Dynamic disaster surveying and mapping and risk early warning system and method based on deep learning and remote sensing technology

Abstract

The invention provides a disaster dynamic surveying and mapping and risk early warning system and method based on deep learning and remote sensing technology, and the system comprises a data collection and processing module which is used for collecting remote sensing monitoring data of various natural disasters, and carrying out the preprocessing of the remote sensing monitoring data based on a preset standardization rule, and obtaining a basic data set; the feature extraction module is used for extracting disaster intensity features and disaster range features based on the basic data set by adopting a deep learning model to obtain a disaster feature set; the spatial distribution processing module is used for obtaining a spatial distribution map layer of disaster influence by adopting a spatial interpolation technology based on the disaster feature set; the time sequence analysis module is used for obtaining a plurality of groups of spatial distribution map layers in a continuous time period to construct a time sequence analysis model, and obtaining a change trend of disaster influence on a time dimension; and the dynamic surveying and mapping early warning module is used for simulating future short-term disaster evolution by adopting a Prophet model short-term prediction algorithm based on the change trend and performing early warning.

Description

Technical Field

[0001] This invention belongs to the field of disaster dynamic evolution and risk early warning technology, specifically relating to a disaster dynamic mapping and risk early warning system and method based on deep learning and remote sensing technology. Background Technology

[0002] Disaster dynamic mapping and risk early warning are crucial components of public safety, directly impacting the protection of life and property and social stability. With the increasing frequency of natural disasters, timely and accurate understanding of the scope and intensity of disaster impacts has become a core requirement for reducing losses and enhancing emergency response capabilities. Research in this field is not only about technological innovation but also a significant test of humanity's ability to cope with natural challenges.

[0003] However, many current methods often struggle to adapt to the fusion of multi-source data and the capture of dynamic changes when faced with complex disaster scenarios. Especially after a disaster, the diversity of data sources and the rapid changes in environmental conditions render traditional methods significantly inadequate in terms of real-time processing and comprehensiveness. Existing technologies often cannot accurately depict the full picture of a disaster in a short period, particularly in the identification of characteristics and impact assessment of different types of disasters, lacking unified quantitative standards and adaptive capabilities.

[0004] Against this backdrop, the main challenge in research focuses on how to achieve accurate quantification of disaster impacts through technological means. The diversity of disaster characteristics requires systems to extract key information from massive amounts of data and transform it into calculable numerical expressions, a process often limited by data complexity and the difficulty of feature recognition. A deeper issue is that the impact of disasters is not only reflected in spatial distribution but also involves dynamic evolution over time. Single static analysis struggles to capture these changing trends, thus hindering the accuracy of early warning and assessment. These two aspects are interconnected; insufficient feature quantification directly affects the ability to track dynamic evolution.

[0005] Therefore, how to construct a method that can accurately extract and quantify disaster characteristics from complex data, and capture the evolution patterns of disasters in time and space, has become a key problem that this study urgently needs to solve. Summary of the Invention

[0006] To address the problems existing in the prior art, this invention provides a disaster dynamic mapping and risk early warning system and method based on deep learning and remote sensing technology. The system aims to accurately extract and quantify disaster characteristics from complex data, capture the evolution patterns of disasters in time and space, and provide timely risk warnings.

[0007] To achieve the above objectives, the present invention provides the following solution:

[0008] A disaster dynamic mapping and risk early warning system based on deep learning and remote sensing technology, the system comprising: a data acquisition and processing module, a feature extraction module, a spatial distribution processing module, a time series analysis module, and a dynamic mapping and early warning module;

[0009] The data acquisition and processing module is used to collect remote sensing monitoring data of various natural disasters, and to preprocess the remote sensing monitoring data based on preset standardization rules to obtain a basic dataset.

[0010] The feature extraction module is used to extract disaster intensity features and disaster range features based on the basic dataset using a deep learning model, thereby obtaining a disaster feature set;

[0011] The spatial distribution processing module is used to obtain a spatial distribution layer of disaster impact based on a disaster feature set and using spatial interpolation techniques.

