Flood event segmentation and feature extraction method and device
By automatically classifying flood processes using adaptive thresholds and dynamic expansion methods, the problem of manual classification errors in existing technologies is solved, and the ability to accurately quantify flood characteristics and analyze flood fluctuations is achieved.
Patent Information
- Application Number
- CN202511591971.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-03
- Publication Date
- 2026-02-06
Smart Images
Figure CN121479183A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of flood feature analysis technology, and in particular to a method and apparatus for flood event segmentation and feature extraction based on adaptive thresholding and dynamic expansion. Background Technology
[0002] Hydrological situation analysis of historical flood data can be used to predict the correlation of flood fluctuations between various hydrological stations, tributaries, and upstream and downstream areas, making it an important aspect of hydrological information forecasting. Existing technologies include adaptive methods for segmenting rainfall-runoff periods and adaptive methods for forecasting floods and flood hydrographs. The former is based on segmenting rainfall-runoff data into rainfall periods, while the latter divides flood processes based on measured daily average flow and peak flow. However, existing technologies can only adaptively segment rainfall-runoff and adaptively forecast flood hydrographs. Because subsequent data compilation and analysis are required, errors inevitably arise from the artificial flood process segmentation. Therefore, a method for automatically segmenting actual flood periods is lacking. Summary of the Invention
[0003] To overcome the shortcomings of existing technologies, this invention provides a method and apparatus for flood event segmentation and feature extraction.
[0004] A method for flood event segmentation and feature extraction includes:
[0005] S1, Obtain basic time-flow data about the flood, and perform data standardization and time granularity consistency processing on the basic time-flow data to obtain a dataset with time series as independent variable and flow corresponding to the time series as dependent variable;
[0006] S2. Set a dynamic flood threshold, and filter the dataset using the dynamic flood threshold to obtain multiple time periods that continuously exceed the dynamic flood threshold; wherein, each time period that exceeds the threshold corresponds to the occurrence of a flood event;
[0007] S3. Perform spatiotemporal expansion detection on the flood event: taking the threshold-exceeding time period as the midpoint, expand it by N time intervals before and after to obtain the spatiotemporally expanded threshold-exceeding time period; where N is a natural number greater than 1;
[0008] S4. Compare multiple flood events during the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with overlapping start times to obtain a target flood event; or, compare multiple flood events during the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with an interval time between two adjacent floods that is less than a preset threshold to obtain a target flood event.
[0009] S5. Quantify the target flood event using multidimensional flood characteristics.
[0010] A flood event segmentation and feature extraction apparatus, the apparatus being used to implement the flood event segmentation and feature extraction method as described above, and comprising:
[0011] The data standardization preprocessing and time granularity consistency processing module is used to acquire basic time-flow data about floods, and to perform data standardization and time granularity consistency processing on the basic time-flow data to obtain a dataset with time series as independent variable and flow corresponding to the time series as dependent variable;
[0012] A dynamic flood threshold calculation module is used to set a dynamic flood threshold and filter the dataset using the dynamic flood threshold to obtain multiple threshold-exceeding time periods that continuously exceed the dynamic flood threshold; wherein each threshold-exceeding time period corresponds to the occurrence of a flood event;
[0013] A flood event spatiotemporal expansion detection module is used to detect the spatiotemporal expansion of the flood event: taking the threshold-exceeding time period as the midpoint, expanding it by N time intervals before and after to obtain the spatiotemporally expanded threshold-exceeding time period; wherein, N is a natural number greater than 1;
[0014] The event overlap adaptive merging module is used to compare multiple flood events in the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with overlapping start times to obtain a target flood event; or, it is used to compare multiple flood events in the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with an interval time between two adjacent floods that is less than a preset threshold to obtain a target flood event.
[0015] The multidimensional flood feature quantization module is used to represent the target flood event using multidimensional flood feature quantization.
[0016] The flood event segmentation and feature extraction method and apparatus provided by this invention can automatically divide flood processes and events for flood characteristics in different watersheds, and quantify the features of the segmented flood processes: for the flow process of hydrological stations, the flood events in the small watershed are adaptively divided to correspond with the rainfall period, which facilitates the compilation of flood data and the study of the correlation between rainfall, runoff and flood; by utilizing the characteristic peak value, continuity, correlation and integrity of flood in flow, and inputting dynamic thresholds for features in different watersheds, the flood is segmented using difference arrays and spatiotemporal detection methods, and the flood process is expanded to include the rise and fall of the flood, and adjacent floods with small intervals are merged into a multi-peak flood, and finally the feature values of each flood are obtained, providing a reliable data foundation for subsequent flood prediction, frequency analysis and other studies.
