Multi-disaster multi-mode flood flow intelligent monitoring and early warning method, system and device

Abstract

The invention provides a multi-disaster multi-mode flood flow intelligent monitoring and early warning method, system and device, and relates to the field of disaster early warning. Establishing a rainstorm-runoff coupling model for predicting the surface runoff yield, the confluence time and the runoff peak value based on the soil water content, the vegetation coverage and the surface slope data under different rainfall capacities; constructing a river channel confluence coupling model based on the flow velocity, the water level and the section form of river channel section monitoring for predicting river channel flood confluence time, confluence amount and on-way water level change; judging whether there is a riverway confluence risk according to the real-time predicted riverway flood confluence time, confluence amount and on-way water level change; when the riverway confluence risk exists, the flood inundation range, the water depth distribution and the evolution speed are predicted through a pre-trained flood evolution coupling model based on the surface runoff, the confluence time and the change speed of the runoff peak value in combination with the landform data. According to the invention, the accuracy, timeliness and reliability of disaster early warning can be improved.

Description

Technical Field

[0001] This invention relates to the field of disaster early warning, and more specifically, to a method, system, and device for intelligent monitoring and early warning of multi-hazard, multi-modal floods. Background Technology

[0002] In the field of natural disaster monitoring and early warning, traditional monitoring methods often rely on equipment with limited functionality, and early warning systems lack comprehensive perception and intelligent early warning capabilities for floods and disaster chains. Currently, there is a need to propose a multi-hazard, multi-modal intelligent monitoring and early warning method, system, and device for floods. This aims to overcome the limitations of traditional monitoring and early warning systems, achieving comprehensive perception of all elements of floods and disaster chains, multi-modal early warning, and full-process response, thereby significantly improving the accuracy, timeliness, and reliability of disaster early warning. Summary of the Invention

[0003] The purpose of this invention is to provide a multi-hazard, multi-modal intelligent monitoring and early warning method, system, and device for floods. It can overcome the limitations of traditional monitoring and early warning systems, realize full-element perception of floods and disaster chains, multi-modal early warning, and full-process response, and significantly improve the accuracy, timeliness, and reliability of disaster early warning.

[0004] The embodiments of the present invention are implemented as follows:

[0005] This application provides a method for intelligent monitoring and early warning of multi-hazard, multi-modal floods, which includes the following steps:

[0006] A flood threshold early warning model is established based on the triggering of early warnings when meteorological, hydrological, and infrasound data collected by multiple sensors exceed preset thresholds.

[0007] AI scene early warning and monitoring of hydrological disaster videos, geological disaster videos, and urban disaster videos are carried out based on edge-side visual analysis algorithms, and an AI recognition model for flood videos is established.

[0008] When either the flood threshold early warning model or the flood video AI recognition model issues an early warning signal, a rainstorm-runoff coupling model is established based on soil moisture content, vegetation coverage and surface slope data under different rainfall amounts to predict surface runoff, confluence time and runoff peak.

[0009] Based on the flow velocity, water level and cross-sectional morphology of the river cross section, a river confluence coupling model is constructed to predict the confluence time, confluence volume and water level changes along the river.

[0010] The risk of river confluence is determined based on real-time predictions of the river flood confluence time, flow rate, and water level changes along the course.

[0011] When there is a risk of river confluence, the surface runoff, confluence time, and rate of change of peak runoff are obtained by applying the storm-runoff coupling model based on real-time rainfall.

[0012] Based on the changes in surface runoff, confluence time, and peak runoff rate, combined with topographic data, a pre-trained flood evolution coupling model is used to predict the flood inundation range, water depth distribution, and evolution rate.

[0013] Meteorological data collected by sensors includes rainfall and rainfall duration; hydrological data collected by sensors includes river level, flow rate, flow velocity, and soil moisture content.

