multi-source heterogeneous data according to the first time granularity, fill the low-frequency data in the multi-source heterogeneous data according to the second time granularity, and perform spatial interpolation on the meteorological data to establish a unified time index. (3) Map the spatial coordinates of the multi-source heterogeneous data to the composite spatial index to generate a spatial key, and align the timestamps of the multi-source heterogeneous data to the unified time index to generate a standard timestamp; (4) The spatial key is concatenated with the standard timestamp to generate a unified spatiotemporal primary key for the multi-source heterogeneous data.
[0023] In this embodiment, the macroscopic layer in the multi-layer coupled grid is used to fuse environmental monitoring data, meteorological data, and satellite inversion data; the mesoscopic layer is used to fuse pollution discharge permits, administrative penalties, and total electricity consumption; and the microscopic layer is used to fuse concentration monitoring data, video data, and minute-level electricity consumption data.
[0024] For example, the fusion process is as follows: Macro level: Kriging or inverse distance interpolation (IDW) is performed on the air quality monitoring station data according to the air mass unit to obtain the grid-level air quality concentration; the wind field model is called to calculate the pollutant adsorption and diffusion coefficient between adjacent macro grids; meteorological data (wind speed, wind direction, temperature, stability, etc.) are interpolated to the center of the macro level grid to form a grid-level meteorological field.
[0025] Mesoscopic layer: Project industry distribution, administrative divisions and industrial park boundaries onto mesoscopic grids; aggregate pollution discharge permits and historical penalties according to grid affiliation; downscale macroscopic meteorological fields to the mesoscopic layer and form an "emission potential matrix" with mesoscopic emission permit parameters; divide the total electricity consumption of the park into each mesoscopic grid according to area weight.
[0026] Micro-level layer: A buffer zone is constructed with sewage outlets and monitoring points as the center, and minute-level automatic monitoring data is mapped to the corresponding micro-grids; visual events are projected onto the micro-grids using the geographic coordinates of video surveillance; and the industry attributes, sewage discharge permits, and violation tendencies of the meso-level layer are pushed down to the micro-level layer according to grid proximity.
[0027] S2. Extract features from the gridded fused data to obtain multi-level structured features, and perform multi-level anomaly detection based on the multi-level structured features to identify abnormal events.
[0028] In this embodiment, the feature extraction of the gridded fused data includes: The numerical monitoring data is standardized, and a sliding window is used to perform trend analysis on the standardized data to extract trend and fluctuation features. A natural language processing model is used to perform semantic classification on discharge permits and penalty records to obtain pollution behavior labels. A target detection model is used to perform event recognition on video data to obtain a stream of visual anomalies. Frequency domain analysis is used to identify anomalies in electricity consumption data to obtain a proxy variable for discharge intensity.
[0029] In this embodiment, the multi-level anomaly detection includes: Gradual trend detection is performed based on the aforementioned trend characteristics to identify slow-rising anomalies; multi-indicator coupling detection is performed based on the aforementioned meteorological data, concentration monitoring data, and pollution intensity proxy variables to identify abnormal coupling between meteorology, emissions, and energy consumption; and cross-media collaborative detection is performed based on the aforementioned environmental monitoring data and meteorological data to identify compound pollution. The anomaly detection results include abnormal events such as abnormal trends, abnormal coupling, and complex contamination.
[0030] For example, by calculating the cumulative difference sum of time-series data, if the cumulative difference sum is greater than a preset threshold for 5 consecutive hours and the slope in the trend feature is greater than 0.1 units / hour, it is marked as a slow-rising anomaly; an interaction matrix is constructed based on meteorological data, concentration monitoring data and pollution intensity proxy variables, then the Pearson correlation coefficient is calculated, and the Granger causality test is passed. If the three are synchronously abnormal, it is marked as anomalous coupling; if the air quality monitoring data and water environment monitoring data in the same grid synchronously exceed the standard, and the meteorological conditions support cross-media transmission (such as high temperature and low wind speed), it is marked as meeting the pollution criteria.
[0031] In some embodiments, during gradual trend detection, a sliding window can be used first to fit the data within the window using piecewise linear regression; the fitted trend line is retained, and the residual sequence is obtained by subtracting the trend from the original data; the residual is denoised using the 3σ principle, and then the trend is superimposed to output the "denoised trend-preserved sequence"; this is used to eliminate meteorological fluctuation interference and highlight anomalies in anthropogenic emissions; and dynamic weights are introduced to weight the cumulative difference sum to highlight abrupt change points and periodic anomalies.
[0032] S3. Based on the abnormal events, perform static spatial matching, dynamic spatiotemporal backtracking, and multi-source causal chain reasoning to generate abnormal clues including candidate pollution sources, causal probabilities, and evidence chains.