[0017] Compared with existing technologies, this solution has the following characteristics:
[0018] 1) Set adaptive thresholds. According to the "Hydrological Information Forecasting Standard" (GB / T22482-2008), the flood magnitude is set according to the recurrence period of hydrological elements (such as peak flow and maximum flood volume in a period), and an adaptive flood threshold is set. This can be used to filter small, medium, large and catastrophic floods in a year as needed, thus improving the screening accuracy.
[0019] 2) Dynamic expansion. Considering the rise and fall of floodwaters, compared to other methods that only consider peak values or flood maxima to determine flood levels, this scheme sets dynamic expansion parameters to make the judgment more comprehensive; in addition, adjustable merging parameters are set to merge floods with small time intervals and separate those with larger intervals, making the segmentation more precise.
[0020] 3) Flood characteristic quantification. For the identified target flood events, multi-dimensional parameter quantification is performed to facilitate the generation of data tables or flood process diagrams for subsequent analysis.
[0021] Compared with existing technologies, this solution has the following advantages: it facilitates the analysis of the annual flood process, and when data from multiple hydrological stations are available, it facilitates the comparative analysis of the fluctuation relationships between upstream and downstream, and between main and tributary floods. By quantifying the characteristics of segmented flood events, it facilitates a preliminary understanding of each flood event in the current year (including the duration of the flood, peak flow, and total flow of the flood process), and can be used to analyze the impact of the flood on local or downstream water resources. Attached Figure Description
[0022] Figure 1 This is a flowchart of the flood event segmentation and feature extraction method in an embodiment of the present invention;
[0023] Figure 2 This is a schematic diagram illustrating the multidimensional flood feature quantification after segmenting and feature extraction of basic time-flow data of a hydrological station in an embodiment of the present invention. Detailed Implementation
[0024] To make the technical problems to be solved, the technical solutions, and the beneficial effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present invention and are not intended to limit the present invention.
[0025] A method for flood event segmentation and feature extraction is provided, the process of which is as follows: Figure 1 As shown, it includes the following steps:
[0026] S1. Obtain basic time-flow data about the flood, and perform data standardization and time granularity consistency processing on the basic time-flow data to obtain a dataset with time series as independent variable and the flow corresponding to the time series as dependent variable.
[0027] The baseline time-discharge data is provided by the hydrological network, but its sources are not limited to this. Data standardization and time granularity consistency processing refers to preprocessing the baseline time-discharge data to obtain the correspondence between flood occurrence time and flood discharge.
[0028] Furthermore, the basic time-based flow data undergoes data standardization and time-granularity consistency processing, including the following steps:
[0029] S11, fill in the missing items in the basic time flow data, and format the time series cells into digital format;
[0030] S12 fills the time series cells after the formatting is digitized into time series with preset time intervals;
[0031] S13. Based on the time series with preset time intervals, the basic time flow data is filled in using an interpolation algorithm to obtain a dataset with the time series as the independent variable and the flow corresponding to the time series as the dependent variable.
[0032] Specifically, taking MATLAB as an example, firstly, long-sequence continuous flow data and related time series of the relevant watershed are collected; then, missing or incomplete data are filled in, and the time series cell format is digitized; the time data is filled in into a time series with 1-hour intervals; the interpolation tool interp1 and the linear interpolation function linear are used for processing, and the flow data is processed with the filled time series as the independent variable to obtain the flow value at each time moment, so that time and flow correspond one-to-one, thus obtaining the dataset.
[0033] S2. Set a dynamic flood threshold and filter the dataset using the dynamic flood threshold to obtain multiple time periods that continuously exceed the dynamic flood threshold; where each time period that exceeds the threshold corresponds to the occurrence of a flood event.
[0034] Specifically, it includes the following steps:
[0035] S21. First, the flood threshold is dynamically determined based on the percentile of the flow (based on flood frequency analysis). The threshold is determined according to the river conditions of each basin, usually between 90% and 95%, with 90% being the preferred threshold.
[0036] Let the dynamic flood threshold be Q_th = Pα(Q).
