Oil spill tracing method and system based on space-time mapping and dynamic backtracking

Abstract

The invention discloses an oil spill tracing method and system based on space-time mapping and dynamic backtracking, and belongs to the technical field of marine environment monitoring and oil spill tracing. The oil spill traceability method is realized through the oil spill traceability system. The method comprises the following steps: acquiring a remote sensing image containing an oil spill area, suspected ship AIS data and marine environment data; calibrating a central skeleton point set of the oil spilling area; space-time mapping is executed for each suspected ship, and reference emission time is calibrated for each skeleton point; performing dynamic reverse backtracking of different durations on the central skeleton point set, and reconstructing a simulated pollution discharge path; and calculating the space-time similarity between the simulated pollution discharge path and the AIS track of the suspected ship, obtaining a comprehensive traceability score, and determining the oil spill source ship. Through space-time mapping and a dynamic backtracking mechanism, a time length basis with physical significance is provided for reverse backtracking, the problem of physical distortion caused by uncertain backtracking time length is solved, and the accuracy and reliability of oil spill traceability of the mobile source are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of marine environmental monitoring technology, and relates to an oil spill source tracing method and system based on spatiotemporal mapping and dynamic backtracking, particularly to an oil spill source tracing method and system based on remote sensing images and ship dynamic data from marine mobile sources. Background Technology

[0002] Effective protection and management of the marine environment requires the ability to respond quickly and accurately trace the source of sudden pollution events such as oil spills at sea. With the increasing shipping activities, mobile source oil spills caused by ships have become one of the major types of pollution. However, the drifting evolution of oil spills under the complex marine dynamics of wind and currents, as well as the spatiotemporal continuity of mobile source pollution behavior, together constitute the fundamental challenges in source tracing.

[0003] To address these challenges, existing technologies have explored various approaches. For example, Chinese invention patent (application number CN202010784950.3) employs a reverse-time tracking numerical model. This model discretizes the oil spill as a whole and calculates the backtracking time for all particles, attempting to find their common initial aggregation point. Chinese invention patent (application number CN118098423A) represents another technical approach, simulating a fixed, hypothetical source point using a forward diffusion model and relying on subsequent on-site sampling analysis for backtracking verification. While these solutions offer some insights, their underlying technical paradigms suffer from fundamental flaws: the "synchronous backtracking" mechanism used in the former contradicts the physical fact that mobile sources discharge pollution sequentially along their flight paths; the latter's "prediction-verification" process is not only slow to respond but also ineffective in handling the dynamic mobility of the source. Ultimately, these methods are confined to a simplified model of an "instantaneous fixed-point source," which directly leads to a lack of basis and serious bias in setting the backtracking physical time. Furthermore, existing technologies often ignore random disturbances in marine environmental parameters, and the source tracing trajectories lack necessary smoothing and probabilistic characterization, thus limiting the accuracy of the source tracing results. Summary of the Invention

[0004] To address the problems of existing technologies, this invention proposes an oil spill source tracing method and system based on spatiotemporal mapping and dynamic backtracking, aiming to solve the problem of low accuracy in mobile source oil spill source tracing caused by missing backtracking time references and physical distortions. The core of this invention lies in the following: First, by extracting the central skeleton point set of the oil spill area and spatiotemporally anchoring it with the AIS track of the suspected vessel, a physically meaningful and differentiated baseline discharge time is determined for the discrete points on the skeleton. This "spatiotemporal mapping" process provides an accurate time scale for subsequent reverse backtracking. Subsequently, a simulated cluster with multi-parameter perturbations is constructed and environmental uncertainties are injected. Based on this scale, dynamic reverse backtracking is performed. A high-precision simulated discharge path is reconstructed through trajectory post-processing and smoothing algorithms. Kernel density estimation (KDE) is used to generate a source tracing probability heatmap to accurately reconstruct the simulated discharge path. Finally, by performing multi-dimensional spatiotemporal similarity quantification comparison with the real track, reliable identification of the pollution source is achieved.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] An oil spill source tracing method based on spatiotemporal mapping and dynamic backtracking includes the following steps:

[0007] Step S1: Obtain remote sensing images containing the actual oil spill area, AIS track data from the Automatic Identification System (AIS), and marine environmental data.

[0008] Step S2 involves preprocessing the remote sensing imagery and identifying the central skeleton point set of the actual oil spill area; specifically:

[0009] Step S2.1 involves preprocessing the remote sensing image, including geometric correction and image filtering.

[0010] Step S2.2: Through the human-computer interaction interface, the actual oil spill area on the preprocessed remote sensing image is manually calibrated to obtain a central skeleton representing the direction of the oil spill.

[0011] Step S2.3: Perform equally spaced discretized sampling on the central skeleton to obtain a central skeleton point set P={p1,p2,...,pn} composed of several discrete points, where n is the total number of sampling points, p1 is the first central skeleton point, p2 is the second central skeleton point, and pn is the nth central skeleton point.

[0012] Step S3: Based on the imaging time of the remote sensing image and the location information of the actual oil spill area, and combined with AIS track data, suspect vessels are screened out; specifically:

[0013] Step S3.1: Based on the imaging time T0 of the remote sensing image, set a time threshold forward, for example, T = 24 hours, and construct a time filtering window [T0-T, T0], where T represents the length of the ship filtering time; at the same time, based on the center coordinates of the actual oil spill area, set a coverage radius, for example, R = 10 nautical miles, and construct a spatial filtering area.