[0014] The aforementioned storm-runoff coupling model, based on soil moisture content, vegetation cover, and surface slope data under different rainfall amounts, is used to predict surface runoff, confluence time, and peak runoff. Specifically, it includes:

[0015] Based on the spatiotemporal distribution data of soil moisture content, vegetation cover and surface slope, a multivariate nonlinear regression method was used to establish the response relationship between surface runoff, runoff time and peak runoff and rainfall, so as to construct a rainstorm-runoff coupling model.

[0016] The aforementioned river confluence coupling model, constructed based on river cross-section monitoring of flow velocity, water level, and cross-sectional morphology, is used to predict river flood confluence time, confluence volume, and water level changes along the river course. Specifically, it includes:

[0017] Using river cross-section monitoring data such as flow velocity, water level, and cross-sectional morphology as inputs, and river flood confluence time, confluence volume, and water level changes along the course as outputs, a flood confluence prediction model is constructed using machine learning methods.

[0018] The method of determining whether there is a risk of river confluence based on real-time predicted river flood confluence time, confluence volume, and changes in water level along the river course specifically includes:

[0019] By analyzing real-time river flood confluence time, flow rate, and water level changes along the river, and by determining whether the historical safety threshold of the river's carrying capacity is exceeded, it is possible to determine whether there is a risk of river confluence.

[0020] The method involves triggering an early warning based on meteorological, hydrological, and infrasound data collected from multiple sensors when they exceed preset thresholds, and establishing a flood threshold early warning model. Specifically, this includes setting graded thresholds for water level, flow rate, and rainfall for different hydrological year types and different watershed zones; and triggering an early warning signal and outputting the threshold early warning result when meteorological, hydrological, and infrasound data exceed the dynamic thresholds of the corresponding zones and year types.

[0021] The AI ​​scene monitoring of hydrological disaster videos, geological disaster videos, and urban disaster videos based on edge-side visual analysis algorithms, and the establishment of a flood video AI recognition model, specifically includes:

[0022] The YOLOv8 object detection algorithm was used to identify key indicators in videos of hydrological disasters, geological disasters, and urban disasters.

[0023] A CNN-LSTM hybrid model is used to perform time-series analysis on key indicators of hydrological disaster videos, geological disaster videos, and urban disaster videos to predict the changing trends of these key indicators.

[0024] When the changing trend of the identified indicators reaches the preset risk threshold, the AI ​​scenario early warning monitoring results are output.

[0025] A multi-hazard, multi-modal intelligent flood monitoring and early warning system, characterized in that it includes: a flood threshold early warning unit, a flood video AI recognition unit, and a disaster risk early warning unit;

[0026] The flood threshold early warning unit is used to trigger an early warning based on meteorological, hydrological and infrasound data collected by multiple sensors exceeding a preset threshold, and to establish a flood threshold early warning model.

[0027] The flood video AI recognition unit is used to perform AI scene early warning monitoring of hydrological disaster videos, geological disaster videos and urban disaster videos based on edge-side visual analysis algorithms, and to establish a flood video AI recognition model;

[0028] The disaster risk early warning unit is used to establish a rainstorm-runoff coupling model to predict surface runoff, confluence time and runoff peak based on soil moisture content, vegetation coverage and surface slope data under different rainfall amounts when either the flood threshold early warning model or the flood video AI recognition model issues an early warning signal.

[0029] Based on the flow velocity, water level and cross-sectional morphology of the river cross section, a river confluence coupling model is constructed to predict the confluence time, confluence volume and water level changes along the river.

[0030] The risk of river confluence is determined based on real-time predictions of the river flood confluence time, flow rate, and water level changes along the course.

[0031] When there is a risk of river confluence, the surface runoff, confluence time, and rate of change of peak runoff are obtained by applying the storm-runoff coupling model based on real-time rainfall.

[0032] Based on the changes in surface runoff, confluence time, and peak runoff rate, combined with topographic data, a pre-trained flood evolution coupling model is used to predict the flood inundation range, water depth distribution, and evolution rate.