[0012] The time series analysis module is used to acquire multiple spatial distribution layers within a continuous time period, construct a time series analysis model, and obtain the changing trend of disaster impact in the time dimension.

[0013] The dynamic mapping and early warning module is used to simulate the evolution of future disasters in the short term based on changing trends and using the Prophet model short-term prediction algorithm, and to issue early warnings based on the simulation results.

[0014] Preferably, the data acquisition and processing module includes: a data acquisition unit, a data preprocessing unit, and a data fusion unit;

[0015] The data acquisition unit is used to collect remote sensing monitoring data on various natural disasters.

[0016] The data preprocessing unit is used to preprocess remote sensing monitoring data based on preset standardization rules;

[0017] The data fusion unit is used to fuse preprocessed remote sensing monitoring data using multi-source data fusion technology to obtain a basic dataset.

[0018] Preferably, the feature extraction module includes: a data filtering unit and a feature extraction unit;

[0019] The data filtering unit is used to filter the basic dataset to obtain a subset of data containing fields for disaster intensity and disaster extent.

[0020] The feature extraction unit is used to extract disaster intensity features and disaster range features based on a subset of data and employ a deep learning model to obtain a disaster feature set.

[0021] Preferably, the spatial distribution processing module includes:

[0022]

[0023] Where r(h) is the experimental variation function value, h is the separation distance, C0 is the nugget value, C0+C is the sill value, and a is the range.

[0024] Preferably, the dynamic mapping early warning module includes:

[0025] D(t) = g(t) + p(t) + ε(t);

[0026] Where D is the disaster feature set, t is time, g(t) is the trend term, p(t) is the periodic term, and ε(t) is the error term.

[0027] This invention also provides a method for dynamic disaster mapping and risk early warning based on deep learning and remote sensing technology. The method is implemented using the aforementioned system for dynamic disaster mapping and risk early warning based on deep learning and remote sensing technology. The method includes:

[0028] S1. Collect remote sensing monitoring data of various natural disasters, and preprocess the remote sensing monitoring data based on preset standardization rules to obtain a basic dataset;

[0029] S2. Based on the basic dataset, a deep learning model is used to extract disaster intensity features and disaster range features to obtain a disaster feature set;

[0030] S3. Based on the disaster feature set, spatial interpolation technology is used to obtain a spatial distribution layer of disaster impact;

[0031] S4. Obtain multiple spatial distribution layers within a continuous time period, construct a time series analysis model, and obtain the changing trend of disaster impact in the time dimension;

[0032] S5. Based on the changing trend, the Prophet model short-term prediction algorithm is used to simulate the evolution of disasters in the short term, and early warning is given based on the simulation results.

[0033] Preferably, in step S1, remote sensing monitoring data of various natural disasters are collected, and the remote sensing monitoring data are preprocessed based on preset standardization rules to obtain a basic dataset, including:

[0034] S11. Collect remote sensing monitoring data on various natural disasters;

[0035] S12. Preprocess the remote sensing monitoring data based on preset standardization rules;

[0036] S13. Use multi-source data fusion technology to fuse the preprocessed remote sensing monitoring data to obtain a basic dataset.

[0037] Preferably, in step S2, a deep learning model is used to extract disaster intensity features and disaster extent features based on the basic dataset to obtain a disaster feature set, including:

[0038] S21. Filter the basic dataset to obtain a subset of data containing fields for disaster intensity and disaster range;

[0039] S22. Based on a subset of data, a deep learning model is used to extract disaster intensity features and disaster range features to obtain a disaster feature set.

[0040] Preferably, in step S3, a spatial distribution layer of disaster impact is obtained based on a disaster feature set and using spatial interpolation techniques, including:

[0041]

[0042] Where r(h) is the experimental variation function value, h is the separation distance, C0 is the nugget value, C0+C is the sill value, and a is the range.