[0014] Step S3.2: Retrieve the AIS database, extract the list of ships that have passed through the spatial filtering area within the time filtering window, and obtain their ship identification codes (MMSI) and corresponding latitude, longitude, heading, and timestamp data.

[0015] Step S3.3: Select vessels whose heading is consistent with the extension direction of the central frame and identify them as suspect vessels.

[0016] Step S4: For each suspected vessel, perform spatiotemporal mapping calibration to determine the reference emission time for each point in the central skeleton point set; specifically:

[0017] Step S4.1: Map the endpoints of the central skeleton of the oil spill area to the AIS track of the suspected vessel to obtain the first mapping point and the second mapping point; compare the timestamps of the two mapping points, take the mapping point with the earliest timestamp as the unique spatiotemporal anchor point, and determine the discharge start time as T_start.

[0018] Step S4.2: Obtain the total length L of the central skeleton. Starting from the spatiotemporal anchor point determined in step S4.1, project the total length along the AIS track to find the corresponding projection endpoint. Determine the time corresponding to the projection endpoint on the AIS track as the sewage discharge termination time, and define it as T_end.

[0019] Step S4.3: Based on the sewage discharge start time and sewage discharge end time, perform linear time interpolation calculations on each skeleton point in the central skeleton point set, thereby calculating the value for each skeleton point p. i Assign the corresponding baseline emission time t i The calculation formula is as follows: .

[0020] Step S5: Based on the baseline discharge time obtained for the suspected vessel in Step S4, and combined with the marine environmental data from Step S1, dynamically backtrack the central skeleton point set with differentiated durations to reconstruct a simulated discharge path corresponding to the suspected vessel; specifically:

[0021] Step S5.1: Each skeleton point in the central skeleton point set is set as an independent particle, and the spatial coordinates of each particle at the time of remote sensing image imaging are used as the initial position for its reverse backtracking. To improve statistical reliability, multiple sub-particles are randomly generated within a preset radius with each skeleton point as the center to construct a simulated cluster.

[0022] Step S5.2: For each particle, calculate the simulated backtracking duration based on the baseline emission time of its corresponding skeleton point; the simulated backtracking duration ΔT i Defined as the distance between the remote sensing image imaging time T0 and the corresponding baseline emission time t of the particle. i The time difference between them, i.e. ΔT i = T0 - t i This allows for the creation of differentiated time step constraints for each particle.

[0023] Step S5.3: Combining marine environmental data, control all particles to synchronously begin reverse motion from a unified imaging moment. The particle position update follows the inverse process of the following kinematic equations: , where, where, x i (t-Δt) x represents the updated particle position vector. i (t) Ci represents the position vector of particle i at time t, Cc represents the ocean current correction coefficient, and Ci (t) Cw represents the ocean current vector at the particle's location at time t, and Wi represents the wind deflection coefficient. (t) Δt represents the wind speed vector at the particle's location at time t, Δt represents the calculation time step during the solution process, Ud represents the random turbulent diffusion term, and t represents the current time in the simulation backtracking calculation process. During the backtracking process, an uncertainty injection mechanism is executed: through Monte Carlo sampling, differentiated horizontal diffusion coefficients, wind deflection factors, and ocean current field disturbance parameters are randomly assigned to each particle within a preset range. Each particle stops moving and locks its emission source point when it reaches the corresponding historical emission time according to its own simulation backtracking time.

[0024] Step S5.4: Summarize the coordinates of all emission source points after all particles stop moving, and use the Savitzky-Golay filter to denoise the particle swarm centroid sequence corresponding to each skeleton point to reconstruct a simulated discharge path that reflects the real physical process of continuous discharge of pollutants by the suspected vessel in the corresponding waterway segment.

[0025] Step S6: By calculating the spatiotemporal similarity between the simulated sewage discharge path corresponding to the suspected vessel and the actual AIS track of the suspected vessel, a comprehensive source tracing score for the suspected vessel is obtained, wherein the comprehensive source tracing score is proportional to the spatiotemporal similarity; specifically:

[0026] Step S6.1: Using the kernel density estimation (KDE) algorithm, the spatial distribution of particles after all backtracking stops is processed by fieldization to generate a heat map of the probability distribution of the source, so as to intuitively represent the high-probability core area of ​​the emission source.

[0027] Step S6.2: Using the Dynamic Time Warping (DTW) algorithm, calculate the cumulative distance between the smooth simulated sewage discharge path reconstructed in step S5 and the actual AIS track of the suspected vessel; automatically match the corresponding point pairs between the two paths using the DTW algorithm to effectively handle the local spatiotemporal stretching or compression caused by changes in vessel speed, and quantify the consistency of the two paths in motion patterns.

[0028] Step S6.3: The DTW distance obtained in step S6.2 is normalized by combining the trajectory length of the simulated path and the spatial scale of the study area. The normalization calculation formula is: normalized_dtw = D_total / L_path. This eliminates the influence of differences in different dimensions and path lengths. The dimensionless relative error E_dt is calculated by combining the characteristic scale L_char of the trajectory, i.e., E_dtw = normalized_dtw / L_char, where normalized_dtw represents the normalized DTW distance; D_total represents the total DTW distance with the minimum cumulative matching cost between the two sequences; and L_path represents the length of the optimal matching path.

