A deep learning-based dynamic flood risk map intelligent drawing method

Abstract

The present application relates to the technical field of image rendering, and particularly relates to a dynamic flood risk map intelligent rendering method based on deep learning. The method comprises the following steps: acquiring real-time hydrological and radar rainfall data of a target area, constructing a physically constrained flood frequency dynamic prediction model to generate an initial risk map, combining spatial inundation information extracted from remote sensing images to analyze inundation propagation relationship, confirming flood risk evolution state, finally optimizing the initial risk map to form a dynamic flood risk map. The present application combines physically constrained deep learning prediction with remote sensing inundation propagation evolution analysis, realizes accurate perception and dynamic intelligent rendering of the spatio-temporal evolution process of flood risk, thereby significantly improving the prediction reliability, spatial consistency and actual decision support capability of the flood risk map.

Description

Technical Field

[0001] This invention relates to the field of image rendering technology, and in particular to a method for intelligent rendering of dynamic flood risk maps based on deep learning. Background Technology

[0002] Flood risk maps are a crucial foundation for urban flood control scheduling, emergency response, and disaster prevention decision-making. They visually represent the potential inundation area and the evolution of risk, providing a scientific basis for timely protective measures. However, existing flood risk mapping methods primarily rely on three types of technologies: first, numerical hydrological models that calculate river peak flood levels and inundation areas using historical rainfall and topographic data; second, statistical analysis of historical flood data to construct flood frequency curves and static risk maps; and third, analysis using remote sensing imagery and GIS overlay to extract inundation areas. Early methods largely depended on single models or manual experience. While the introduction of remote sensing and GIS has enabled a more visual representation of inundation areas, these methods remain at the static analysis stage, failing to reflect the spatial expansion of floods over time. Furthermore, the lack of a systematic characterization of propagation paths and evolution patterns between geographical units can easily lead to delayed or misjudged risk assessments. Summary of the Invention

[0003] Therefore, it is necessary to provide a deep learning-based intelligent method for drawing dynamic flood risk maps to solve at least one of the aforementioned technical problems.

[0004] To achieve the above objectives, a method for intelligently drawing dynamic flood risk maps based on deep learning is proposed, the method comprising the following steps: Step S1: Acquire real-time hydrological monitoring data and radar rainfall data for the target area; Step S2: Physical constraints are embedded using real-time hydrological monitoring data and radar rainfall data to construct a dynamic flood frequency prediction model; the initial predicted risk map of the target area is output using the dynamic flood frequency prediction model. Step S3: Acquire remote sensing images of the target area; Based on the spatial inundation information of the remote sensing images and the inundation propagation relationship with the initial predicted risk map, confirm the flood risk evolution status of the target area; Step S4: Optimize the initial predicted risk map based on the evolution of flood risk to obtain a dynamic flood risk map.

[0005] This application has the following beneficial effects: By introducing a technical approach combining deep learning with physical mechanism constraints, the flood risk map has been transformed from "static prediction" to "dynamic evolution perception and intelligent mapping." On the one hand, by utilizing real-time hydrological monitoring data and radar rainfall data, a dynamic flood frequency prediction model is constructed under the embedding of flow velocity-water depth physical constraints, effectively improving the stability and reliability of initial risk predictions under extreme rainfall conditions. On the other hand, by integrating spatial inundation information extracted from remote sensing imagery, and based on the inundation propagation relationships and propagation path locations between geographical units, the spatial transmission, diffusion, and attenuation processes of floods are meticulously depicted, thereby achieving dynamic confirmation and correction of the flood risk evolution status. By introducing the propagation structure division of leading, intermediate, and terminal locations, and combining change markers and inundation response discrimination mechanisms, the implicit propagation bias and unmanifested diffusion risks during flood evolution can be identified, significantly reducing the probability of misjudgment caused by a single moment or a single data source. While ensuring physical rationality, the overall method enhances the ability of flood risk maps to express temporal changes and spatial propagation characteristics, making the generated dynamic flood risk maps superior to existing technologies in terms of accuracy, continuity, and interpretability. It is suitable for various application scenarios such as urban flood control scheduling, risk early warning, and emergency decision-making. Attached Figure Description

[0006] Figure 1 This is a flowchart illustrating the steps of a deep learning-based intelligent method for drawing dynamic flood risk maps. Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3. Figure 3 This is the initial predicted risk map for the intelligent drawing method of dynamic flood risk map based on deep learning proposed in this application; Figure 4 This application presents a dynamic flood risk map based on a deep learning-based intelligent drawing method for dynamic flood risk maps. The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0007] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0008] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0009] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0010] To achieve the above objectives, please refer to Figures 1 to 4 A method for intelligently drawing dynamic flood risk maps based on deep learning, the method comprising the following steps: Step S1: Acquire real-time hydrological monitoring data and radar rainfall data for the target area; In one embodiment, several hydrological monitoring stations are pre-established within the target area to monitor the hydrological information of the target area in real time. These hydrological monitoring stations may include river level monitoring stations, flow monitoring sections, and rainfall monitoring stations. Each hydrological monitoring station collects water level data, flow data, and rainfall data according to a preset sampling period and uploads the collected data to a data processing unit.

[0011] In some embodiments, the real-time hydrological monitoring data may include water level and flow rates at different times, which may be divided into minute-level or hour-level intervals. The data processing unit may organize the received real-time hydrological monitoring data in chronological order to obtain hydrological monitoring time-series data for the target area.

[0012] For example, when a monitoring station fails to upload valid hydrological data at the current time, the data processing unit can use the hydrological data from the previous time or data from adjacent monitoring stations for compensation processing to ensure the continuity of hydrological monitoring data.

[0013] In some embodiments, radar rainfall data covering the target area is acquired via a meteorological radar system. This radar rainfall data can be used to characterize the rainfall intensity distribution in the target area at different times. The radar rainfall data can be divided according to a preset spatial resolution and correlated with the geographical location of the target area.

[0014] In some embodiments, the data processing unit can perform time alignment processing on the radar rainfall data to ensure that the radar rainfall data and real-time hydrological monitoring data are consistent in the time dimension, which facilitates subsequent joint analysis.

[0015] For example, when the temporal resolution of radar rainfall data is higher than that of hydrological monitoring data, time aggregation processing can be performed on the radar rainfall data; when the temporal resolution of radar rainfall data is lower than that of hydrological monitoring data, time can be filled in by interpolation.