[0043] Preferably, in step S5, based on the changing trend, the Prophet model short-term prediction algorithm is used to simulate the evolution of future disasters in the short term, and an early warning is issued based on the simulation results, including:

[0044] D(t) = g(t) + p(t) + ε(t);

[0045] Where D is the disaster feature set, t is time, g(t) is the trend term, p(t) is the periodic term, and ε(t) is the error term.

[0046] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0047] This invention discloses a disaster dynamic mapping and risk early warning system and method based on deep learning and remote sensing technologies. Addressing the challenges of complex disaster scenarios, insufficient feature quantification, and high identification difficulty, this invention first constructs a structured disaster baseline dataset through multi-source data fusion and standardization. Then, it uses a deep learning model to extract key features and identifies high-risk features based on a classification mechanism. Next, it employs spatial interpolation techniques to generate a spatial distribution layer of disaster impacts and combines this with a time series analysis model to determine the temporal variation trend of disaster impacts. Finally, it applies a prediction algorithm to simulate the evolution of future short-term disasters. This invention achieves effective monitoring, spatial distribution analysis, and dynamic evolution prediction of complex disaster scenarios, providing crucial technical support for disaster prevention and control decision-making. Attached Figure Description

[0048] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 This is a schematic diagram of a disaster dynamic mapping and risk early warning system module based on deep learning and remote sensing technology according to an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of the disaster dynamic mapping and risk early warning method based on deep learning and remote sensing technology according to an embodiment of the present invention. Detailed Implementation

[0051] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] Example 1

[0054] like Figure 1 As shown, the present invention provides a disaster dynamic mapping and risk early warning system based on deep learning and remote sensing technology, including: a data acquisition and processing module, a feature extraction module, a spatial distribution processing module, a time series analysis module, and a dynamic mapping and early warning module;

[0055] The data acquisition and processing module is used to collect remote sensing monitoring data of various natural disasters, and to preprocess the remote sensing monitoring data based on preset standardization rules to obtain a basic dataset.

[0056] The feature extraction module is used to extract disaster intensity features and disaster range features based on the basic dataset using a deep learning model, thereby obtaining a disaster feature set;

[0057] The spatial distribution processing module is used to obtain a spatial distribution layer of disaster impact based on a disaster feature set and using spatial interpolation techniques.

[0058] The time series analysis module is used to acquire multiple spatial distribution layers within a continuous time period, construct a time series analysis model, and obtain the changing trend of disaster impact in the time dimension.

[0059] The dynamic mapping and early warning module is used to simulate the evolution of future disasters in the short term based on changing trends and using the Prophet model short-term prediction algorithm, and to issue early warnings based on the simulation results.

[0060] Specifically, the data acquisition and processing module includes: a data acquisition unit, a data preprocessing unit, and a data fusion unit;

[0061] The data acquisition unit is used to collect remote sensing monitoring data of various natural disasters. It focuses on collecting remote sensing data such as multi-temporal, multispectral / hyperspectral / radar satellite remote sensing images and UAV aerial images. At the same time, it integrates ground sensor network data (such as rain gauges, water level stations, GNSS), meteorological data, basic geographic data and socio-economic statistics.

[0062] The data preprocessing unit is used to preprocess remote sensing monitoring data based on preset standardized rules, and to perform professional preprocessing processes such as radiometric calibration, atmospheric correction, geometric fine correction / orthorectification, image fusion, mosaicking, and cropping on the remote sensing data.

[0063] The data fusion unit is used to fuse preprocessed remote sensing monitoring data using multi-source data fusion technology to obtain a basic dataset.

[0064] Specifically, multi-source data fusion technologies include:

[0065] The weighted average method assigns weights to the data based on the reliability of each data source and then sums the data in a weighted manner.