[0029] Step S6.4: Convert the dimensionless relative error into a comprehensive source tracing score S, where S = 1 / (1 + E_dtw); through a preset mapping function, make the comprehensive source tracing score proportional to the spatiotemporal similarity, that is, the smaller the relative error, the higher the similarity, and the higher the comprehensive source tracing score, which serves as the final quantitative basis for determining the degree of suspicion of the suspected vessel's emissions.

[0030] Step S7: Sort all suspected vessels by their comprehensive source tracing scores and determine the source of the oil spill as the vessel with the highest score.

[0031] An oil spill tracing system based on spatiotemporal mapping and dynamic backtracking is disclosed. This system implements the aforementioned oil spill tracing method based on spatiotemporal mapping and dynamic backtracking. The system includes a data acquisition module, a preprocessing and skeleton labeling module, a suspect screening module, a spatiotemporal labeling module, a backtracking reconstruction module, a similarity calculation module, and a source determination module. Specifically:

[0032] The data acquisition module is used to acquire remote sensing images containing the actual oil spill area, ship AIS data, and marine environmental data.

[0033] The preprocessing and skeleton calibration module is used to preprocess the remote sensing image and determine the central skeleton point set of the oil spill area.

[0034] The suspect screening module is used to screen out suspect vessels based on the imaging time of the remote sensing image, the location information of the actual oil spill area, and the AIS track data.

[0035] The spatiotemporal calibration module is used to perform spatiotemporal mapping calibration for each suspected vessel in order to determine the reference emission time of each central skeleton point.

[0036] The backtracking reconstruction module is used to dynamically backtrack the central skeleton point set based on the baseline emission time in order to reconstruct the simulated sewage discharge path.

[0037] The similarity calculation module is used to calculate the spatiotemporal similarity between the simulated sewage discharge path and the real AIS flight track, and to obtain a comprehensive source tracing score.

[0038] The source tracing module is used to sort all suspected vessels based on the comprehensive source tracing score to determine the source of the oil spill.

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] (1) High fidelity of the source tracing mechanism, improving the physical authenticity of the source tracing: The spatiotemporal mapping calibration mechanism proposed in this invention couples the spatial geometric features of the oil spill morphology with the AIS track of the suspect vessel, providing a differentiated time scale with clear physical meaning for reverse tracing. This method overcomes the problem of physical process distortion caused by the lack of basis for setting the tracing time in traditional models, fundamentally improving the accuracy of the source tracing results.

[0041] (2) The simulation strategy is scientific and rigorous, enhancing the anti-interference capability: By constructing a simulation cluster and injecting uncertainties in environmental parameters such as wind, current, and diffusion, this invention upgrades single trajectory backtracking to probabilistic backtracking based on statistical distribution, which can effectively resist interference caused by marine environmental observation errors. At the same time, Savitzky-Golay filtering and spline interpolation techniques are used to smooth the reconstructed path, eliminating artifacts generated by discrete simulation and making the simulated path more consistent with the physical continuity of ship navigation.

[0042] (3) The quantitative assessment is intuitive and objective, realizing probabilistic representation: The kernel density estimation (KDE) algorithm is introduced to generate a heatmap of the source tracing probability, which transforms the complex particle distribution into an intuitive probability field, providing a scientific basis for evidence collection. At the same time, the dynamic time warping (DTW) algorithm is used to quantitatively assess the spatiotemporal similarity between the simulated path and the real trajectory, and a normalization method based on the trajectory physical scale is introduced in an original way. This method transforms the absolute DTW distance into a dimensionless relative error, effectively eliminating the evaluation bias caused by the difference in trajectory spatial scale in different source tracing events, so that the source tracing score has the comparability across scenarios.

[0043] In summary, this invention combines innovative spatiotemporal mapping calibration, high-fidelity dynamic asynchronous backtracking, and an objective and robust spatiotemporal quantitative evaluation system, significantly improving the automation level, accuracy, and reliability of mobile source oil spill tracing. Attached Figure Description

[0044] Figure 1 This is a flowchart of an oil spill tracing method based on spatiotemporal mapping and dynamic backtracking provided in an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram illustrating the screening of suspected vessels based on spatiotemporal constraints according to an embodiment of the present invention;

[0046] Figure 3 This is a schematic diagram of the source tracing simulation results for different suspected vessels provided by an embodiment of the present invention;

[0047] Figure 4 This is a schematic diagram of the structure of an oil spill tracing system based on spatiotemporal mapping and dynamic backtracking provided in an embodiment of the present invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0049] This invention proposes an oil spill source tracing method and system based on spatiotemporal mapping and dynamic backtracking. Figure 1 This document demonstrates the complete process of an oil spill source tracing method based on spatiotemporal mapping and dynamic backtracking, provided by an embodiment of the present invention. See also... Figure 1 Specifically, it includes the following steps:

[0050] Step S1: Obtain remote sensing images containing the actual oil spill area, AIS track data from the Automatic Identification System (AIS), and marine environmental data.