[0016] Step S2: Physical constraints are embedded using real-time hydrological monitoring data and radar rainfall data to construct a dynamic flood frequency prediction model; the initial predicted risk map of the target area is output using the dynamic flood frequency prediction model. In one embodiment, after acquiring the real-time hydrological monitoring data and radar rainfall data in step S1, the data processing unit performs joint processing on the two types of data to construct a basic dataset for flood prediction.

[0017] In some embodiments, the physical constraint embedding is used to constrain the basic hydrological laws in the flood evolution process. These basic hydrological laws may include the relationship between rainfall and runoff generation, the relationship between water level and discharge, and continuity constraints during the confluence process. The data processing unit can perform consistency verification on real-time hydrological monitoring data and radar rainfall data based on the aforementioned hydrological laws.

[0018] For example, when the rainfall intensity represented by radar rainfall data is inconsistent with the water level change trend at the corresponding time, the data processing unit can adjust the weight or mark the abnormal data according to the preset physical constraint rules in order to avoid the abnormal data having a significant impact on the subsequent prediction results.

[0019] In some embodiments, after completing the physical constraint embedding, the data processing unit constructs a flood frequency dynamic prediction model based on the processed real-time hydrological monitoring data and radar rainfall data. The flood frequency dynamic prediction model is used to characterize the frequency variation features of floods occurring in the target area at different times.

[0020] In some embodiments, the flood frequency dynamic prediction model can be initialized based on historical hydrological data and dynamically updated by combining real-time hydrological monitoring data and radar rainfall data at the current moment, so that the model output can reflect the current hydrological status changes in the target area.

[0021] For example, when the target area detects a continuous increase in rainfall intensity and a faster rate of water level rise at multiple consecutive moments, the flood frequency dynamic prediction model can correspondingly increase the predicted frequency of floods occurring in the target area at subsequent moments.

[0022] In some embodiments, based on the output of the flood frequency dynamic prediction model, the data processing unit generates an initial predicted risk map of the target area. The initial predicted risk map is used to characterize the flood risk distribution at different locations within the target area during the current prediction period.

[0023] In some embodiments, the initial predicted risk map can be divided according to a preset spatial resolution, with different spatial units corresponding to different flood risk levels; the flood risk level can be determined based on the flood frequency prediction results.

[0024] For example, when the flood frequency prediction value corresponding to a certain spatial unit is higher than a preset threshold, the spatial unit can be marked as a high-risk area; when the flood frequency prediction value is lower than the threshold, the spatial unit can be marked as a medium-risk or low-risk area.

[0025] In another embodiment, reference can be made to Figure 3 The initial predicted risk map is generated and displayed in the form of a 3D topographic map. The overall 3D model of the initial predicted risk map is constructed based on the digital elevation topography of the target area, including continuous undulating surface elevation units to express topographic features such as mountains and hills. Obvious water bodies or potential flooding areas are superimposed in the low-lying areas to represent the spatial distribution of rivers, water bodies and easily flooded areas. At the same time, different colors or color levels are used to render the surface units to reflect the differences in flood risk intensity calculated by the flood frequency dynamic prediction model, so that the 3D model can comprehensively present the topographic features, water body distribution and initial flood risk spatial pattern within the same spatial framework.

[0026] Step S3: Acquire remote sensing images of the target area; Based on the spatial inundation information of the remote sensing images and the inundation propagation relationship with the initial predicted risk map, confirm the flood risk evolution status of the target area; In one embodiment, the data processing unit acquires remote sensing image data covering the target area. The remote sensing image may include optical remote sensing image or synthetic aperture radar remote sensing image, used to characterize the surface state of the target area at the current moment.

[0027] In some embodiments, the remote sensing images are acquired at preset time intervals and aligned with the initial predicted risk map generated in step S2 on a spatial coordinate system, so that the remote sensing images and the initial predicted risk map can be compared and analyzed under the same spatial reference.

[0028] In some embodiments, the data processing unit extracts spatial flooding information of the target area based on remote sensing images. The spatial flooding information is used to characterize the actual flooding range and its spatial distribution at different times within the target area.

[0029] For example, when a significant change in the reflection or backscattering characteristics of a continuous area is detected in a remote sensing image, the corresponding area can be identified as an area with surface water accumulation or flooding, thus forming spatial flooding information.

[0030] In some embodiments, the data processing unit performs correlation analysis between the spatial flooding information and the initial predicted risk map to determine the propagation direction and expansion trend of the spatial flooding information within the target area.

[0031] In some embodiments, the flooding propagation relationship is used to characterize the expansion, contraction, or maintenance of the flooded area between adjacent times, and the propagation relationship can be determined based on the degree of spatial overlap of spatial flooding information at adjacent times.

[0032] For example, when the spatial extent of a flooded area is larger in a later time than in a previous time, and the direction of expansion is consistent with the spatial distribution direction of high-risk areas in the initial predicted risk map, it can be determined that the flooded area is in a state of expansion and propagation.

[0033] For example, when the spatial inundation information remains largely consistent between adjacent time points and the risk level of the corresponding area in the initial predicted risk map does not change significantly, it can be determined that the flood risk of the target area is in a stable evolution state.

[0034] In some embodiments, the data processing unit determines the flood risk evolution status of the target area based on the inundation propagation relationship between spatial inundation information and the initial predicted risk map. The flood risk evolution status may include an enhanced, weakened, or stable state.

[0035] Step S4: Optimize the initial predicted risk map based on the evolution of flood risk to obtain a dynamic flood risk map.

[0036] In one embodiment, the data processing unit obtains the flood risk evolution status confirmed in step S3, and uses the flood risk evolution status as the basis for regulation and control for risk map drawing optimization.

[0037] In some embodiments, the drawing optimization is used to adjust the risk level distribution, spatial boundary morphology, or risk change magnitude in the initial predicted risk map, so that the risk map can reflect the dynamic change process of flood risk over time.

[0038] In some embodiments, when the flood risk evolution state is in an enhanced state, the data processing unit may adjust the risk level of the corresponding area in the initial predicted risk map upward or expand the spatial range of the high-risk area.