[0033] A multimodal intelligent flood monitoring device includes an integrated pole; the integrated pole integrates a multi-hazard multimodal intelligent flood monitoring and early warning system, multiple sensors, a video monitoring system, a solar power supply system, and a waterproof speaker;

[0034] Multiple sensors are used to collect the meteorological, hydrological, or infrasound data;

[0035] The video surveillance system is used to collect videos of hydrological disasters, geological disasters, and urban disasters;

[0036] The waterproof speaker is used to provide voice alerts for the disaster risk warning results;

[0037] The solar power system is used to power multiple sensors, the video surveillance system, the waterproof speaker, and the multi-hazard, multi-modal flood intelligent monitoring and early warning system.

[0038] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects:

[0039] This application relates to a multi-hazard, multi-modal intelligent monitoring and early warning method for floods. This method integrates multiple monitoring technologies and models to achieve intelligent monitoring and early warning of various disasters. Based on meteorological, hydrological, and infrasound data collected from multiple sensors, preset thresholds are set. When the data exceeds these thresholds, an early warning is triggered, thus establishing a flood threshold early warning model to ensure the efficiency of meteorological, hydrological, and geological early warnings. Simultaneously, edge-side visual analysis algorithms are used to perform AI scene early warning monitoring on hydrological disaster videos, geological disaster videos, and urban disaster videos, thereby establishing a flood video AI recognition model, which can improve the accuracy of early warnings for hydrological, geological, and urban safety. When one of the flood threshold early warning model and the flood video AI recognition model issues an early warning, due to the complexity of sensor data sources and the limitations of video recognition scenarios, further measures are taken to improve the prediction accuracy of flood disasters. To improve accuracy, a storm-runoff coupling model is established based on soil moisture content, vegetation cover, and surface slope data under different rainfall amounts to predict runoff changes caused by storms. A river confluence coupling model is constructed based on river cross-section velocity, water level, and cross-sectional morphology to predict river flood confluence processes and assess river confluence risk. When a risk is identified, the storm-runoff coupling model is used to obtain the surface runoff, confluence time, and rate of change of peak runoff from real-time rainfall. Combined with topographic data, a pre-trained flood evolution coupling model is used to predict the flood inundation range, water depth distribution, and evolution rate, thereby simulating the flood propagation path and impact range using storm-runoff changes and topography. This application not only optimizes the accuracy of threshold models but also significantly improves the latency of video AI recognition. Furthermore, it monitors whether heavy rainfall will further cause flooding. By assessing river channel confluence risk based on river cross-sections and predicting flood risk based on changes in current rainfall, and when flood risk is present, it monitors the flood evolution process by combining rainfall-related runoff data and topographic data. This improves the accuracy and timeliness of disaster early warning, contributing to enhanced emergency response capabilities. This method can be applied to flood monitoring and early warning systems for multiple disasters, including geological disasters, urban safety hazards, and hydrological disasters, providing disaster decision support for relevant departments. Attached Figure Description

[0040] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0041] Figure 1 This is a flowchart of the intelligent monitoring and early warning method for multi-hazard, multi-modal floods according to an embodiment of the present invention. Detailed Implementation

[0042] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0043] Example

[0044] Please refer to Figure 1 The following is an example of a multi-hazard, multi-modal intelligent monitoring and early warning method for floods provided in this embodiment, which includes the following steps:

[0045] A flood threshold early warning model is established based on the triggering of early warnings when meteorological, hydrological, and infrasound data collected by multiple sensors exceed preset thresholds.

[0046] AI scene early warning and monitoring of hydrological disaster videos, geological disaster videos, and urban disaster videos are carried out based on edge-side visual analysis algorithms, and an AI recognition model for flood videos is established.

[0047] When either the flood threshold early warning model or the flood video AI recognition model issues an early warning signal, a rainstorm-runoff coupling model is established based on soil moisture content, vegetation coverage and surface slope data under different rainfall amounts to predict surface runoff, confluence time and runoff peak.

[0048] Based on the flow velocity, water level and cross-sectional morphology of the river cross section, a river confluence coupling model is constructed to predict the confluence time, confluence volume and water level changes along the river.

[0049] The risk of river confluence is determined based on real-time predictions of the river flood confluence time, flow rate, and water level changes along the course.