[0051] Specifically, remote sensing imagery, such as synthetic aperture radar (SAR) imagery or hyperspectral imagery, is an effective data source for large-scale marine oil spill monitoring. AIS data provides dynamic information such as vessel identification, real-time location, speed, and heading, serving as the basis for screening suspect vessels and obtaining their actual tracks. Marine environmental data mainly includes wind and current field data covering the target sea area within a specific time period, used to drive subsequent reverse-engineering physical models. In one embodiment, sea surface wind field data can be derived from the ERA5 reanalysis dataset of the European Centre for Medium-Range Weather Forecasts (ECMWF), while ocean current field data can be derived from high-resolution regional ocean model products, such as the 800-meter resolution model data covering nearshore areas provided by the Norwegian Meteorological Service.

[0052] Step S2 involves preprocessing the remote sensing image and identifying the central skeleton point set of the actual oil spill area; specifically:

[0053] Step S2.1, preprocessing the original remote sensing image includes geometric correction and image filtering.

[0054] Since the oil film regions in SAR images are irregular in shape and may contain artifacts (such as biofilms and low-wind-speed areas), direct region matching will introduce uncertainty.

[0055] Step S2.2: Through the human-computer interaction interface, the center line of the real oil spill area on the preprocessed remote sensing image is manually calibrated to generate a central skeleton that can represent the direction and core shape of the oil spill.

[0056] Step S2.3 involves discretizing the central skeleton at equal intervals to obtain a central skeleton point set P = {p1, p2, ..., pn}, where n is the total number of sampling points, p1 is the first central skeleton point, p2 is the second central skeleton point, and pn is the nth central skeleton point. This process simplifies the two-dimensional planar region into a one-dimensional linear geometric feature, providing a high-precision spatial reference for subsequent spatiotemporal mapping calibration.

[0057] S3. Based on the imaging time of the remote sensing image and the location information of the actual oil spill area, and combined with the AIS tracking data, suspect vessels are screened out; specifically including:

[0058] Step S3.1: Based on the imaging time T0 of the remote sensing image, set a time threshold forward, for example, T = 24 hours, to construct a time filtering window of [T0-T, T0], where T represents the length of the ship filtering time; at the same time, based on the center coordinates of the actual oil spill area, set a coverage radius, for example, R = 10 nautical miles, to construct a spatial filtering area.

[0059] Step S3.2: Retrieve the AIS database, extract the list of ships that pass through the spatial filtering area within the time filtering window, and obtain their MMSI (ship identification code) and corresponding latitude, longitude, heading and timestamp data.

[0060] Step S3.3: Select vessels whose heading is consistent with the extension direction of the central frame and identify them as suspect vessels.

[0061] In this embodiment, this step eliminates a large number of irrelevant vessels through spatiotemporal constraints. The historical AIS tracks of suspected vessels must meet the following requirements:

[0062] Spatial proximity: Its historical flight path overlaps or is adjacent to the central skeleton in geographic space.

[0063] Heading consistency: Its direction of travel roughly coincides with the overall extension direction of the central frame.

[0064] Time-leading: The timestamps of the track points passing through the area must be earlier than the imaging time of the remote sensing image and fall within a reasonable time window based on oil spill drift estimation.

[0065] Figure 2 A schematic diagram is shown illustrating the screening of suspected vessels based on AIS information and oil spill instances. (Refer to...) Figure 2 The dark striped areas represent the actual oil spill areas in the remote sensing image. The green tracks represent the historical tracks of two vessels, labeled as candidate suspect vessels A and B, respectively. Because their tracks both meet the aforementioned spatiotemporal constraints, vessels A and B are identified as suspect vessels in this source tracing investigation.

[0066] Step S4: For each suspected vessel, perform spatiotemporal mapping calibration to determine the reference emission time for each point in the central skeleton point set; specifically:

[0067] Step S4.1: Map the endpoints of the central skeleton of the oil spill area to the AIS track of the suspected vessel to obtain the first mapping point and the second mapping point; compare the timestamps of the two mapping points, take the mapping point with the earliest timestamp as the unique spatiotemporal anchor point, and determine the timestamp as the start time of the pollution discharge.

[0068] Step S4.2: Obtain the total length of the central skeleton. Starting from the spatiotemporal anchor point, project the total length along the AIS track to find the corresponding projection endpoint. Determine the time corresponding to the projection endpoint on the AIS track as the sewage discharge termination time.

[0069] Step S4.3: Based on the discharge start time and discharge end time, perform linear time interpolation calculation on each skeleton point in the central skeleton point set, thereby assigning a corresponding baseline discharge time to each skeleton point.

[0070] In one specific embodiment, the spatiotemporal mapping calibration is a space-time coupled calibration method, the core of which is:

[0071] Spatiotemporal anchor point determination: The endpoints of the central skeleton are mapped to the AIS track of the suspected vessel, and the matching point with the earlier timestamp is used as the spatiotemporal anchor point, and its timestamp is determined as the sewage discharge start time T_start.

[0072] Path length projection: Calculate the total length L of the central skeleton. Then, project the path length along the AIS track of the suspected vessel from the spatiotemporal anchor point. When the cumulative curve length of the projected path equals L, the time corresponding to its endpoint is determined as the discharge termination time T_end.