[0033] In this embodiment, S3 specifically includes: S301. Calculate the spatial correlation between the grid where the abnormal event is located and the administrative region, industrial park and sewage outlet, and determine the first candidate pollution source based on the spatial correlation. Specifically, when performing static spatial matching, the spatial correlation degree is calculated as follows: ,in, This represents the area where the grid containing the abnormal event intersects with the administrative district, industrial park, and sewage outlet. This represents the area of administrative districts, industrial parks, and sewage outlets. The higher the spatial correlation, the greater the probability that the abnormal event occurred in that administrative district, industrial park, or sewage outlet.
[0034] S302. Calculate the theoretical concentration and arrival time of the first candidate pollution source to the grid where the abnormal event is located based on the Gaussian plume diffusion model, calculate the transport path and peak arrival time of pollutants along the river based on the one-dimensional river diffusion model, and perform spatiotemporal matching of the calculation results with the abnormal event to screen the second candidate pollution source. Specifically, when the abnormal event involves air pollutants (such as PM, SO2, NO2, NH3), a Gaussian plume diffusion model is used to calculate the theoretical concentration of the first candidate pollution source at the current wind speed and wind direction, as well as the theoretical arrival time of the pollutant to the grid where the abnormal event is located. If "theoretical arrival time ∈ (abnormal event time ± 2 hours)" and "theoretical concentration ≥ preset ratio of monitored concentration (e.g., 0.7)", then the candidate pollution source is confirmed as the second candidate pollution source. When the abnormal event involves water pollutants (COD, TP, etc.), a one-dimensional river diffusion model is used to calculate whether the source tracing path is along the upstream of the river; whether the pollution peak arrival time falls within the abnormal time window; if the above spatiotemporal consistency criteria are met, then the water discharge outlet is confirmed as the second candidate pollution source.
[0035] S303. Construct a multi-source evidence node including pollution behavior tags, video abnormal event streams, and pollution intensity proxy variables. Calculate the causal probability of the second candidate pollution source under various evidence conditions based on a Bayesian inference model. Perform weighted fusion of the causal probabilities of the second candidate pollution source under various evidence conditions to generate an abnormal clue containing a list of candidate pollution sources, corresponding causal probabilities, and evidence chains.
[0036] Compared with existing technologies, the method for generating anomaly clues based on multi-source data provided in this embodiment achieves cross-media, cross-scale, and cross-modal data fusion by uniformly aligning multi-source heterogeneous data in time and space and mapping it to a multi-layered coupled grid of macroscopic, mesoscopic, and microscopic dimensions, thus solving the "data silo" problem. By generating multi-level structured features from the fused data and performing multi-level anomaly detection, it overcomes the problem of insufficient anomaly identification capability caused by traditional methods relying on single-point thresholds or single-variable trends. Based on anomaly events, it performs static spatial matching, dynamic time and space backtracking, and multi-source causal chain reasoning, unifying multi-modal information into a causal analysis framework to generate anomaly clues containing candidate pollution sources, causal probabilities, and evidence chains. This gives anomaly events clear source orientation and causal evidence support, significantly improving the accuracy of anomaly push and the efficiency of law enforcement response.
[0037] In some embodiments, the method further includes a comprehensive assessment of the pollution intensity, cross-media impact, and industry risks of the abnormal event based on the causal probability calculation results of the second candidate pollution source, in order to improve the risk differentiation capability and regulatory targeting of abnormal clues, specifically including: (1) Pollution intensity assessment The pollution intensity index is calculated based on the duration, scope of impact, and increase in pollutant concentration of the abnormal event, and is expressed as: ,in, Indicates the increase in pollutant concentration. T Indicates the duration of the abnormal event. SThe pollution intensity index indicates the scope of impact of an abnormal event and is used to quantify the hazard level of the abnormal event itself.
[0038] (2) Cross-media impact assessment Based on the concentration characteristics of anomalous events in water and atmospheric monitoring, the cross-media coupling coefficient is calculated and expressed as: ,in, and Representing the medium i and j If the concentration of pollutants in the water, K A value >1 indicates an amplification effect during cross-media propagation, and the abnormal event is marked as a "high-risk cross-media event".
[0039] (3) Calculation of industry correction coefficient To improve the consistency of risk assessments for different types of enterprises, an industry correction factor can be introduced to adjust the pollution intensity index, expressed as: ,in, This represents the industry adjustment factor. For example, the industry adjustment factor for chemical companies is 1.2, and the industry adjustment factor for manufacturing companies is 1.0.
[0040] In some embodiments, the method further includes generating structured anomaly clue results and visualizing them based on the aforementioned causal reasoning, pollution intensity assessment, and pollution intensity index correction. Specifically, this involves constructing a link graph of "anomaly event—monitoring point—candidate pollution source—causal probability," where nodes represent monitoring points, anomaly events, and candidate pollution sources; edge weights represent causal probability or the degree of evidence support; and the graph can be stored and retrieved using an adjacency matrix or graph database. Based on this link graph, it is possible to achieve rapid location and correlation analysis of anomalies, ranking of candidate pollution sources and generation of key monitoring lists, multi-source correlation tracing between different events, risk level classification of anomalies, and early warning reporting.