[0039] For example, when the flood risk evolution state is in a weakening state, the data processing unit can downgrade the risk level of the corresponding area in the initial predicted risk map, or reduce the spatial range of the high-risk area.

[0040] In some embodiments, when the flood risk evolution state is stable, the data processing unit may keep the risk level distribution of the initial predicted risk map unchanged and only update the time stamp of the risk map.

[0041] In some embodiments, when performing drawing optimization, the data processing unit may smooth the risk level changes between adjacent spatial units based on preset spatial continuity rules to avoid unreasonable abrupt boundaries in the risk map.

[0042] For example, when the risk level difference between adjacent spatial units exceeds a preset threshold, the risk level can be transitioned to make the dynamic flood risk map more continuous in spatial distribution.

[0043] In some embodiments, the optimized risk map is determined as a dynamic flood risk map, which is used to characterize the spatial distribution and evolution trend of flood risk in the target area during the current forecast period.

[0044] In some embodiments, the dynamic flood risk map can be updated at preset time intervals and adjusted in real time as the flood risk evolves to support subsequent risk warnings or decision support.

[0045] In another embodiment, reference can be made to Figure 4After obtaining the flood risk evolution status confirmed in step S3, the data processing unit further introduces topographic zoning information of the target area as an auxiliary constraint for drawing optimization. This topographic zoning information includes, but is not limited to, the water areas, mid-altitude mountains, low-altitude hills, and high-altitude mountain ranges marked on the map. Different colored lines represent the flood risk levels corresponding to different spatial units within the target area, with the line colors progressively changing from low to high risk. Specifically, during the drawing optimization process, the data processing unit differentiates the risk level adjustment strategy within different topographic zones based on the flood risk evolution status. When the flood risk evolution status is in an enhanced state, for spatial units near water areas or located near mid-altitude mountain confluence channels, the data processing unit can prioritize increasing the risk level color of their corresponding lines and appropriately expand the coverage of high-risk lines along the river network or confluence path direction to reflect the evolutionary characteristics of flood diffusion towards low-lying areas along topographic channels. Simultaneously, for low-altitude hilly areas, the colors of adjacent lines can be enhanced in a coordinated manner based on their spatial connectivity with water areas.

[0046] When the flood risk evolution state is in a weakening state, the data processing unit can lower the risk level color of the lines corresponding to the high-altitude mountain areas and the medium-high potential areas far from water in the map, and gradually shrink the distribution range of high-risk lines in the low-altitude hilly areas, so as to reflect the spatial characteristics of risk receding from low-lying areas to high-potential areas during the flood receding process.

[0047] When the flood risk evolution is in a stable state, the data processing unit can keep the risk level distribution corresponding to the line colors in each topographic zone of the map basically unchanged, and only update the time marker or evolution stage marker of the dynamic flood risk map in combination with the current prediction time, so that the risk map can continuously reflect the spatial risk pattern under the stable stage.

[0048] Furthermore, in this embodiment, when adjusting line colors and risk levels, the data processing unit can also combine preset spatial continuity rules to perform risk level transition processing on adjacent spatial units that cross different terrain zones (such as water areas and low-altitude hills, medium-altitude mountains and high-altitude mountain ranges). When the risk level difference corresponding to adjacent line colors exceeds a preset threshold, transition lines are generated through interpolation or gradation to ensure that the dynamic flood risk map maintains visual and risk expression continuity at terrain boundaries, avoiding abrupt boundaries that do not conform to the actual flood propagation pattern.

[0049] Ultimately, the risk map generated after the above-mentioned drawing and optimization process was determined to be a dynamic flood risk map. This dynamic flood risk map can comprehensively reflect the spatial evolution characteristics of different terrain units, water distribution, and flood risk level lines over time, and can be continuously updated according to preset time intervals to support flood risk early warning and disaster prevention and mitigation auxiliary decision-making.

[0050] As an example of the present invention, reference is made to Figure 2 As shown, step S3 in this example includes: Step S31: Acquire remote sensing images of the target area, and perform geometric correction and spatial registration on the remote sensing images to obtain corrected remote sensing images; Step S32: Extract spatial inundation information of the target area based on the corrected remote sensing image to obtain the inundation distribution results reflecting the inundation status of each geographic unit; Step S33: Spatially correlate the inundation distribution results with the initial predicted risk map and analyze the inundation propagation relationship between each geographic unit; Step S34: Based on the inundation propagation relationship, confirm the flood risk evolution status of the target area at the current moment.

[0051] In one embodiment, based on the corrected remote sensing image, the surface features of each geographic unit within the target area are identified, and the corresponding spatial inundation information is extracted according to the changes in surface reflection or backscattering features in the image, thereby obtaining the inundation distribution result reflecting the inundation state of each geographic unit. The extraction of inundation information from the remote sensing image can employ image interpretation or feature extraction methods known to those skilled in the art, and will not be elaborated upon here.

[0052] In another embodiment, the inundation distribution results are spatially correlated with the initial predicted risk map, so that the inundation state of each geographic unit is correlated with its risk level in the initial predicted risk map, and the inundation propagation relationship between each geographic unit is analyzed based on the changes in the inundation state between adjacent geographic units.

[0053] Inundation propagation relationships are used to characterize the direction and trend of inundation spread in the spatial dimension, reflecting the propagation path and intensity of floods within a target area. For example, when a geographic unit changes from an uninundated state to an inundated state, and its neighboring geographic units were already in an inundated state at the previous moment, it can be determined that the inundation state of the geographic unit was formed by the propagation from the neighboring geographic units.

[0054] In this embodiment of the invention, based on the inundation propagation relationship, the change in the inundation range within the target area at the current time relative to the previous time is comprehensively judged, thereby confirming the flood risk evolution state of the target area at the current time. The flood risk evolution state may include a risk enhancement state, a risk reduction state, or a risk stability state.

[0055] Preferably, step S34 includes: Based on the inundation propagation relationship, the relative positional relationship of each geographic unit in the target area is extracted in the propagation relationship; Based on their relative positions, the location of each geographical unit along the propagation path during flood propagation is determined; By combining the propagation path location corresponding to each geographical unit with the spatial inundation information at the current moment, the flood evolution trend data of the target area can be confirmed. Based on flood evolution data, the flood risk evolution status of the target area at the current moment is confirmed.