[0066] Kalman filtering, based on a state-space model, optimizes the fusion result through recursive estimation;

[0067] DS evidence theory addresses uncertainty through basic probability allocation and evidence combination rules;

[0068] Deep learning methods utilize neural networks to automatically learn complex relationships between data;

[0069] Bayesian networks use conditional probability distributions to represent the dependencies between variables and perform probabilistic inference.

[0070] The data fusion unit employs machine learning algorithms to dynamically allocate weights using entropy weighting and analytic hierarchy process, ensuring that high-value data sources receive higher weights and automatically adjusting the weights of different data sources to improve the final accuracy and flexibility.

[0071] The feature extraction module includes: a data filtering unit and a feature extraction unit;

[0072] The data filtering unit is used to filter the basic dataset to obtain a subset of data containing fields for disaster intensity and disaster extent.

[0073] The feature extraction unit is used to extract disaster intensity features and disaster range features based on a subset of data and employ a deep learning model to obtain a disaster feature set.

[0074] Using deep learning models (such as convolutional neural networks CNN), key disaster features are automatically extracted, including: disaster range features (such as flood-inundated areas, wildfire-burned areas, and landslide areas), disaster intensity features (such as water depth / turbidity estimation, fire intensity, and surface deformation rate), etc., to generate a disaster feature set containing spatial location attributes.

[0075] The spatial distribution processing module employs an improved kriging method to obtain a spatial distribution layer of disaster impacts.

[0076] Kriging, proposed by Matheron (1971), is a statistical method that provides unbiased and optimal estimates of the values ​​of regionalized variables based on their correlation and variability. Its optimal applicability is when spatial correlation exists among the variables. Considering that the variogram only needs to be positive definite to be considered effective, an improvement to the ordinary kriging method is proposed, as shown in the following model formula:

[0077]

[0078] Where r(h) is the experimental variation function value, h is the separation distance, C0 is the nugget value, C0+C is the sill value, and a is the range.

[0079] The corresponding quality control meets the following requirements:

[0080] |E i -Q i |≤f·σ;

[0081] Among them, E i Q is an estimate. i σ is the original value, f is the prediction standard error, and f is the quality control parameter, the value of which can vary depending on the application requirements.

[0082] The time series analysis module is used to obtain multiple spatial distribution layers within a continuous time period, construct a time series analysis model, and obtain the changing trend of disaster impact in the time dimension. Specifically, the time-space autoregressive model (STAR) is used as the time series analysis model to obtain the changing trend of disaster impact in the time dimension.

[0083] The dynamic mapping and early warning module uses the Prophet model short-term prediction algorithm to simulate the evolution of disasters in the near future, and obtains risk values ​​based on the simulation results. The risk values ​​are compared with preset risk thresholds, and an early warning is issued in a timely manner when the risk value exceeds the risk threshold.

[0084] Specifically, the Prophet model's short-term prediction algorithm uses an additive model, assuming that the disaster feature set D is a time series composed of a trend term, a periodicity term, and an error term, with the specific relationship shown in the following formula:

[0085] D(t) = g(t) + p(t) + ε(t);

[0086] Where D is the disaster feature set, t is time, g(t) is the trend term, p(t) is the periodic term, and ε(t) is the error term.

[0087] In summary, this invention addresses the challenges of complex disaster scenarios, insufficient feature quantification, and high identification difficulty. First, it constructs a structured disaster baseline dataset through multi-source data fusion and standardization. Then, it extracts key features using a deep learning model and identifies high-risk features based on a classification mechanism. Next, it employs spatial interpolation techniques to generate a spatial distribution layer of disaster impacts and combines this with a time series analysis model to determine the temporal trend of disaster impacts. Finally, it applies a prediction algorithm to simulate the evolution of future short-term disasters. This invention achieves effective monitoring, spatial distribution analysis, and dynamic evolution prediction of complex disaster scenarios, providing crucial technical support for disaster prevention and control decision-making.