[0041] In some embodiments, high-confidence anomaly samples can also be used as training or validation sets for subsequent anomaly detection and causal reasoning models to form a closed-loop optimization and continuously improve the accuracy and stability of anomaly detection and source tracing reasoning.
[0042] Example 2 This embodiment provides an apparatus for generating anomaly clues based on multi-source data, such as... Figure 2 As shown, it includes: The data collection and processing module is used to collect multi-source heterogeneous data, and after the multi-source heterogeneous data undergoes spatiotemporal alignment processing, it maps the multi-source heterogeneous data onto a multi-layer coupled grid to obtain gridded fused data. An abnormal event identification module is used to extract features from the gridded fused data to obtain multi-level structured features, and to perform multi-level anomaly detection based on the multi-level structured features to identify abnormal events. The anomaly clue generation module is used to perform static spatial matching, dynamic spatiotemporal backtracking, and multi-source causal chain reasoning based on the anomaly event, and generate anomaly clues including candidate pollution sources, causal probabilities, and evidence chains.
[0043] Example 3 This embodiment provides an electronic device, such as... Figure 3 As shown, it includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, implements the method for generating anomaly clues based on multi-source data as described in any of the above embodiments.
[0044] Example 4 This embodiment provides a computer-readable storage medium storing a computer program thereon. When the program is executed by a processor, it implements the method for generating anomaly clues based on multi-source data as described in any of the above embodiments.
[0045] Computer-readable storage media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0046] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0047] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented in hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.
[0048] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for generating anomaly clues based on multi-source data, characterized in that, include: Collect multi-source heterogeneous data including environmental monitoring data, pollution source data, meteorological data, geospatial data and auxiliary data, and map the multi-source heterogeneous data into a multi-layer coupled grid after spatiotemporal alignment processing to obtain gridded fused data. The multi-layer coupled grid includes macroscopic layer grid, mesoscopic layer grid and microscopic layer grid. The spatiotemporal alignment process includes: dividing watershed units based on a digital elevation model, dividing air mass units based on wind field and topography, and dividing administrative grid units based on administrative divisions; establishing a composite spatial index for any point in space, composed of the corresponding watershed unit identifier, air mass unit identifier, and administrative grid unit identifier; aggregating high-frequency data from the multi-source heterogeneous data according to a first time granularity, filling low-frequency data from the multi-source heterogeneous data according to a second time granularity, and performing spatial interpolation on the meteorological data to establish a unified time index; mapping the spatial coordinates of the multi-source heterogeneous data to the composite spatial index to generate a spatial key; aligning the timestamps of the multi-source heterogeneous data to the unified time index to generate a standard timestamp; and concatenating the spatial key with the standard timestamp to generate a unified spatiotemporal primary key for the multi-source heterogeneous data. Feature extraction is performed on the gridded fused data to obtain multi-level structured features, which include trend features and fluctuation features, pollution behavior labels, visual abnormal event streams and pollution intensity proxy variables. Multi-level anomaly detection is performed based on the multi-level structured features to identify abnormal events. The multi-level anomaly detection includes: performing gradual trend detection based on the trend characteristics to identify slow-rising anomalies; performing multi-indicator coupling detection based on the meteorological data, concentration monitoring data, and pollution intensity proxy variables to identify abnormal coupling between meteorology, emissions, and energy consumption; and performing cross-media collaborative detection based on the environmental monitoring data and meteorological data to identify compound pollution. Based on the anomalous events, static spatial matching, dynamic spatiotemporal backtracking, and multi-source causal chain inference are performed to generate anomalous clues including candidate pollution sources, causal probabilities, and evidence chains. This includes: calculating the spatial correlation between the grid where the anomalous event is located and administrative regions, industrial parks, and sewage outlets; determining the first candidate pollution source based on the spatial correlation; calculating the theoretical concentration and arrival time from the first candidate pollution source to the grid where the anomalous event is located based on a Gaussian plume diffusion model; calculating the transport path and peak arrival time of pollutants along the river based on a one-dimensional river diffusion model; performing spatiotemporal matching of the calculation results with the anomalous events to screen for second candidate pollution sources; constructing multi-source evidence nodes including pollution behavior tags, video anomalous event streams, and sewage discharge intensity proxy variables; calculating the causal probability of the second candidate pollution source under various evidence conditions based on a Bayesian inference model; weighted fusion of the causal probabilities of the second candidate pollution source under various evidence conditions; and generating anomalous clues including a list of candidate pollution sources, corresponding causal probabilities, and evidence chains.