[0073] Linear time interpolation: for any point p on the central skeleton i Its baseline emission time t i Determined using the following linear interpolation formula:

[0074] (1)

[0075] Among them, l i For point p i The cumulative length along the central skeleton path from the starting endpoint. This mechanism directly and linearly maps the spatial geometric length of the oil spill morphology to the dynamic time span of ship navigation, providing a physically grounded time scale for subsequent differentiated retrospective analysis.

[0076] S5. Based on the baseline discharge time obtained for the suspected vessel in step S4, and combined with the marine environmental data from step S1, dynamically backtrack the central skeleton point set with differentiated durations to reconstruct a simulated discharge path corresponding to the suspected vessel; specifically:

[0077] Step S5.1: Each skeleton point in the central skeleton point set is set as an independent virtual particle, and the spatial coordinates of each particle at the time of remote sensing image imaging are used as the initial position for its reverse backtracking. To improve statistical reliability, multiple sub-particles are randomly generated within a preset radius with each skeleton point as the center to construct a simulated cluster.

[0078] Specifically, the initialization of the simulated cluster is represented as follows: for each skeleton point pi, k sub-particles are generated to form a particle swarm, and the coordinates of all sub-particles at the initial time T0 are distributed in a circular domain with pi as the center and a preset radius R as the radius.

[0079] Step S5.2: For each virtual particle, calculate its simulated backtracking duration independently based on the baseline emission time of its corresponding skeleton point; the simulated backtracking duration is defined as the time difference between the remote sensing image imaging time and the baseline emission time corresponding to the particle, thereby constructing differentiated time step constraints for each particle.

[0080] In one specific embodiment, the dynamic reverse backtracking is a model based on Lagrange particle tracking, the core of which is:

[0081] Independent particle source setting: For each point p on the central skeleton... i Consider it as the termination position of an independent Lagrange particle i.

[0082] Differential backtracking duration calculation: For each particle i (whose termination position is p) i Its differentiated backtracking duration ΔT starts from the uniform remote sensing image imaging time T0. i It was identified as:

[0083] ΔT i = T0 - t i (2)

[0084] Among them, t i This is the baseline emission time for that point as determined in step S4.

[0085] Step S5.3: Combining marine environmental data, control all virtual particles to synchronously start reverse motion from a unified imaging moment; during the backtracking process, execute an uncertainty injection mechanism: through Monte Carlo sampling, randomly assign differentiated horizontal diffusion coefficients, wind deflection factors, and ocean current field disturbance parameters to each particle within a preset range; each particle stops moving and locks its emission source point when it reaches the corresponding historical emission moment, according to its own simulation backtracking time.

[0086] Asynchronous stopping backtracking: All particles start from their respective termination positions p at the target time T0. i The particle begins its reverse motion synchronously, and its trajectory is solved by time stepping. Within each time step Δt, the particle's position update follows the inverse process of the following kinematic equations:

[0087] (3)

[0088] Where, x i (t-Δt) The updated particle position vector is represented by Δt, which is the computation time step during the solution process; x i (t) C is the position vector of particle i at time t; i (t) and W i(t) t represents the ocean current and wind speed vectors at the particle's location at time t; Cc and Cw are drift coefficients, with Cc being the ocean current correction coefficient and Cw being the wind deflection coefficient; Ud is the random turbulence diffusion term; and t represents the current time in the simulation backtracking calculation process.

[0089] In this embodiment, Cc, Cw, and Ud are the injected dynamic uncertainty parameters, and their value rules are as follows:

[0090] (1) Ocean current uncertainty coefficient Cc: random sampling for each particle in the range of [0.02, 0.05] to simulate random bias in ocean current observation data.

[0091] (2) Wind deflection coefficient Cw: Randomly sampled for each particle in the range of [0.02, 0.04] to characterize the differences in wind-driven oil film properties.

[0092] (3) Random turbulent diffusion term Ud: calculated based on the horizontal diffusion coefficient Ddiff, where Ddiff is randomly sampled in the range of [0.01, 0.1] m2 / s. By assigning different parameter combinations to each particle in the cluster, the backtracking results cover the real physical uncertainty range.

[0093] Each particle i experienced its own differentiated backtracking time ΔT i Afterward, the particles reach their historical emission source and cease movement. The collection of all historical source points after the particles cease movement together constitutes the simulated discharge path of the suspected vessel.

[0094] Step S5.4: Summarize the coordinates of all emission source points after all virtual particles stop moving, and use the Savitzky-Golay filter to denoise the particle swarm centroid sequence corresponding to each skeleton point to reconstruct a simulated discharge path that reflects the real physical process of continuous discharge of pollutants by the suspected vessel in the corresponding waterway segment.

[0095] Step S6: By calculating the spatiotemporal similarity between the simulated sewage discharge path corresponding to the suspected vessel and the actual AIS track of the suspected vessel, the comprehensive source tracing score of the suspected vessel is obtained, wherein the comprehensive source tracing score is proportional to the spatiotemporal similarity; specifically.

[0096] Step S6.1: Using the kernel density estimation (KDE) algorithm, the spatial distribution of particles after all backtracking stops is processed by fieldization to generate a heat map of the probability distribution of the source, so as to intuitively represent the high-probability core area of ​​the emission source.