[0056] In one embodiment, based on the flood propagation relationship obtained in step S33, the relative positional relationship of each geographic unit within the target area in the propagation relationship is extracted. The relative positional relationship is used to characterize the sequential order and spatial adjacency of each geographic unit in the direction of flood propagation.

[0057] In another embodiment, the relative positional relationship can be determined by analyzing the spatial adjacency relationship between each geographic unit and the temporal sequence of changes in flooding status. For example, when the first geographic unit is flooded at a previous moment, and the second geographic unit adjacent to it is flooded at a later moment, it can be determined that the first geographic unit is located upstream of the second geographic unit.

[0058] In another embodiment, the propagation path position of each geographical unit in the flood propagation is determined based on the relative positional relationship. The propagation path position is used to characterize the node position or hierarchical position of each geographical unit in the overall flood propagation path.

[0059] The location of propagation paths can reflect the main and secondary propagation channels of floods within a target area, thus distinguishing the degree of influence of different geographical units in the flood propagation process. For example, changes in the inundation state of geographical units located on the main propagation path have a more significant indicative significance for the overall flood evolution.

[0060] In another embodiment, the propagation path location corresponding to each geographical unit is combined with the spatial inundation information at the current moment to form flood evolution trend data for the target area. This flood evolution trend data is used to comprehensively reflect the changes in the spatial expansion direction, propagation intensity, and coverage area of ​​the flood.

[0061] In this embodiment of the invention, the flood evolution trend data can be obtained by summarizing and analyzing the inundation status of geographical units at different propagation paths, so as to characterize whether the flood is in a rapid expansion stage, a slow expansion stage, or a gradual receding stage at the current moment.

[0062] In another embodiment, based on the flood evolution data, the flood risk evolution status of the target area at the current moment is determined. This flood risk evolution status can be used to characterize the changing trend of flood risk over time.

[0063] For example, when flood evolution data shows that multiple geographic units along the main propagation path change from an uninundated state to an inundated state at the current moment, the flood risk evolution state can be confirmed as an enhanced state; when the number of inundated geographic units along the propagation path decreases, the flood risk evolution state can be confirmed as a weakened state; when the inundation state along the propagation path remains basically unchanged, the flood risk evolution state can be confirmed as a stable state.

[0064] By introducing a combined analysis of propagation path location and spatial inundation information, a structured characterization of the flood propagation process can be achieved. This enables the confirmation of the flood risk evolution state to be based not only on single-point inundation information, but also on a comprehensive consideration of the flood propagation path and the overall evolution trend, thereby improving the accuracy and reliability of flood risk evolution judgment.

[0065] Preferably, the propagation path location corresponding to each geographical unit is combined with the spatial inundation information at the current moment to confirm the flood evolution trend data of the target area, including: Based on the inundation propagation relationship, each geographic unit within the target area is identified as the preceding position, intermediate position, and terminal position in the corresponding propagation path; Using the geographical unit at the middle position as the state reference, its current spatial inundation information is compared side-by-side with the spatial inundation information at the corresponding preceding and ending positions. Based on the juxtaposition and comparison results, the directional characteristics of spatial flooding information changes between the leading, intermediate, and terminal positions are identified. By analyzing the changing bias characteristics, we can confirm the flood evolution trend data of the target area at the current moment.

[0066] In one embodiment, based on the flood propagation relationship determined in step S33, the location of each geographic unit within the target area is identified along its corresponding propagation path, classifying them into leading positions, intermediate positions, and terminal positions. The location identification is used to characterize the relative hierarchical relationship of each geographic unit in the direction of flood propagation.

[0067] In another embodiment, the leading position is used to characterize the geographical unit that the flood first reaches or is first affected by, the intermediate position is used to characterize the geographical unit located in the middle of the propagation path that has a receiving role in the upstream and downstream inundation changes, and the terminal position is used to characterize the geographical unit that is subsequently affected or responds with a lag during the propagation of the flood.

[0068] In another embodiment, the geographic unit at the middle position is used as the state comparison benchmark, and its spatial flooding information at the current time is compared with the spatial flooding information of the corresponding preceding and ending positions to form a flooding state comparison result between multiple positions.

[0069] Juxtaposition and comparison are used to eliminate the interference caused by the fluctuation of the inundation state of a single geographic unit, so that the judgment of inundation changes is based on the relative change relationship within the propagation path, rather than isolated spatial single-point information.

[0070] In another embodiment, based on the juxtaposition comparison results, the variation bias characteristics of spatial inundation information between the leading, intermediate, and terminal positions are identified. These variation bias characteristics are used to characterize the changing trend of inundation degree along the propagation path.

[0071] For example, when the flooding level at the leading position is higher than that at the middle position, and the flooding level at the middle position is higher than that at the terminal position, the change bias feature can be identified as forward enhancement type; when the flooding level at the terminal position gradually becomes higher than that at the middle position, the change bias feature can be identified as backward expansion type.

[0072] In another embodiment, the flood evolution trend data corresponding to the target area at the current moment is confirmed by the aforementioned change bias characteristics. The flood evolution trend data is used to comprehensively reflect the expansion direction and evolution intensity of the flood along its spatial propagation path.

[0073] Flood evolution data can be used to distinguish whether a flood is in a rapid expansion phase, a steady propagation phase, or a gradual attenuation phase, providing a basis for confirming the subsequent evolution of flood risk.

[0074] By introducing a juxtaposition and comparison mechanism that uses the intermediate position as the reference, the confirmation of flood evolution status data is based on the internal structural relationship of the propagation path, avoiding the risk of misjudgment caused by single-point judgment, thereby improving the stability and reliability of flood evolution status identification.

[0075] Preferably, based on the juxtaposition and comparison results, the characteristics of spatial flooding information variation bias between the leading, intermediate, and terminal positions include: Based on the comparison results, the spatial flooding information records corresponding to the preceding position, intermediate position and the terminal position at the current time are obtained respectively; By comparing the recorded values ​​of spatial flooding information over time, it can be determined whether the flooding status of each location has changed relative to the previous moment. When only one location experiences a change in flooding status, that location is designated as the location of the changed load-bearing capacity. When the flooding state changes at at least two locations, determine whether the changed bearing location has shifted from the preceding location to the intermediate location or from the intermediate location to the end location; Based on the shift of changing load-bearing locations, the changing bias characteristics of spatial inundation information between the initial, intermediate, and final locations are identified.

