WO2022184088A1 - 一种基于集成学习的洪水敏感性风险评估方法 - Google Patents

一种基于集成学习的洪水敏感性风险评估方法 Download PDF

Info

Publication number
WO2022184088A1
WO2022184088A1 PCT/CN2022/078765 CN2022078765W WO2022184088A1 WO 2022184088 A1 WO2022184088 A1 WO 2022184088A1 CN 2022078765 W CN2022078765 W CN 2022078765W WO 2022184088 A1 WO2022184088 A1 WO 2022184088A1
Authority
WO
WIPO (PCT)
Prior art keywords
flood
data
risk
sensitivity
ensemble learning
Prior art date
Application number
PCT/CN2022/078765
Other languages
English (en)
French (fr)
Inventor
胡鹤轩
胡强
张晔
胡震云
Original Assignee
河海大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 河海大学 filed Critical 河海大学
Priority to US17/995,018 priority Critical patent/US20230141886A1/en
Publication of WO2022184088A1 publication Critical patent/WO2022184088A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/10—Terrestrial scenes
    • G06V20/13—Satellite images
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
    • G06F16/21—Design, administration or maintenance of databases
    • G06F16/215—Improving data quality; Data cleansing, e.g. de-duplication, removing invalid entries or correcting typographical errors
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/211—Selection of the most significant subset of features
    • G06F18/2113—Selection of the most significant subset of features by ranking or filtering the set of features, e.g. using a measure of variance or of feature cross-correlation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G06N20/20—Ensemble learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/01—Dynamic search techniques; Heuristics; Dynamic trees; Branch-and-bound
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
    • G06V10/7747—Organisation of the process, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/776—Validation; Performance evaluation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
    • G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
    • G06V10/809—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of classification results, e.g. where the classifiers operate on the same input data
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A10/00—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE at coastal zones; at river basins
    • Y02A10/40—Controlling or monitoring, e.g. of flood or hurricane; Forecasting, e.g. risk assessment or mapping