[0050] When there is a risk of river confluence, the surface runoff, confluence time, and rate of change of peak runoff are obtained by applying the storm-runoff coupling model based on real-time rainfall.

[0051] Based on the changes in surface runoff, confluence time, and peak runoff rate, combined with topographic data, a pre-trained flood evolution coupling model is used to predict the flood inundation range, water depth distribution, and evolution rate.

[0052] For example, the YOLOv8 target detection algorithm used in edge-side visual analysis algorithms can support the identification of key indicators in hydrological disaster videos, such as the rate of change of river surface area, length of embankment cracks, depth of road water accumulation, river siltation, bridge and culvert blockage, illegal construction in river channels and river junctions, people entering dangerous waters, and people entering flood channels; it can support the identification of key indicators in geological disaster videos, such as slope collapse volume, slope displacement monitoring, rockfall at the foot of mountains, identification of illegal construction in dangerous areas, and abnormal road interruption; and it can support the identification of key indicators in urban disaster videos, such as road water accumulation, road icing, severe snow accumulation in temporary buildings, hidden dangers of dense crowds, crowd pushing and shoving, fighting for supplies, illegal intrusion, illegal use of fire at resettlement sites, identification of people stranded in disaster-stricken areas, and illegal intrusion into disaster control areas.

[0053] The response relationship between rainfall and surface runoff is obtained through a storm-runoff coupling model to predict flood risk; the response relationship between river cross-section and flood confluence is obtained through a river confluence coupling model to predict the river flood confluence process; based on the river flood confluence risk, the surface runoff, confluence time, and rate of change of peak runoff corresponding to rainfall are obtained using the storm-runoff coupling model, and then combined with topographic data to predict the flood evolution process and provide real-time early warning.

[0054] The sensor-based meteorological data includes rainfall and rainfall duration; the sensor-based hydrological data includes river level, flow rate, flow velocity, and soil moisture content.

[0055] For example: meteorological sensors use tipping bucket rain gauges to collect rainfall and duration. Hydrological sensors use submersible hydrostatic level gauges to collect river water levels; Doppler current meters to collect river flow velocities; water level sensors to collect water levels, and the river flow can be calculated by combining the collected flow velocity and water level; and FDR soil moisture sensors to collect soil moisture content.

[0056] Ground acoustic and infrasound data collected by ground acoustic and infrasound sensors can be used to monitor early signs of geological disasters such as debris flows and landslides. Optionally, the ground acoustic sensors are piezoelectric / seismic detectors, deployed underground at the rear edge of the landslide body and on both sides of the debris flow channel; the infrasound sensors are capacitive / MEMS type, deployed in open, high locations. The timestamps of ground acoustic / infrasound data are aligned with those of rainfall, displacement, and mud level data to facilitate multi-parameter fusion verification.

[0057] The aforementioned storm-runoff coupling model, based on soil moisture content, vegetation cover, and surface slope data under different rainfall amounts, is used to predict surface runoff, confluence time, and peak runoff. Specifically, it includes:

[0058] Based on the spatiotemporal distribution data of soil moisture content, vegetation cover and surface slope, a multivariate nonlinear regression method was used to establish the response relationship between surface runoff, runoff time and peak runoff and rainfall, so as to construct a rainstorm-runoff coupling model.

[0059] Nonlinear regression methods are employed, incorporating spatiotemporal distribution data to capture complex coupling effects between variables. Nonlinear regression methods include multinomial regression or generalized additive models. Model parameters can be optimized through cross-validation and residual analysis to improve the model's predictive accuracy and stability under different rainfall amounts.

[0060] The aforementioned river confluence coupling model, constructed based on river cross-section monitoring of flow velocity, water level, and cross-sectional morphology, is used to predict river flood confluence time, confluence volume, and water level changes along the river course. Specifically, it includes:

[0061] Using river cross-section monitoring data such as flow velocity, water level, and cross-sectional morphology as inputs, and river flood confluence time, confluence volume, and water level changes along the course as outputs, a flood confluence prediction model is constructed using machine learning methods.