[0076] In one embodiment, spatial inundation information of the target area within the same time period is extracted and analyzed by juxtaposing and comparing the results. During the extraction process, the target area is divided into three typical spatial locations: the leading position, the intermediate position, and the terminal position, and the inundation information record value of each location at the current time is obtained. The inundation information record value may include cell water depth, water accumulation area, or inundation status identifier, which are used to characterize the water coverage of each location under the two-dimensional inundation model.

[0077] Subsequently, the recorded flooding information values ​​for each location were analyzed in chronological order. By comparing the flooding status of each location at the current moment with that at the previous moment, it was determined whether a change in the flooding status had occurred at each location. Specifically, when the flooding status of a location changes from dry to waterlogged or from waterlogged to dry, it is determined that a change in the flooding status has occurred at that location.

[0078] Based on the determination of changes in inundation status, the location of the change in bearing capacity is further identified. When only a single location experiences a change in inundation status, that location is directly identified as the location of the change in bearing capacity. However, when at least two locations experience changes in inundation status simultaneously, it is necessary to further analyze the direction of the change in bearing capacity, i.e., to determine whether the change has shifted from the initial location to the intermediate location, or from the intermediate location to the final location, thereby reflecting the spatial development trend of inundation expansion or contraction.

[0079] Finally, based on the changing location and its transfer, the directional characteristics of spatial inundation information across the preceding, intermediate, and terminal locations can be identified. These directional characteristics reflect the migration trend of floodwaters within the target area, providing data support for urban flood forecasting, emergency dispatching, and risk analysis. For example, when the directional characteristics show a continuous transfer from the preceding to the terminal location, it means that the floodwaters are continuously advancing along a predetermined channel, and the inundation risk in downstream areas should be a primary focus.

[0080] Preferably, when only one location experiences a change in flooding status, after determining that location as the location of the changed bearing capacity, the method further includes: Obtain the spatial flooding information change marker corresponding to the location of the change; Search for the presence of a flooding response corresponding to the change marker at the middle and end positions of the propagation path; When a change marker does not generate a corresponding response in the middle or end position, the change marker is determined to be an unpropagated change. When a change marker generates a corresponding response at an intermediate or terminal position, the change marker is identified as a propagated change. Based on the propagation status of the change markers, the evolution of flood risk in the target area at the current moment is corrected and confirmed.

[0081] In one embodiment, when a comparison analysis determines that only one spatial location has experienced a change in flooding status, that location is identified as the changed bearing capacity location. Subsequently, a spatial flooding information change marker corresponding to the changed bearing capacity location is obtained. This marker may include the changed flooding depth, water area, or flooding status change type, and is used to uniquely identify the flooding change event at that location.

[0082] Next, in the propagation path of the target area, the spatial flooding information at the intermediate and terminal positions is searched to see if there is a flooding response corresponding to the change marker. Specifically, by comparing the flooding information records at the intermediate and terminal positions at the current time with the change marker characteristics, it is determined whether the change shows continuous or similar flooding state changes along the propagation path.

[0083] If the search results show that the change marker does not generate a corresponding response at the intermediate or terminal location, the change marker is determined to be a non-propagating change, meaning that the flooding event is confined to the initial location and has not yet spread downstream or to adjacent areas. Conversely, if the change marker generates a corresponding response at the intermediate or terminal location, the change marker is determined to be a propagating change, indicating that the flood has spread downstream or to adjacent areas along a specific path.

[0084] Finally, based on the propagation status of the change markers, non-propagated and propagated changes are applied to correct the flood risk evolution trend in the target area. In the case of propagated changes, the risk level prediction for downstream areas can be appropriately increased; in the case of non-propagated changes, the risk prediction for the corresponding location is maintained or decreased, thereby achieving dynamic correction and confirmation of the urban flood risk evolution trend.

[0085] Preferably, the flooding response behavior corresponding to the change marker specifically includes: Before the change marker is generated, obtain the spatial flooding information corresponding to the middle and end positions, and record the predetermined expression method of the spatial flooding information; At the current moment, based on the established expression method, the spatial flooding information corresponding to the middle and end positions is expressed again; Determine whether the spatial flooding information at the current moment can be fully expressed through the established expression method; if the spatial flooding information at the current moment cannot be fully expressed through the established expression method, determine that the location forms a flooding response to the change marker; if the spatial flooding information at the current moment can be fully expressed through the established expression method, determine that the location does not form a flooding response to the change marker.

[0086] In one embodiment, regarding the flood response performance corresponding to the change marker, spatial flood information corresponding to the middle and end positions in the target area is first obtained before the change marker is generated, and this spatial flood information is recorded according to a predetermined expression method. The predetermined expression method may include flood depth distribution, water accumulation area ratio, or flood state encoding, etc., to characterize the flood characteristics of the location.

[0087] Subsequently, at the current moment, based on the aforementioned established expression method, the spatial flooding information at the intermediate and terminal positions is expressed again to obtain a description of the spatial flooding characteristics at the current moment. Specifically, the current flooding data can be mapped to an indicator system of the established expression method to determine whether the data is fully presented within the range or pattern allowed by the expression method.

[0088] Then, it is determined whether the spatial inundation information at the current moment can be fully expressed using the established representation method. If it cannot be fully expressed, that is, if the current inundation state or water depth change exceeds the characterization range of the established representation method, then it is determined that the location has formed an inundation response to the change marker, indicating that the inundation state has undergone a identifiable change along the propagation path. Conversely, if the spatial inundation information at the current moment can be fully expressed using the established representation method, then it is determined that the location has not formed an inundation response to the change marker, that is, no identifiable inundation change has occurred.

[0089] Preferably, determining the propagation path location of each geographical unit during flood propagation based on relative positional relationships includes: Obtain the spatial location parameters of each geographic unit and construct a set of relative positional relationships between any geographic unit and its neighboring geographic units; Based on the set of relative positional relationships, the relative order of each geographical unit in the direction of flood propagation is determined; Using any geographic unit as a reference unit, determine the trend of its relative ranking status among adjacent geographic units; When the trend of change shows expansion in one direction, the geographic unit is identified as the leading position; when the trend of change shows bidirectional correlation, the geographic unit is identified as the intermediate position; when the trend of change shows no correlation, the geographic unit is identified as the ending position.