[0037] Where Pα represents the α percentile of the flow sequence Q, α∈[90,95], preferably α=90.
[0038] S22 uses the dynamic flood threshold as a baseline to traverse the dataset and detect flood events that exceed the dynamic flood threshold.
[0039] Iterate through all the data such that Q > Q_th, and obtain a boolean array edges, in which 1 represents true and 0 represents false.
[0040] S23, use the differential algorithm to find the start and end indices of the flood event, and the start and end indices correspond to the time periods exceeding the threshold.
[0041] The difference edges are calculated using the diff function. 1 (0→1) represents the start of the flood (starts), and −1 (1→0) represents the end of the flood (ends). Starts and ends are arrays of the same length, representing the start and end indices of each event exceeding the threshold, respectively. For example, the duration interval of a flood event is obtained as [starts(i), ends(i)].
[0042] S3. Spatiotemporal expansion detection of flood events: Taking the time period exceeding the threshold as the midpoint, expand it by N time intervals before and after to obtain the time period exceeding the threshold after spatiotemporal expansion; where N is a natural number greater than 1.
[0043] Since each flood involves both rising and receding water levels, the time interval should be extended by N to ensure the completeness of the flood event. The specific rising and receding water levels should be determined based on local hydrological data; preferably, N is 6.
[0044] Specifically, the spatiotemporal spread detection algorithm satisfies:
[0045]
[0046] in, For flood events exceeding the threshold, and For the original start and stop cables
[0047] lead, To extend the time window, N represents the total amount of data.
[0048] S4. Compare multiple flood events during the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with overlapping start times to obtain the target flood event; or, compare multiple flood events during the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with an interval time between two adjacent floods that is less than a preset threshold to obtain the target flood event.
[0049] Considering that typical flood events are not single-peaked but often multi-peaked, to avoid excessive complexity in flood events and to study similar floods together, the flood events identified through the above steps will be merged. Specifically, if the start times of floods overlap, the overlapping flood events will be merged. Alternatively, the start and end times of the above flood events will be detected; if the interval between floods is less than a preset threshold, the two adjacent floods will be merged. The preset threshold can be 6 hours, 24 hours, etc.
[0050] S5. Quantify the target flood event using multidimensional flood characteristics.
[0051] Multidimensional flood characteristic quantification representation plots the flow process of the final flood event over a certain period of time (such as a whole year) on a graph, and automatically obtains the start and end time, flood flow, flood peak, flood duration, etc. of each flood, and obtains the flood process line, which facilitates subsequent research.
[0052] Specifically, the target flood event is represented by multidimensional flood feature quantification using the following feature model:
[0053] Peak traffic is:
[0054] Let the total flood volume be:
[0055] Let the duration be:
[0056] Where t is time, The end time of the event. This is the start time of the event.
[0057] like Figure 2 The image shows the results of analyzing the flood characteristics of a hydrological station using the method described in this case.
[0058] A flood event segmentation and feature extraction device is proposed to implement the aforementioned flood event segmentation and feature extraction method, and includes the following modules:
[0059] The data standardization preprocessing and time granularity consistency processing module is used to obtain basic time-flow data about floods, and to perform data standardization and time granularity consistency processing on the basic time-flow data to obtain a dataset with time series as independent variable and the flow corresponding to the time series as dependent variable;
[0060] The dynamic flood threshold calculation module is used to set dynamic flood thresholds and filter the dataset using these thresholds to obtain multiple time periods that continuously exceed the dynamic flood thresholds; each time period that exceeds the threshold corresponds to the occurrence of a flood event.
[0061] The flood event spatiotemporal expansion detection module is used to detect the spatiotemporal expansion of flood events: taking the time period exceeding the threshold as the midpoint, it expands by N time intervals before and after to obtain the time period exceeding the threshold after spatiotemporal expansion; where N is a natural number greater than 1.
[0062] The event overlap adaptive merging module is used to compare multiple flood events in the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with overlapping start times to obtain the target flood event; or, it is used to compare multiple flood events in the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with an interval time between two adjacent floods that is less than a preset threshold to obtain the target flood event.
[0063] The multidimensional flood feature quantization module is used to represent the target flood event using multidimensional flood feature quantization.
[0064] The aforementioned functional modules can be flexibly deployed on computer equipment or storage media. The functional implementation of this flood event segmentation and feature extraction device has been described in the corresponding method and will not be repeated here.