[0062] Machine learning methods such as Support Vector Machines (SVM), Random Forests (RF), or Long Short-Term Memory Networks (LSTM) can be used to train the nonlinear response relationship of river cross-section monitoring data to flood events. Simultaneously, data assimilation techniques (such as EnKF, Ensemble Kalman Filter) can be employed to update model state parameters in real time, improving the model's responsiveness to real-time flood events and enabling dynamic prediction of confluence time, flow rate, and water level changes.

[0063] The method of determining whether there is a risk of river confluence based on real-time predicted river flood confluence time, confluence volume, and changes in water level along the river course specifically includes:

[0064] By analyzing real-time river flood confluence time, flow rate, and water level changes along the river, and by determining whether the historical safety threshold of the river's carrying capacity is exceeded, it is possible to determine whether there is a risk of river confluence.

[0065] The method involves triggering an early warning based on meteorological, hydrological, and infrasound data collected from multiple sensors when they exceed preset thresholds, and establishing a flood threshold early warning model. Specifically, this includes setting graded thresholds for water level, flow rate, and rainfall for different hydrological year types and different watershed zones; and triggering an early warning signal and outputting the threshold early warning result when meteorological, hydrological, and infrasound data exceed the dynamic thresholds of the corresponding zones and year types.

[0066] The hydrological year type includes high-water years, normal-water years, and low-water years. The watershed is divided into upstream, midstream, and downstream sections, and a multi-level early warning system is set according to tiered thresholds. Threshold early warning results can include the warning type, triggering indicator, and the extent of exceedance.

[0067] The AI ​​scene monitoring of hydrological disaster videos, geological disaster videos, and urban disaster videos based on edge-side visual analysis algorithms, and the establishment of a flood video AI recognition model, specifically includes:

[0068] The YOLOv8 object detection algorithm was used to identify key indicators in videos of hydrological disasters, geological disasters, and urban disasters.

[0069] A CNN-LSTM hybrid model is used to perform time-series analysis on key indicators of hydrological disaster videos, geological disaster videos, and urban disaster videos to predict the changing trends of these key indicators.

[0070] When the changing trend of the identified indicators reaches the preset risk threshold, the AI ​​scenario early warning monitoring results are output.

[0071] When applied, a multimodal intelligent flood monitoring device can be used, including an integrated pole, a system that integrates the multi-hazard multimodal intelligent flood monitoring and early warning method, multiple sensors, a video monitoring system, a solar power supply system, and a waterproof speaker;

[0072] Multiple sensors are used to collect the meteorological, hydrological, or infrasound data;

[0073] The video surveillance system is used to collect videos of hydrological disasters, geological disasters, and urban disasters;

[0074] The waterproof speaker is used to provide voice alerts for the disaster risk warning results;

[0075] The solar power system is used to power multiple sensors, the video surveillance system, the waterproof speaker, and the multi-hazard, multi-modal flood intelligent monitoring and early warning system.

[0076] Optionally, the integrated pole is also equipped with a microphone to assist in collecting relevant detection sounds, and a communication module for remote connection to the supporting system terminal.