[0090] In one embodiment, the spatial location parameters of each geographic unit within the target area are first obtained, including the unit center coordinates, boundary coordinates, and adjacency relationships. Based on these parameters, a set of relative positional relationships between any geographic unit and its neighboring geographic units is constructed. This set of relative positional relationships can be used to describe the spatial relationships between units in the direction of water flow propagation, such as upstream and downstream relationships, lateral adjacency relationships, and potential hydraulic connection paths.

[0091] Subsequently, based on the set of relative positional relationships, the relative order of each geographical unit in the direction of flood propagation is analyzed. That is, by comparing the relative elevation, drainage sequence or adjacent connection path of each unit, the front-back and left-right order characteristics of each unit in the process of flood propagation are determined.

[0092] Next, using any geographic unit as a reference unit, the trend of its relative ranking status among adjacent geographic units is determined. Specifically, by performing trend analysis on the spatial sequence of adjacent units, the direction and range of flood expansion from the reference unit to adjacent units are determined, and the propagation pattern characteristics of the units are extracted.

[0093] Furthermore, when the trend of change is a unidirectional expansion, that is, the flood mainly spreads continuously in one direction, the geographical unit is identified as the leading position; when the trend of change is a bidirectional correlation, that is, the flood spreads simultaneously in two main directions, the geographical unit is identified as the intermediate position; when the trend of change is uncorrelated, that is, the flood spread does not significantly extend to neighboring units, the geographical unit is identified as the terminal position.

[0094] Preferably, the correction and confirmation of the flood risk evolution trend of the target area at the current moment based on the propagation status of the change markers includes: Based on the propagation status of the change markers, random points are selected to analyze the flood risk evolution of the target area at the current moment, resulting in several risk evolution data points. Based on the changes in flood risk at the risk evolution data points, determine the range of flood risk changes at each data point. The extent of flood risk impact at current risk evolution data points is determined by the range of changes in flood risk. The maximum impact point within the range of flood risk changes is identified by using the flood risk impact degree of several risk evolution data points; The flood risk evolution trend of the target area at the current moment is corrected and confirmed based on the point of greatest impact.

[0095] In one embodiment, firstly, based on the propagation status of the change markers, random points are selected to assess the flood risk evolution of the target area at the current moment. This random point selection involves discretely sampling different locations within the target area's spatial range, taking into account the distribution density and propagation direction of the change markers, to avoid risk assessment bias caused by relying on a single location, thereby obtaining several risk evolution data points.

[0096] In some embodiments, for each acquired risk evolution data point, information on the flood risk level, water depth change trend, or risk category change at the current and adjacent times is extracted, and the flood risk change of each risk evolution data point is determined based on this information. On this basis, the changes in the risk evolution data points are summarized to confirm the spatial and temporal range of flood risk changes for each risk evolution data point.

[0097] In some embodiments, based on the confirmed range of flood risk changes, the degree of risk enhancement or reduction at each risk evolution data point during the change process is further analyzed to determine the degree of flood risk impact corresponding to each risk evolution data point. The degree of flood risk impact is used to characterize the strength of the contribution of that point to the spread or mitigation of flood risk in the surrounding area.

[0098] In some embodiments, a comprehensive comparative analysis of the flood risk impact of several risk evolution data points is performed to select the points that have the most significant impact on the overall risk evolution within the range of flood risk changes, as the points of maximum impact. These points of maximum impact are typically located in areas where risk changes are most drastic or where change markers are most concentrated.

[0099] In some embodiments, the flood risk evolution of the target area at the current moment is corrected and confirmed based on the determined maximum impact point. Specifically, the risk change characteristics of the maximum impact point are used as a correction reference to adjust the original flood risk evolution of the target area, so that the corrected flood risk evolution is more consistent with the actual propagation state of the change marker, thereby improving the accuracy and reliability of flood risk assessment and early warning results.

[0100] Preferably, the correction and confirmation of the flood risk evolution trend of the target area at the current moment based on the point of maximum impact includes: Identify the flood risk evolution path of the target area based on the discrete distance of the point of greatest impact; When the flood risk evolution path is inconsistent with the risk propagation direction of the flood risk evolution situation at the current moment, the flood risk evolution path is used to confirm the direction angle convergence correction of the risk propagation direction of the flood risk evolution situation. When the flood risk evolution path is consistent with the risk propagation direction of the flood risk evolution status at the current moment, the direction of the flood risk evolution status of the target area at the current moment is maintained and confirmed.

[0101] In one embodiment, the discrete distance distribution between the point of maximum impact and other risk units within the target area is first calculated based on the spatial location of the point of maximum impact within the target area. By analyzing the discrete distances from nearest to farthest, the main propagation directions and preferred propagation channels of flood risk spreading outward from the point of maximum impact are identified, thereby determining the flood risk evolution path of the target area. This flood risk evolution path is used to characterize the dominant spatial expansion trend of flood risk.

[0102] In some embodiments, the determined flood risk evolution path is compared and analyzed with the existing risk propagation direction in the current flood risk evolution situation. When a deviation is found between the propagation direction indicated by the flood risk evolution path and the risk propagation direction in the current flood risk evolution situation, it is determined that there is a directional error in the current risk propagation direction. At this time, based on the flood risk evolution path, the original risk propagation direction is gradually adjusted to converge the risk propagation direction to the direction indicated by the flood risk evolution path, thereby completing the correction and confirmation of the directional angle and avoiding abrupt changes or reverse shifts in the risk propagation direction.

[0103] In some embodiments, when the flood risk evolution path is substantially consistent with or within the allowable directional deviation range of the current flood risk evolution situation, the current flood risk evolution situation is considered to accurately reflect the risk propagation process. In this case, a directional maintenance constraint is imposed on the flood risk evolution situation of the target area at the current moment to ensure that the risk propagation direction remains stable and continuous in subsequent moments, preventing unnecessary fluctuations in the risk propagation direction due to local disturbances or noisy data.

[0104] By using the above methods, the risk propagation direction of the flood risk evolution trend in the target area can be modified or constrained by the point of greatest impact, so that the flood risk evolution trend is more in line with the actual propagation law in the spatial direction, and the stability and credibility of the flood risk evolution analysis results can be improved.