[0097] Step S6.2: Using the Dynamic Time Warping (DTW) algorithm, calculate the cumulative distance between the smooth simulated sewage discharge path reconstructed in step S5 and the actual AIS track of the suspected vessel; automatically match the corresponding point pairs between the two paths using the DTW algorithm to effectively handle the local spatiotemporal stretching or compression caused by changes in vessel speed, and quantify the consistency of the two paths in motion patterns.

[0098] Step S6.3: The DTW distance obtained in step S6.2 is normalized by combining the trajectory length of the simulated path and the spatial scale of the study area to eliminate the influence of different dimensions and path length differences, thereby obtaining a dimensionless relative error index that can reflect the degree of matching between the two paths.

[0099] Step S6.4: Convert the dimensionless relative error into a comprehensive source tracing score; through a preset mapping function, make the comprehensive source tracing score proportional to the spatiotemporal similarity (i.e., the smaller the relative error, the higher the similarity, and the higher the comprehensive source tracing score), and use this as the final quantitative basis for determining the degree of suspicion of the suspected vessel's emissions.

[0100] In a preferred embodiment, to obtain a more objective and cross-scenario comparable comprehensive score, this invention proposes a comprehensive source tracing method based on Dynamic Time Warping (DTW) algorithm and physical feature normalization. Its core calculation steps are as follows:

[0101] Step S6.4.1, Calculate the trajectory feature scale: First, calculate the geospatial bounding box of the suspect vessel's actual AIS track, and determine the diagonal length of this bounding box as the trajectory feature scale L_char. This scale objectively characterizes the spatial scale of this source tracing event.

[0102] Step S6.4.2, calculate the normalized DTW distance: The simulated sewage discharge path X = (x1, ..., x...) is... n The actual AIS track Y = (y1, ..., y) is compared with the actual track Y = (y1, ..., y). m These are considered as two spatiotemporal sequences.

[0103] First, this invention employs the Dynamic Time Warping (DTW) algorithm to find the optimal matching path between two spatiotemporal sequences through dynamic programming. The algorithm calculates and accumulates the spatiotemporal distances between all matching pairs along this optimal matching path, thus obtaining a total DTW distance, denoted as D_total, representing the minimum cumulative matching cost between the two sequences.

[0104] Secondly, after finding the optimal matching path, the total number of matching point pairs contained in the path is obtained. This number is defined as the length of the optimal matching path, denoted as L_path.

[0105] To eliminate the impact of differences in the number of trajectory points and the length of the matching path, the total DTW distance D_total is divided by the length L_path of the optimal matching path to obtain the normalized DTW distance normalized_dtw, which represents the average matching cost per step on the optimal matching path.

[0106] normalized_dtw = D_total / L_path (5)

[0107] Step S6.4.3, calculate the relative error and overall score:

[0108] Dividing the normalized DTW distance (normalized_dtw) by the feature scale (L_char) yields a dimensionless relative average error (E_dtw):

[0109] E_dtw = normalized_dtw / L_char (6)

[0110] The relative average error E_dtw is inversely proportional to the matching degree between the simulated sewage discharge path and the actual AIS track. To obtain a comprehensive source tracing score S that is proportional to the matching degree, the relative average error E_dtw is transformed by applying a decreasing function. A specific embodiment of the decreasing function is as follows:

[0111] S = 1 / (1 + E_dtw) (7)

[0112] The higher the score S, the better the match between the simulated sewage discharge path and the actual AIS track, and the greater the likelihood that the suspected vessel is the source of pollution.

[0113] Step S7: Sort the comprehensive source tracing scores of all suspected vessels and determine the source of the oil spill as the vessel with the highest score.

[0114] Figure 3 A schematic diagram of the source determination result based on spatiotemporal mapping and dynamic backtracking is shown. The diagram intuitively compares the results after performing steps S5 and S6 on two different suspected vessels (suspect vessel A and suspect vessel B).

[0115] Figure (a) on the left illustrates the source tracing process for suspected vessel A. According to step S4, the central skeleton point set representing the actual oil spill pattern (blue hollow dots in the figure) is spatiotemporally mapped to the actual AIS track of suspected vessel A (green dashed line in the figure). Then, according to step S5, dynamic reverse tracing is performed, and a source tracing probability distribution heatmap (red area in the figure) is generated through step S6.1, with its center line representing the simulated discharge path. As can be seen from the figure, the core area of ​​this source tracing probability distribution heatmap highly overlaps spatially with the central skeleton point set representing the actual oil spill.

[0116] Figure (b) on the right illustrates the source tracing process for suspect vessel B. Using the same procedure, the central skeleton point set of the same oil spill area (blue hollow dots in the figure) was spatiotemporally mapped and dynamically traced back to the actual AIS track of suspect vessel B (green dashed line in the figure). As can be seen from the figure, the generated source probability distribution heatmap (red area in the figure) has a significant spatial deviation from the central skeleton point set representing the actual oil spill.

[0117] To conduct a quantitative assessment, this embodiment executes step S6, calculating that the comprehensive source tracing score of suspect vessel A(a) is a relatively high value (e.g., 0.90), while the comprehensive source tracing score of suspect vessel B(b) is a relatively low value (e.g., 0.55). Based on the determination rules in step S7, since the score of suspect vessel A is significantly higher than that of suspect vessel B, suspect vessel A is ultimately determined to be the source of the oil spill in this incident.