[0077] In summary, embodiments of the present invention provide a multi-hazard, multi-modal intelligent monitoring and early warning method, system, and device for floods: by integrating multiple monitoring technologies and models, intelligent monitoring and early warning of various disasters can be achieved. Based on meteorological, hydrological, and infrasound data collected by multiple sensors, preset thresholds are set. When the data exceeds these thresholds, an early warning is triggered, thereby establishing a flood threshold early warning model to ensure the efficiency of meteorological, hydrological, and geological early warnings. Simultaneously, edge-side visual analysis algorithms are used to perform AI scene early warning monitoring on hydrological disaster videos, geological disaster videos, and urban disaster videos, thereby establishing a flood video AI recognition model, which can improve the accuracy of early warnings for hydrological, geological, and urban safety. When one of the flood threshold early warning model and the flood video AI recognition model issues an early warning, due to the complexity of sensor data sources and the limitations of video recognition scenarios, in order to further improve the prediction accuracy of flood disasters... To improve accuracy, a storm-runoff coupling model is established based on soil moisture content, vegetation cover, and surface slope data under different rainfall amounts to predict runoff changes caused by storms. A river confluence coupling model is constructed based on river cross-section velocity, water level, and cross-sectional morphology to predict river flood confluence processes and assess river confluence risk. When a risk is identified, the storm-runoff coupling model is used to obtain the surface runoff, confluence time, and rate of change of peak runoff from real-time rainfall. Combined with topographic data, a pre-trained flood evolution coupling model is used to predict the flood inundation range, water depth distribution, and evolution rate, thereby simulating the flood propagation path and impact range using storm-runoff changes and topography. This application not only optimizes the accuracy of threshold models but also significantly improves the latency of video AI recognition. Furthermore, it monitors whether heavy rainfall will further cause flooding. By assessing river channel confluence risk based on river cross-sections and predicting flood risk based on changes in current rainfall, and when flood risk is present, it monitors the flood evolution process by combining rainfall-related runoff data and topographic data. This improves the accuracy and timeliness of disaster early warning, contributing to enhanced emergency response capabilities. This method can be applied to flood monitoring and early warning systems for multiple disasters, including geological disasters, urban safety hazards, and hydrological disasters, providing disaster decision support for relevant departments.

[0078] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligent monitoring and early warning of multi-hazard, multi-modal floods, characterized in that, Includes the following steps: A flood threshold early warning model is established based on the triggering of early warnings when meteorological, hydrological, and infrasound data collected by multiple sensors exceed preset thresholds. AI scene early warning and monitoring of hydrological disaster videos, geological disaster videos, and urban disaster videos are carried out based on edge-side visual analysis algorithms, and an AI recognition model for flood videos is established. When either the flood threshold early warning model or the flood video AI recognition model issues an early warning signal, a rainstorm-runoff coupling model is established based on soil moisture content, vegetation coverage and surface slope data under different rainfall amounts to predict surface runoff, confluence time and runoff peak. Based on the flow velocity, water level and cross-sectional morphology of the river cross section, a river confluence coupling model is constructed to predict the confluence time, confluence volume and water level changes along the river. The risk of river confluence is determined based on real-time predictions of the river flood confluence time, flow rate, and water level changes along the course. When there is a risk of river confluence, the surface runoff, confluence time, and rate of change of peak runoff are obtained by applying the storm-runoff coupling model based on real-time rainfall. Based on the changes in surface runoff, confluence time, and peak runoff rate, combined with topographic data, a pre-trained flood evolution coupling model is used to predict the flood inundation range, water depth distribution, and evolution rate.

2. The intelligent monitoring and early warning method for multi-hazard, multi-modal floods according to claim 1, characterized in that: Meteorological data collected by sensors includes rainfall and rainfall duration; hydrological data collected by sensors includes river level, flow rate, flow velocity, and soil moisture content.

3. The intelligent monitoring and early warning method for multi-hazard, multi-modal floods according to claim 1, characterized in that: The aforementioned storm-runoff coupling model, based on soil moisture content, vegetation cover, and surface slope data under different rainfall amounts, is used to predict surface runoff, confluence time, and peak runoff. Specifically, it includes: Based on the spatiotemporal distribution data of soil moisture content, vegetation cover and surface slope, a multivariate nonlinear regression method was used to establish the response relationship between surface runoff, runoff time and peak runoff and rainfall, so as to construct a rainstorm-runoff coupling model.

4. The intelligent monitoring and early warning method for multi-hazard, multi-modal floods according to claim 1, characterized in that: The aforementioned river confluence coupling model, constructed based on river cross-section monitoring of flow velocity, water level, and cross-sectional morphology, is used to predict river flood confluence time, confluence volume, and water level changes along the river course. Specifically, it includes: Using river cross-section monitoring data such as flow velocity, water level, and cross-sectional morphology as inputs, and river flood confluence time, confluence volume, and water level changes along the course as outputs, a flood confluence prediction model is constructed using machine learning methods.