[0105] Of particular importance is determining the relative order of each geographic unit in the direction of flood propagation, based on the set of relative positional relationships: Select any geographic unit as the reference unit and extract its relative positional relationship with adjacent geographic units; Based on the relative positional relationship, the initial sorting state of the baseline unit relative to its adjacent geographic units is generated; Using the initial sorting state as a reference, the sorting state of the baseline unit is sequentially passed to its adjacent geographic units to obtain the corresponding associated sorting state. Determine whether the associated sorting status remains consistent during the transmission process; when the sorting status remains consistent, confirm that the sorting status is a stable sorting status and use it as the relative sorting status of the geographic unit in the direction of flood propagation.

[0106] In one embodiment, taking a low-lying, flood-prone area of ​​a city as the research object, the target area is divided into several regular or irregular geographic units according to a uniform spatial resolution. Each geographic unit corresponds to a surface grid or sub-catchment unit. Based on elevation data, river course, and surface runoff direction, a set of relative positional relationships containing adjacency and relative orientation information between geographic units is pre-constructed to describe the upstream / downstream, front / back, or high / low relationships of each geographic unit in space.

[0107] In this embodiment, a geographic unit is first selected from the geographic units as a reference unit, for example, a geographic unit located near the main river channel is selected as the reference unit. By querying the set of relative positional relationships, the relative positional relationships between the reference unit and its neighboring geographic units are extracted, including relative elevation difference, slope aspect consistency, and runoff orientation relationship, thereby forming a spatial comparison result between the reference unit and its neighboring geographic units.

[0108] Subsequently, based on the extracted relative positional relationships, the order of the benchmark unit relative to its adjacent geographic units in the direction of flood propagation is determined, generating the initial sorting state of the benchmark unit. For example, when the elevation of the benchmark unit is lower than that of its adjacent geographic units and it is located downstream in the direction of runoff convergence, its sorting state is marked as "subsequent unit"; otherwise, it is marked as "preceding unit", thus obtaining the initial sorting state with the benchmark unit as the core.

[0109] Based on this, the initial sorting state is used as a reference, and according to the adjacency relationship between geographic units, the sorting state is progressively passed to other geographic units adjacent to the base unit, and further extended to more distant related geographic units, forming multiple related sorting states. During the transmission process, the consistency between the newly generated related sorting states and the existing sorting states is continuously compared to determine the stability of the sorting logic during spatial propagation.

[0110] When a certain ordering state remains consistent throughout its transmission among multiple geographic units without any contradictions or cyclical conflicts, it is considered a stable ordering state and is determined as the relative ordering state of the corresponding geographic unit in the direction of flood propagation. This method allows for the generation of geographic unit ordering results reflecting the flood propagation path and sequence relationships across the entire target area, providing reliable spatial order constraints for subsequent flood evolution simulation, risk propagation analysis, and dynamic flood forecasting.

[0111] Of particular importance is that, based on relative positional relationships, the initial sorting state of the baseline unit relative to its neighboring geographic units includes: Obtain the relative positional relationship parameters between the baseline unit and each adjacent geographic unit; For each adjacent geographic unit, determine whether the relative positional relationship parameters meet the preset sorting criteria; Adjacent geographic units that meet the sorting criteria are marked as sorting front units, and adjacent geographic units that do not meet the sorting criteria are marked as sorting back units. Based on the distribution results of the pre-sorting and post-sorting units, the initial sorting state of the baseline unit relative to its adjacent geographic units is generated.

[0112] In one embodiment, a typical flood-prone area of ​​a city is selected as the study area. The target area is divided into multiple adjacent geographical units, each corresponding to a surface grid unit or sub-catchment. For each geographical unit, a set of relative positional relationship parameters, including spatial adjacency, relative elevation, and runoff orientation, is pre-established to characterize the spatial correlation characteristics between different geographical units during flood propagation.

[0113] In this embodiment, any one of the geographic units is first selected as a reference unit. By querying the relative positional relationship parameter set, the relative positional relationship parameters between the reference unit and its neighboring geographic units are obtained. The relative positional relationship parameters include at least the relative elevation difference, the surface slope aspect consistency index, and the runoff orientation relationship calculated based on the digital elevation model, which are used to comprehensively reflect the potential flow trend of floods during their propagation on the surface.

[0114] Subsequently, for each adjacent geographic unit, the relative positional relationship parameters are compared with preset sorting criteria. These criteria may include: whether the baseline unit is at a lower elevation relative to adjacent geographic units, whether runoff direction is from adjacent geographic units to the baseline unit, and whether there is a direct surface runoff channel between them. When at least one or more of these conditions are met, it is determined that the flood is more likely to converge from adjacent geographic units to the baseline unit during its propagation.

[0115] After the judgment is completed, adjacent geographical units that meet the ranking criteria are marked as ranking preceding units, i.e., geographical units located before the baseline unit in the direction of flood propagation; adjacent geographical units that do not meet the ranking criteria are marked as ranking following units, i.e., geographical units located after the baseline unit in the direction of flood propagation. Through this marking process, adjacent geographical units around the baseline unit can be initially classified according to the order of flood propagation.

[0116] Finally, based on the spatial distribution of the pre- and post-order units, an initial ordering state of the baseline unit relative to its adjacent geographical units is generated. This initial ordering state characterizes the flood propagation sequence position of the baseline unit within a local spatial range and serves as the fundamental constraint for subsequent ordering state transmission, stability assessment, and flood propagation path construction.

[0117] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0118] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A method for intelligently drawing dynamic flood risk maps based on deep learning, characterized in that, Includes the following steps: Step S1: Acquire real-time hydrological monitoring data and radar rainfall data for the target area; Step S2: Physical constraints are embedded using real-time hydrological monitoring data and radar rainfall data to construct a flood frequency dynamic prediction model; the initial predicted risk map of the target area is output using the flood frequency dynamic prediction model. Step S3: Acquire remote sensing images of the target area; Based on the spatial inundation information of the remote sensing images and the inundation propagation relationship with the initial predicted risk map, confirm the flood risk evolution status of the target area; Step S4: Optimize the initial predicted risk map based on the evolution of flood risk to obtain a dynamic flood risk map.