[0065] The above is a description of the present invention to help understand it; however, the implementation of the present invention is not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the principle of the present invention should be considered equivalent substitutions and are included within the protection scope of the present invention.
Claims
1. A method for flood event segmentation and feature extraction, characterized in that, include: S1, Obtain basic time-flow data about the flood, and perform data standardization and time granularity consistency processing on the basic time-flow data to obtain a dataset with time series as independent variable and flow corresponding to the time series as dependent variable; S2. Set a dynamic flood threshold, and filter the dataset using the dynamic flood threshold to obtain multiple time periods that continuously exceed the dynamic flood threshold; wherein, each time period that exceeds the threshold corresponds to the occurrence of a flood event; S3. Perform spatiotemporal expansion detection on the flood event: taking the threshold-exceeding time period as the midpoint, expand it by N time intervals before and after to obtain the spatiotemporally expanded threshold-exceeding time period; where N is a natural number greater than 1; S4. Compare multiple flood events during the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with overlapping start times to obtain a target flood event; or, compare multiple flood events during the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with an interval time between two adjacent floods that is less than a preset threshold to obtain a target flood event. S5. Quantify the target flood event using multidimensional flood characteristics.
2. The flood event segmentation and feature extraction method as described in claim 1, characterized in that, The data standardization and time granularity consistency processing includes: S11, fill in the missing items in the basic time flow data, and perform digital formatting on the time series cells; S12 fills the time series cells after the formatting is digitized into time series with preset time intervals; S13, Based on the time series with preset time intervals, the basic time flow data is filled in using an interpolation algorithm to obtain a dataset with the time series as the independent variable and the flow corresponding to the time series as the dependent variable.
3. The flood event segmentation and feature extraction method as described in claim 1, characterized in that, Setting the dynamic flood threshold includes: S21, The dynamic flood threshold is dynamically determined based on the percentile of the flow rate; The filtering of the dataset using the dynamic flood threshold includes: S22, using the dynamic flood threshold as a baseline, traverse the dataset to detect flood events that exceed the dynamic flood threshold; S23, use the differential algorithm to find the start and end index of the flood event, and the start and end index corresponds to the time period exceeding the threshold.
4. The flood event segmentation and feature extraction method as described in claim 1, characterized in that, The time interval N is 6.
5. The flood event segmentation and feature extraction method as described in claim 1, characterized in that, The multidimensional flood characteristics include flood flow, flood peak value, flood duration, and the start and end times of the flood.
6. The flood event segmentation and feature extraction method as described in claim 1, characterized in that, The preset threshold is 24 hours.
7. The flood event segmentation and feature extraction method as described in any one of claims 1 to 6, characterized in that, The target flood event is represented by a multidimensional flood feature quantification using the following feature model: Peak traffic is: ; Let the total flood volume be: Let the duration be: Where t is time, The end time of the event. This is the start time of the event.
8. A flood event segmentation and feature extraction device, characterized in that, The apparatus is used to implement the flood event segmentation and feature extraction method as described in any one of claims 1 to 7, and includes: The data standardization preprocessing and time granularity consistency processing module is used to acquire basic time-flow data about floods, and to perform data standardization and time granularity consistency processing on the basic time-flow data to obtain a dataset with time series as independent variable and flow corresponding to the time series as dependent variable; A dynamic flood threshold calculation module is used to set a dynamic flood threshold and filter the dataset using the dynamic flood threshold to obtain multiple threshold-exceeding time periods that continuously exceed the dynamic flood threshold; wherein each threshold-exceeding time period corresponds to the occurrence of a flood event; A flood event spatiotemporal expansion detection module is used to detect the spatiotemporal expansion of the flood event: taking the threshold-exceeding time period as the midpoint, expanding it by N time intervals before and after to obtain the spatiotemporally expanded threshold-exceeding time period; wherein, N is a natural number greater than 1; The event overlap adaptive merging module is used to compare multiple flood events in the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with overlapping start times to obtain a target flood event; or, it is used to compare multiple flood events in the time period exceeding the threshold after spatiotemporal expansion, and merge flood events with an interval time between two adjacent floods that is less than a preset threshold to obtain a target flood event. The multidimensional flood feature quantization module is used to represent the target flood event using multidimensional flood feature quantization.