2. The method for generating anomaly clues based on multi-source data according to claim 1, characterized in that, The environmental monitoring data includes air quality monitoring data and water environment monitoring data; the pollution source data includes discharge permits, administrative penalties, concentration monitoring data, electricity consumption data, and video data; the meteorological data includes wind direction, wind speed, and temperature; the geospatial data includes administrative divisions and the distribution of discharge outlets; and the auxiliary data includes the distribution of industrial parks and zoning control units.
3. The method for generating anomaly clues based on multi-source data according to claim 1, characterized in that, The macro layer is used to integrate environmental monitoring data, meteorological data, and satellite inversion data; the meso layer is used to integrate pollution discharge permits, administrative penalties, and total electricity consumption; and the micro layer is used to integrate concentration monitoring data, video data, and minute-level electricity consumption data.
4. The method for generating anomaly clues based on multi-source data according to claim 2, characterized in that, The feature extraction of the gridded fused data includes: The numerical monitoring data is standardized, and a sliding window is used to perform trend analysis on the standardized data to extract trend and fluctuation features. A natural language processing model is used to perform semantic classification on discharge permits and penalty records to obtain pollution behavior labels. A target detection model is used to perform event recognition on video data to obtain a stream of visual anomalies. Frequency domain analysis is used to identify anomalies in electricity consumption data to obtain a proxy variable for discharge intensity.
5. A device for generating anomaly clues based on multi-source data, characterized in that, The device includes: The data collection and processing module is used to collect multi-source heterogeneous data, including environmental monitoring data, pollution source data, meteorological data, geospatial data, and auxiliary data. After spatiotemporal alignment processing, the multi-source heterogeneous data is mapped to a multi-layer coupled grid to obtain gridded fused data. The spatiotemporal alignment processing includes: dividing watershed units based on a digital elevation model, air mass units based on wind field and topography, and administrative grid units based on administrative divisions; establishing a composite spatial index for any point in space, composed of the corresponding watershed unit identifier, air mass unit identifier, and administrative grid unit identifier; aggregating high-frequency data from the multi-source heterogeneous data according to a first time granularity, filling low-frequency data according to a second time granularity, and spatially interpolating the meteorological data to establish a unified time index; mapping the spatial coordinates of the multi-source heterogeneous data to the composite spatial index to generate a spatial key; aligning the timestamps of the multi-source heterogeneous data to the unified time index to generate a standard timestamp; and concatenating the spatial key with the standard timestamp to generate a unified spatiotemporal primary key for the multi-source heterogeneous data. An anomaly identification module is used to extract features from the gridded fused data to obtain multi-level structured features. These multi-level structured features include trend and fluctuation features, pollution behavior labels, visual anomaly event streams, and pollution intensity proxy variables. Multi-level anomaly detection is performed based on these multi-level structured features to identify anomaly events. Specifically, the multi-level anomaly detection includes: gradual trend detection based on the trend features to identify slowly rising anomalies; multi-indicator coupling detection based on meteorological data, concentration monitoring data, and pollution intensity proxy variables to identify abnormal coupling between meteorology, emissions, and energy consumption; and cross-media collaborative detection based on environmental monitoring data and meteorological data to identify compound pollution. An anomaly clue generation module is used to perform static spatial matching, dynamic spatiotemporal backtracking, and multi-source causal chain inference based on the anomaly event, generating anomaly clues including candidate pollution sources, causal probabilities, and evidence chains. This includes: calculating the spatial correlation between the grid where the anomaly event is located and administrative regions, industrial parks, and sewage outlets; determining a first candidate pollution source based on the spatial correlation; calculating the theoretical concentration and arrival time from the first candidate pollution source to the grid where the anomaly event is located based on a Gaussian plume diffusion model; calculating the transport path and peak arrival time of pollutants along the river based on a one-dimensional river diffusion model; performing spatiotemporal matching of the calculation results with the anomaly event to screen for a second candidate pollution source; constructing multi-source evidence nodes including pollution behavior tags, video anomaly event streams, and sewage discharge intensity proxy variables; calculating the causal probability of the second candidate pollution source under various evidence conditions based on a Bayesian inference model; weighted fusion of the causal probabilities of the second candidate pollution source under various evidence conditions; and generating anomaly clues including a list of candidate pollution sources, corresponding causal probabilities, and evidence chains.
6. An electronic device, characterized in that, The system includes a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, implements the method for generating anomaly clues based on multi-source data as described in any one of claims 1-4.
7. A storage medium, characterized in that, It stores a computer program, which, when executed by a processor, implements the method for generating anomaly clues based on multi-source data as described in any one of claims 1-4.
Citation Information
Patent Citations
Mine data fusion and anomaly intelligent mining method, system and equipment and medium
CN121389019A