[0088] Example 2

[0089] like Figure 2 As shown, the present invention also provides a method for dynamic disaster mapping and risk early warning based on deep learning and remote sensing technology, implemented using the disaster dynamic mapping and risk early warning system based on deep learning and remote sensing technology described in the foregoing embodiments. The method includes:

[0090] S1. Collect remote sensing monitoring data of various natural disasters, and preprocess the remote sensing monitoring data based on preset standardization rules to obtain a basic dataset;

[0091] S2. Based on the basic dataset, a deep learning model is used to extract disaster intensity features and disaster range features to obtain a disaster feature set;

[0092] S3. Based on the disaster feature set, spatial interpolation technology is used to obtain a spatial distribution layer of disaster impact;

[0093] S4. Obtain multiple spatial distribution layers within a continuous time period, construct a time series analysis model, and obtain the changing trend of disaster impact in the time dimension;

[0094] S5. Based on the changing trend, the Prophet model short-term prediction algorithm is used to simulate the evolution of disasters in the short term, and early warning is given based on the simulation results.

[0095] Specifically, S1 collects remote sensing monitoring data on various natural disasters, preprocesses the data based on preset standardization rules, and obtains a basic dataset, including:

[0096] S11. Collect remote sensing monitoring data on various natural disasters;

[0097] S12. Preprocess the remote sensing monitoring data based on preset standardization rules;

[0098] S13. Use multi-source data fusion technology to fuse the preprocessed remote sensing monitoring data to obtain a basic dataset.

[0099] Based on the basic dataset, S2 uses a deep learning model to extract disaster intensity features and disaster extent features, obtaining a disaster feature set, including:

[0100] S21. Filter the basic dataset to obtain a subset of data containing fields for disaster intensity and disaster range;

[0101] S22. Based on a subset of data, a deep learning model is used to extract disaster intensity features and disaster range features to obtain a disaster feature set.

[0102] Based on a disaster feature set, S3 uses spatial interpolation techniques to obtain a spatial distribution layer of disaster impacts, including:

[0103]

[0104] Where r(h) is the experimental variation function value, h is the separation distance, C0 is the nugget value, C0+C is the sill value, and a is the range.

[0105] S5 uses the Prophet model short-term prediction algorithm based on changing trends to simulate the evolution of future disasters in the short term, and issues early warnings based on the simulation results, including:

[0106] D(t) = g(t) + p(t) + ε(t);

[0107] Where D is the disaster feature set, t is time, g(t) is the trend term, p(t) is the periodic term, and ε(t) is the error term.

[0108] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A disaster dynamic mapping and risk early warning system based on deep learning and remote sensing technology, characterized in that, The system includes: a data acquisition and processing module, a feature extraction module, a spatial distribution processing module, a time series analysis module, and a dynamic mapping and early warning module; The data acquisition and processing module is used to collect remote sensing monitoring data of various natural disasters, and to preprocess the remote sensing monitoring data based on preset standardization rules to obtain a basic dataset. The feature extraction module is used to extract disaster intensity features and disaster range features based on the basic dataset using a deep learning model, thereby obtaining a disaster feature set; The spatial distribution processing module is used to obtain a spatial distribution layer of disaster impact based on a disaster feature set and using spatial interpolation techniques. The time series analysis module is used to acquire multiple spatial distribution layers within a continuous time period, construct a time series analysis model, and obtain the changing trend of disaster impact in the time dimension. The dynamic mapping and early warning module is used to simulate the evolution of future disasters in the short term based on changing trends and using the Prophet model short-term prediction algorithm, and to issue early warnings based on the simulation results.

2. The disaster dynamic mapping and risk early warning system based on deep learning and remote sensing technology according to claim 1, characterized in that, The data acquisition and processing module includes: a data acquisition unit, a data preprocessing unit, and a data fusion unit; The data acquisition unit is used to collect remote sensing monitoring data on various natural disasters; The data preprocessing unit is used to preprocess remote sensing monitoring data based on preset standardization rules; The data fusion unit is used to fuse preprocessed remote sensing monitoring data using multi-source data fusion technology to obtain a basic dataset.