Definitions

  • the invention belongs to the technical field of flood disaster risk assessment, in particular to a flood sensitivity risk assessment method based on integrated learning.
  • flood disaster is a destructive, sudden and frequent natural disaster. China is one of the countries with the most frequent flood disasters. Every year, flood disasters cause a lot of economic losses and casualties. Therefore, research in the field of flood risk sensitivity assessment is of great significance.
  • Flood risk sensitivity assessment is a comprehensive evaluation of the natural and social attributes of regional flood disasters, aiming to more accurately grasp the spatial distribution of flood risk and its occurrence law. Because flood risk sensitivity assessment is a very complex process, and its assessment process involves multiple evaluation indicators, it has always been one of the difficulties and hot spots in disaster research at home and abroad.
  • patent application CN106651211A discloses a method for regional flood disaster risk assessment at different scales, using AHP analytic hierarchy process and entropy weight method coupled model to evaluate the flood disaster risk value in the study area and classify the risk level.
  • this method needs to collect a large amount of natural and social data as input. Once the amount of data is low or the quality of data is not high, it will cause relatively large deviations in the results.
  • this method requires higher professional knowledge of operators, and when there are a large number of flood influencing factors, it will cause confusion in the judgment of operators, thus affecting the assessment results.
  • the existing flood sensitivity risk assessment methods have the following defects: (1) It requires a large amount of natural and social data, and the data collection workload is large. (2) The requirements for the professional knowledge of the operators are relatively high. (3) The operation time is long and the precision is relatively low.
  • the purpose of the present invention is to overcome the defects of the prior art, and provide a flood sensitivity risk assessment method based on integrated learning, which can effectively establish a flood disaster risk assessment model and solve flood disaster prevention and mitigation measures for meteorological departments and relevant local governments.
  • the method avoids a large amount of manual data collection, has high efficiency, is easy to operate, and has short operation time and high precision.
  • a flood sensitivity risk assessment method based on ensemble learning including the following steps:
  • Step 1 Collection and arrangement of initial data at sample sites: draw flood location maps of the watershed and create a flood-related spatial database by using literature data and field surveys; and select adjustment factors through data obtained from literature and field surveys; Select several flood adjustment factors for sensitivity analysis and build a spatial database of these factors;
  • Step 2 Clean, standardize and assign the collected initial data to each evaluation unit, and convert it into a raster data storage format. All data are subjected to projection transformation and resampling operations;
  • the corresponding hydrological station obtains historical flow data, finds out the peak date of the annual flood flow, and selects the MODIS image of the corresponding date to reflect the inundation status of the flood process; the inundation range reflected by several images corresponding to the peak flow is superimposed to generate a The combined largest inundation range map is used as the inundation range map corresponding to the peak flow, that is, the maximum inundation range; within the maximum inundation range, the number of randomly selected flood inundation sample points is N, and in the non-maximum flood inundation range, the number is randomly selected as N non-flooded sample points, which together constitute a total of 2N sample points; the above sample points are divided into training set and test set, of which 70% of the sample points are used as the training set and 30% of the sample points are used as the test set;
  • Step 3 Calculate the Laplacian score to determine the final feature subset: Use the Laplacian score to score the features of the training set samples described in Step 2, get the score of each feature, and finally take the one with the highest score.
  • the k features are used as the selected feature subset; the feature subset is extracted from the sample points with a total of 2N in step 2 to form a new training set and test set;
  • Step 4 Use the new training set in Step 3 to train the ensemble learning LightGBM model; obtain the accuracy of the ensemble learning LightGBM model in the new training set and test set;
  • Step 5 Use the trained model to calculate the entire watershed to obtain the probability value of the flood risk sensitivity of the entire watershed;
  • step 1 several factors described in step 1 include: atmosphere, evaporation, topography, and river network; 10 flood risk sensitivity assessment indicators can be proposed from these 4 factors, including elevation, slope, curvature, TWI, SPI, Distance from river, soil, vegetation, slope aspect and rainfall; according to the mechanism of watershed flooding; these factors are calculated and processed based on ArcGIS software, where SPI and TWI are calculated using the following formulas:
  • ⁇ is the cumulative slope drainage through a point
  • As is the specific watershed area
  • tan ⁇ is the slope angle at that point.
  • the initial data standardization processing in the step 2 includes:
  • step 3 the process of calculating the Laplacian score to determine the final feature subset described in step 3 includes:
  • the resulting matrix is the weight matrix S of the training set, where
  • L r is the Laplace score of the r-th feature
  • f ri -f rj is the difference between the r-th feature of the i-th sample and the j-th sample
  • S ij is the corresponding value in the weight matrix
  • Var(f r ) is the variance of the rth feature over all samples.
  • the flood disaster risk research area level is divided into five levels: low risk area, low risk area, medium risk area, high risk and extremely high risk area.