[0118] Figure 4 This is a schematic diagram of an oil spill tracing system based on spatiotemporal mapping and dynamic backtracking, provided by an embodiment of the present invention. This system is used to execute the methods described in the above embodiments, such as... Figure 4 As shown, the system includes the following modules:

[0119] The data acquisition module is used to acquire remote sensing images containing the actual oil spill area, ship AIS data, and marine environmental data.

[0120] The preprocessing and skeleton calibration module is used to preprocess the remote sensing image and determine the central skeleton point set of the oil spill area.

[0121] The suspect screening module is used to screen out suspect vessels based on the imaging time of the remote sensing image, the location information of the actual oil spill area, and the AIS track data.

[0122] The spatiotemporal calibration module is used to perform spatiotemporal mapping calibration for each suspected vessel in order to determine the reference emission time of each central skeleton point.

[0123] The backtracking reconstruction module is used to dynamically backtrack the central skeleton point set based on the baseline emission time in order to reconstruct the simulated sewage discharge path.

[0124] The similarity calculation module is used to calculate the spatiotemporal similarity between the simulated sewage discharge path and the real AIS flight track, and to obtain a comprehensive source tracing score.

[0125] The source tracing module is used to sort all suspected vessels based on the comprehensive source tracing score to determine the source of the oil spill.

[0126] The above embodiments are merely illustrative of the implementation methods of the present invention, but should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the protection scope of the present invention.

Claims

1. A method for tracing oil spill sources based on spatiotemporal mapping and dynamic backtracking, characterized in that, The oil spill tracing method includes the following steps: Step S1: Acquire remote sensing images containing the actual oil spill area, AIS track data from the Automatic Identification System (AIS), and marine environmental data. Step S2: Preprocess the remote sensing image and mark the central skeleton point set of the actual oil spill area; Step S3: Based on the imaging time of the remote sensing image and the location information of the actual oil spill area, and combined with AIS track data, suspect vessels are screened out. Step S4: For each suspected vessel, perform spatiotemporal mapping calibration to determine the reference emission time for each point in the central skeleton point set; Step S5: Based on the baseline discharge time obtained for the current suspected vessel in step S4, and combined with the marine environmental data in step S1, perform dynamic reverse backtracking of the central skeleton point set with differentiated durations to reconstruct a simulated sewage discharge path corresponding to the suspected vessel. Step S6: By calculating the spatiotemporal similarity between the simulated sewage discharge path corresponding to the suspected vessel and the actual AIS track of the suspected vessel, the comprehensive source tracing score of the suspected vessel is obtained, wherein the comprehensive source tracing score is proportional to the spatiotemporal similarity. Step S7: Sort all suspected vessels by their comprehensive source tracing scores and determine the source of the oil spill as the vessel with the highest score.

2. The oil spill source tracing method based on spatiotemporal mapping and dynamic backtracking according to claim 1, characterized in that, The specific steps of S2 are as follows: Step S2.1 involves preprocessing the remote sensing image, including geometric correction and image filtering; Step S2.2: Manually calibrate the actual oil spill area on the preprocessed remote sensing image to obtain a central skeleton representing the direction of the main oil spill. Step S2.3: Discretize the central skeleton at equal intervals to obtain a central skeleton point set P consisting of several discrete points, where P={p1,p2,...,pn}, where n is the total number of sampling points, p1 is the first central skeleton point, p2 is the second central skeleton point, and pn is the nth central skeleton point.

3. The oil spill source tracing method based on spatiotemporal mapping and dynamic backtracking according to claim 2, characterized in that, The specific steps of step S3 are as follows: Step S3.1: Based on the imaging time T0 of the remote sensing image, a time threshold is set forward to construct a time filtering window of [T0-T, T0], where T represents the length of the ship filtering time; at the same time, a spatial filtering area is constructed by setting the coverage radius R with the center coordinates of the actual oil spill area as the center. Step S3.2: Search the AIS database, extract the list of ships that have passed through the spatial filtering area within the time filtering window, and obtain their ship identification code MMSI and corresponding latitude, longitude, heading and timestamp data; Step S3.3: Select vessels whose heading is consistent with the extension direction of the central frame and identify them as suspect vessels.

4. The oil spill source tracing method based on spatiotemporal mapping and dynamic backtracking according to claim 3, characterized in that, The specific steps of S4 are as follows: Step S4.1: Map the endpoints of the central skeleton of the oil spill area to the AIS track of the suspected vessel to obtain the first mapping point and the second mapping point; compare the timestamps of the two mapping points, take the mapping point with the earliest timestamp as the unique spatiotemporal anchor point, and determine the discharge start time as T_start; Step S4.2: Obtain the total length L of the central skeleton. Starting from the spatiotemporal anchor point determined in step S4.1, project the total length along the AIS track to find the corresponding projection endpoint. Determine the time corresponding to the projection endpoint on the AIS track as the sewage discharge termination time, and define it as T_end. Step S4.3: Based on the sewage discharge start time and sewage discharge end time, perform linear time interpolation calculations on each skeleton point in the central skeleton point set, thereby calculating the value for each skeleton point p. i Assign the corresponding baseline emission time t i The calculation formula is as follows: .