3. The disaster dynamic mapping and risk early warning system based on deep learning and remote sensing technology according to claim 1, characterized in that, The feature extraction module includes: a data filtering unit and a feature extraction unit; The data filtering unit is used to filter the basic dataset to obtain a subset of data containing fields for disaster intensity and disaster extent. The feature extraction unit is used to extract disaster intensity features and disaster range features based on a subset of data and employ a deep learning model to obtain a disaster feature set.

4. The disaster dynamic mapping and risk early warning system based on deep learning and remote sensing technology according to claim 1, characterized in that, The spatial distribution processing module includes: Where r(h) is the experimental variation function value, h is the separation distance, C0 is the nugget value, C0+C is the sill value, and a is the range.

5. The disaster dynamic mapping and risk early warning system based on deep learning and remote sensing technology according to claim 1, characterized in that, The dynamic mapping and early warning module includes: D(t) = g(t) + p(t) + ε(t); Where D is the disaster feature set, t is time, g(t) is the trend term, p(t) is the periodic term, and ε(t) is the error term.

6. A method for dynamic disaster mapping and risk early warning based on deep learning and remote sensing technology, wherein the method is implemented using the dynamic disaster mapping and risk early warning system based on deep learning and remote sensing technology as described in any one of claims 1-5, characterized in that, The method includes: S1. Collect remote sensing monitoring data of various natural disasters, and preprocess the remote sensing monitoring data based on preset standardization rules to obtain a basic dataset; S2. Based on the basic dataset, a deep learning model is used to extract disaster intensity features and disaster range features to obtain a disaster feature set; S3. Based on the disaster feature set, spatial interpolation technology is used to obtain a spatial distribution layer of disaster impact; S4. Obtain multiple spatial distribution layers within a continuous time period, construct a time series analysis model, and obtain the changing trend of disaster impact in the time dimension; S5. Based on the changing trend, the Prophet model short-term prediction algorithm is used to simulate the evolution of disasters in the short term, and early warning is given based on the simulation results.

7. The disaster dynamic mapping and risk early warning method based on deep learning and remote sensing technology according to claim 6, characterized in that, In step S1, remote sensing monitoring data of various natural disasters are collected. Based on preset standardization rules, the remote sensing monitoring data is preprocessed to obtain a basic dataset, including: S11. Collect remote sensing monitoring data on various natural disasters; S12. Preprocess the remote sensing monitoring data based on preset standardization rules; S13. Use multi-source data fusion technology to fuse the preprocessed remote sensing monitoring data to obtain a basic dataset.

8. The disaster dynamic mapping and risk early warning method based on deep learning and remote sensing technology according to claim 6, characterized in that, In S2, based on the basic dataset, a deep learning model is used to extract disaster intensity features and disaster extent features to obtain a disaster feature set, including: S21. Filter the basic dataset to obtain a subset of data containing fields for disaster intensity and disaster range; S22. Based on a subset of data, a deep learning model is used to extract disaster intensity features and disaster range features to obtain a disaster feature set.

9. The disaster dynamic mapping and risk early warning method based on deep learning and remote sensing technology according to claim 6, characterized in that, S3, based on a disaster feature set, uses spatial interpolation techniques to obtain a spatial distribution layer of disaster impacts, including: Where r(h) is the experimental variation function value, h is the separation distance, C0 is the nugget value, C0+C is the sill value, and a is the range.

10. The disaster dynamic mapping and risk early warning method based on deep learning and remote sensing technology according to claim 6, characterized in that, In S5, based on changing trends, the Prophet model short-term prediction algorithm is used to simulate the evolution of future disasters in the short term, and early warning is issued based on the simulation results, including: D(t) = g(t) + p(t) + ε(t); Where D is the disaster feature set, t is time, g(t) is the trend term, p(t) is the periodic term, and ε(t) is the error term.

Images