2. The method for intelligently drawing dynamic flood risk maps based on deep learning according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Acquire remote sensing images of the target area, and perform geometric correction and spatial registration on the remote sensing images to obtain corrected remote sensing images; Step S32: Extract spatial inundation information of the target area based on the corrected remote sensing image to obtain the inundation distribution results reflecting the inundation status of each geographic unit; Step S33: Spatially correlate the inundation distribution results with the initial predicted risk map and analyze the inundation propagation relationship between each geographic unit; Step S34: Based on the inundation propagation relationship, confirm the flood risk evolution status of the target area at the current moment.

3. The method for intelligently drawing dynamic flood risk maps based on deep learning according to claim 2, characterized in that, Step S34 includes: Based on the inundation propagation relationship, the relative positional relationship of each geographic unit in the target area is extracted in the propagation relationship; Based on their relative positions, the location of each geographical unit along the propagation path during flood propagation is determined; By combining the propagation path location corresponding to each geographical unit with the spatial inundation information at the current moment, the flood evolution trend data of the target area can be confirmed. Based on flood evolution data, the flood risk evolution status of the target area at the current moment is confirmed.

4. The method for intelligently drawing dynamic flood risk maps based on deep learning according to claim 3, characterized in that, By combining the propagation path location corresponding to each geographic unit with the spatial inundation information at the current moment, the flood evolution trend data of the target area is confirmed, including: Based on the inundation propagation relationship, each geographic unit within the target area is identified as the preceding position, intermediate position, and terminal position in the corresponding propagation path; Using the geographical unit at the middle position as the state reference, its current spatial inundation information is compared side-by-side with the spatial inundation information at the corresponding preceding and ending positions. Based on the juxtaposition and comparison results, the directional characteristics of spatial flooding information changes between the leading, intermediate, and terminal positions are identified. By analyzing the changing bias characteristics, we can confirm the flood evolution trend data of the target area at the current moment.

5. The method for intelligently drawing dynamic flood risk maps based on deep learning according to claim 4, characterized in that, Based on the juxtaposition and comparison results, the following characteristics were identified regarding the directional bias of spatial inundation information across the preceding, intermediate, and terminal locations: Based on the comparison results, the spatial flooding information records corresponding to the preceding position, intermediate position and the terminal position at the current time are obtained respectively; By comparing the recorded values ​​of spatial flooding information over time, it can be determined whether the flooding status of each location has changed relative to the previous moment. When only one location experiences a change in flooding status, that location is designated as the location of the changed load-bearing capacity. When the flooding state changes at at least two locations, determine whether the changed bearing location has shifted from the preceding location to the intermediate location or from the intermediate location to the end location; Based on the shift of changing load-bearing locations, the changing bias characteristics of spatial inundation information between the initial, intermediate, and final locations are identified.

6. The method for intelligently drawing dynamic flood risk maps based on deep learning according to claim 5, characterized in that, When only one location experiences a change in flooding status, after designating that location as the location of the changed bearing capacity, the following steps are also included: Obtain the spatial flooding information change marker corresponding to the location of the change; Search for the presence of a flooding response corresponding to the change marker at the middle and end positions of the propagation path; When a change marker does not generate a corresponding response in the middle or end position, the change marker is determined to be an unpropagated change. When a change marker generates a corresponding response at an intermediate or terminal position, the change marker is identified as a propagated change. Based on the propagation status of the change markers, the evolution of flood risk in the target area at the current moment is corrected and confirmed.

7. The method for intelligently drawing dynamic flood risk maps based on deep learning according to claim 6, characterized in that, The specific flood response behaviors corresponding to the change markers include: Before the change marker is generated, obtain the spatial flooding information corresponding to the middle and end positions, and record the predetermined expression method of the spatial flooding information; At the current moment, based on the established expression method, the spatial flooding information corresponding to the middle and end positions is expressed again; Determine whether the spatial flooding information at the current moment can be fully expressed through the established expression method; if the spatial flooding information at the current moment cannot be fully expressed through the established expression method, determine that the location forms a flooding response to the change marker; if the spatial flooding information at the current moment can be fully expressed through the established expression method, determine that the location does not form a flooding response to the change marker.

8. The method for intelligently drawing dynamic flood risk maps based on deep learning according to claim 3, characterized in that, Determining the propagation path location of each geographical unit during flood propagation based on relative location relationships includes: Obtain the spatial location parameters of each geographic unit and construct a set of relative positional relationships between any geographic unit and its neighboring geographic units; Based on the set of relative positional relationships, the relative order of each geographical unit in the direction of flood propagation is determined; Using any geographic unit as a reference unit, determine the trend of its relative ranking status among adjacent geographic units; When the trend of change shows expansion in one direction, the geographic unit is identified as the leading position; when the trend of change shows bidirectional correlation, the geographic unit is identified as the intermediate position; when the trend of change shows no correlation, the geographic unit is identified as the ending position.

9. The method for intelligently drawing dynamic flood risk maps based on deep learning according to claim 6, characterized in that, Based on the propagation status of the change markers, the correction and confirmation of the flood risk evolution trend of the target area at the current moment includes: Based on the propagation status of the change markers, random points are selected to analyze the flood risk evolution of the target area at the current moment, resulting in several risk evolution data points. Based on the changes in flood risk at the risk evolution data points, determine the range of flood risk changes at each data point. The extent of flood risk impact at current risk evolution data points is determined by the range of changes in flood risk. The maximum impact point within the range of flood risk changes is identified by using the flood risk impact degree of several risk evolution data points; The flood risk evolution trend of the target area at the current moment is corrected and confirmed based on the point of greatest impact.

10. The method for intelligently drawing dynamic flood risk maps based on deep learning according to claim 9, characterized in that, The correction and confirmation of the flood risk evolution trend of the target area at the current moment based on the point of maximum impact includes: Identify the flood risk evolution path of the target area based on the discrete distance of the point of greatest impact; When the flood risk evolution path is inconsistent with the risk propagation direction of the flood risk evolution situation at the current moment, the flood risk evolution path is used to confirm the direction angle convergence correction of the risk propagation direction of the flood risk evolution situation. When the flood risk evolution path is consistent with the risk propagation direction of the flood risk evolution status at the current moment, the direction of the flood risk evolution status of the target area at the current moment is maintained and confirmed.

Images