5. The oil spill source tracing method based on spatiotemporal mapping and dynamic backtracking according to claim 4, characterized in that, The specific steps of step S5 are as follows: Step S5.1: Each skeleton point in the central skeleton point set is set as an independent particle, and the spatial coordinates of each particle at the time of remote sensing image imaging are used as the initial position for its reverse backtracking. Multiple sub-particles are randomly generated within a preset radius with each skeleton point as the center to construct a simulated cluster. Step S5.2: For each particle, calculate the simulated backtracking duration based on the baseline emission time of its corresponding skeleton point; the simulated backtracking duration ΔT i Defined as the distance between the remote sensing image imaging time T0 and the corresponding baseline emission time t of the particle. i The time difference between them, i.e. ΔT i = T0 - t i To construct differentiated time step constraints for each particle; Step S5.3: Combining marine environmental data, control all particles to synchronously start reverse motion from a unified imaging moment to update particle positions; during the backtracking process, execute an uncertainty injection mechanism: through Monte Carlo sampling, randomly assign differentiated horizontal diffusion coefficients, wind deflection factors, and ocean current field disturbance parameters to each particle within a preset range; each particle stops moving and locks its emission source point when it reaches the corresponding historical emission moment according to its own simulated backtracking time. Step S5.4: Summarize the coordinates of all emission source points after all particles stop moving, denoise the particle swarm centroid sequence corresponding to each skeleton point, and reconstruct a simulated discharge path that reflects the real physical process of continuous discharge of pollutants by the suspected vessel in the corresponding waterway segment.

6. The oil spill source tracing method based on spatiotemporal mapping and dynamic backtracking according to claim 5, characterized in that, In step 5.3, the particle position update follows the inverse process of the following kinematic equations: , where x i (t-Δt) x represents the updated particle position vector. i (t) Ci represents the position vector of particle i at time t; Cc represents the ocean current correction coefficient; (t) The vector representing the ocean current at time t; Cw represents the wind deflection coefficient; Wi (t) Δt represents the wind speed vector at the particle's location at time t; Δt represents the calculation time step during the solution process; Ud represents the random turbulent diffusion term; and t represents the current time in the simulation backtracking calculation process.

7. The oil spill source tracing method based on spatiotemporal mapping and dynamic backtracking according to claim 5, characterized in that, The specific steps of step S6 are as follows: Step S6.1: Using the kernel density estimation algorithm, the spatial distribution of all particles after the backtracking stops is processed by fieldization to generate a heat map of the source probability distribution, which characterizes the high-probability core region of the emission source. Step S6.2: Using the dynamic time warping algorithm, calculate the cumulative distance between the smooth simulated sewage discharge path reconstructed in step S5 and the actual AIS track of the suspected vessel; use the DTW algorithm to automatically match the corresponding point pairs between the two paths, quantify the consistency of the two paths in motion mode, and obtain the DTW distance. Step S6.3: The DTW distance obtained in step S6.2 is normalized by combining the trajectory length of the simulated path and the spatial scale of the study area. The normalization calculation formula is: normalized_dtw = D_total / L_path. The dimensionless relative error E_dt is calculated by combining the characteristic scale L_char of the trajectory, i.e., E_dtw = normalized_dtw / L_char; where normalized_dtw represents the normalized DTW distance; D_total represents the total DTW distance with the minimum cumulative matching cost between the two sequences; and L_path represents the length of the optimal matching path. Step S6.4: Convert the dimensionless relative error into a comprehensive source tracing score S, where S = 1 / (1 + E_dtw); through a preset mapping function, make the comprehensive source tracing score proportional to the spatiotemporal similarity, that is, the smaller the relative error, the higher the similarity, and the higher the comprehensive source tracing score, so as to serve as the final quantitative basis for determining the degree of suspicion of the suspected ship's emissions.

8. An oil spill tracing system based on spatiotemporal mapping and dynamic backtracking, characterized in that, The oil spill tracing system described herein implements the oil spill tracing method based on spatiotemporal mapping and dynamic backtracking as described in any one of claims 1-7. The oil spill tracing system includes a data acquisition module, a preprocessing and skeleton calibration module, a suspect screening module, a spatiotemporal calibration module, a backtracking reconstruction module, a similarity calculation module, and a tracing determination module.

9. An oil spill tracing system based on spatiotemporal mapping and dynamic backtracking as described in claim 8, characterized in that, The oil spill tracing system is specifically as follows: The data acquisition module is used to acquire remote sensing images containing the actual oil spill area, ship AIS data, and marine environmental data. The preprocessing and skeleton calibration module is used to preprocess the remote sensing image and determine the central skeleton point set of the oil spill area. The suspect screening module is used to screen out suspect vessels based on the imaging time of remote sensing images, the location information of the actual oil spill area, and AIS track data. The spatiotemporal calibration module is used to perform spatiotemporal mapping calibration for each suspected vessel and determine the reference emission time of each central skeleton point. The backtracking reconstruction module is used to dynamically backtrack the central skeleton point set based on the baseline emission time and reconstruct the simulated sewage discharge path. The similarity calculation module is used to calculate the spatiotemporal similarity between the simulated sewage discharge path and the real AIS flight track, and to obtain a comprehensive source tracing score. The source tracing module is used to sort all suspected vessels based on the comprehensive source tracing score to determine the source of the oil spill.

Images