5. The intelligent monitoring and early warning method for multi-hazard, multi-modal floods according to claim 1, characterized in that: The method of determining whether there is a risk of river confluence based on real-time predicted river flood confluence time, confluence volume, and changes in water level along the river course specifically includes: By analyzing real-time river flood confluence time, flow rate, and water level changes along the river, and by determining whether the historical safety threshold of the river's carrying capacity is exceeded, it is possible to determine whether there is a risk of river confluence.

6. The intelligent monitoring and early warning method for multi-hazard, multi-modal floods according to claim 1, characterized in that: The method involves triggering an early warning based on meteorological, hydrological, and infrasound data collected from multiple sensors when they exceed preset thresholds, and establishing a flood threshold early warning model. Specifically, this includes setting graded thresholds for water level, flow rate, and rainfall for different hydrological year types and different watershed zones; and triggering an early warning signal and outputting the threshold early warning result when meteorological, hydrological, and infrasound data exceed the dynamic thresholds of the corresponding zones and year types.

7. The intelligent monitoring and early warning method for multi-hazard, multi-modal floods according to claim 1, characterized in that: The AI ​​scene monitoring of hydrological disaster videos, geological disaster videos, and urban disaster videos based on edge-side visual analysis algorithms, and the establishment of a flood video AI recognition model, specifically includes: The YOLOv8 object detection algorithm was used to identify key indicators in videos of hydrological disasters, geological disasters, and urban disasters. A CNN-LSTM hybrid model is used to perform time-series analysis on key indicators of hydrological disaster videos, geological disaster videos, and urban disaster videos to predict the changing trends of these key indicators. When the changing trend of the identified indicators reaches the preset risk threshold, the AI ​​scenario early warning monitoring results are output.

8. A multi-hazard, multi-modal intelligent monitoring and early warning system for floods, characterized in that: include: Flood threshold early warning unit, flood video AI recognition unit, and disaster risk early warning unit; The flood threshold early warning unit is used to trigger an early warning based on meteorological, hydrological and infrasound data collected by multiple sensors exceeding a preset threshold, and to establish a flood threshold early warning model. The flood video AI recognition unit is used to perform AI scene early warning monitoring of hydrological disaster videos, geological disaster videos and urban disaster videos based on edge-side visual analysis algorithms, and to establish a flood video AI recognition model; The disaster risk early warning unit is used to establish a rainstorm-runoff coupling model to predict surface runoff, confluence time and runoff peak based on soil moisture content, vegetation coverage and surface slope data under different rainfall amounts when either the flood threshold early warning model or the flood video AI recognition model issues an early warning signal. Based on the flow velocity, water level and cross-sectional morphology of the river cross section, a river confluence coupling model is constructed to predict the confluence time, confluence volume and water level changes along the river. The risk of river confluence is determined based on real-time predictions of the river flood confluence time, flow rate, and water level changes along the course. When there is a risk of river confluence, the surface runoff, confluence time, and rate of change of peak runoff are obtained by applying the storm-runoff coupling model based on real-time rainfall. Based on the changes in surface runoff, confluence time, and peak runoff rate, combined with topographic data, a pre-trained flood evolution coupling model is used to predict the flood inundation range, water depth distribution, and evolution rate.

9. A multimodal intelligent monitoring device for floods, characterized in that: It includes an integrated pole; the integrated pole integrates a multi-hazard, multi-modal flood intelligent monitoring and early warning system as described in claim 8, multiple sensors, a video monitoring system, a solar power supply system, and a waterproof speaker; Multiple sensors are used to collect the meteorological, hydrological, or infrasound data; The video surveillance system is used to collect videos of hydrological disasters, geological disasters, and urban disasters; The waterproof speaker is used to provide voice alerts for the disaster risk warning results; The solar power system is used to power multiple sensors, the video surveillance system, the waterproof speaker, and the multi-hazard, multi-modal flood intelligent monitoring and early warning system.

Images