  • the present invention has the following advantages and beneficial effects:
  • the LightGBM adopted in the present invention occupies less memory, takes less time for operation, and has higher precision.
  • FIG. 1 is a flow chart of a method according to an embodiment of the present invention.
  • FIG. 2 is a flow chart of calculating a Laplacian score according to an embodiment of the present invention.
  • FIG. 3 is a result diagram of an implementation verification method of the present invention.
  • the invention provides a flood sensitivity risk assessment method based on integrated learning, which includes: collecting the topography, hydrometeorology, soil vegetation and other data of the study area as characteristic data, and standardizing the characteristic data; Remote sensing data extraction to study historical submerged and non-submerged points in the watershed; use Laplace score to select the optimal feature subset; divide the sample points into training set and test set and train the ensemble learning model; use the trained model Flood risk sensitivity calculation is performed for the entire watershed, and a flood sensitivity risk level distribution map of the watershed is generated.
  • the invention uses each characteristic data of the research area as input, adopts a novel integrated learning model, improves the accuracy of flood risk assessment in the basin, and finally generates a flood risk map of the basin, which can intuitively display the flood risk status of the research area.
  • FIG. 1 is a flow chart of a flood sensitivity risk assessment method based on ensemble learning provided by the present invention.
  • the first step is to collect and organize sample data. In order to estimate future flooding events in an area, it is important to analyze its past records. First, map flood locations in the watershed and create a flood-related spatial database by using literature and field surveys. Second, moderators were selected through data obtained from the literature as well as field surveys. Finally, several flood adjustment factors were selected for sensitivity analysis, and a spatial database of these factors was established.
  • ⁇ is the cumulative slope drainage through a point
  • As is the specific watershed area
  • tan ⁇ is the slope angle at that point.
  • Step 2 Clean and standardize the collected initial data to unify the coordinate system; standardize the original sample data described in step 1 and assign it to each evaluation unit, and convert it into a grid data storage format. All have undergone projection transformation and resampling operations. Since flow peaks are the main cause of flood disasters, for each study area, after obtaining historical flow data from its corresponding hydrological station, find out the peak date of annual flood flow, and select the MODIS image of the corresponding date to reflect the The inundation status of the flood process.
  • the above sample points are divided into training set and test set, of which 70% of the sample points are used as the training set and 30% of the sample points are used as the test set.
  • the sample point selection method uses historical remote sensing technology to extract the maximum submerged area map, which has the advantages of good intuition and high accuracy, and also avoids a lot of manual data collection work, which greatly improves the efficiency.
  • Step 3 Calculate the Laplacian score to determine the final feature subset: Use the Laplacian score to score the features of the training set samples described in Step 2, get the score of each feature, and finally take the one with the highest score.
  • the k features are used as the selected feature subset; the feature subset is extracted from the sample points with a total of 2N in step 2 to form a new training set and a test set.
  • FIG. 2 it is a flow chart of calculating the Laplacian score according to an embodiment of the present invention.
  • L r is the Laplace score of the r-th feature
  • f ri -f rj is the difference between the r-th feature of the i-th sample and the j-th sample;
  • S ij is the corresponding value in the weight matrix
  • Var(f r ) is the variance of the r-th feature on all samples
  • each feature will be given a score, and finally the k features with the highest score are taken as the final selected feature subset.
  • the Laplace score method the importance of each flood impact factor on the assessment results can be intuitively seen. After the overall flood risk assessment is carried out, the operator can directly prevent the impact factors that have a high degree of impact on the results. Compared with traditional manual judgment, the Laplacian scoring method greatly reduces the operating threshold.
  • Step 4 Use the new training set in Step 3 to train the ensemble learning LightGBM model; obtain the accuracy of the ensemble learning LightGBM model in the new training set and test set.
  • LightGBM Lightweight Gradient Boosting Tree
  • GOSS one-sided gradient sampling
  • GOSS uses decision tree learning to obtain a function that maps the input space to the gradient space.
  • the feature subset obtained by the Laplacian scoring method in step 3 has a total of n instances
  • the feature dimension is s
  • the negative gradient direction of the loss function of the LightGBM model is expressed as g1, ..., g n
  • the decision The tree divides the sample data into each leaf node through the optimal segmentation point (maximum information gain point), and the segmentation point d of feature j is defined as:
  • n O ⁇ I[x i ⁇ O]
  • O represents the training set of a fixed node.
  • GOSS sorts in descending order according to gradient training, and retains topa sample instances as data subset A.
  • a data subset B of size b is randomly sampled, and then data sets A and B are merged.
  • a weak classifier is trained; then the GOSS algorithm is repeated to train multiple weak classifiers until the formula (5) converges or the number of iteration steps is reached, and finally the information gains of all trained weak classifiers are added to obtain The final ensemble learning model is obtained, and the accuracy of the ensemble learning LightGBM model in the new training set and test set is obtained.
  • Step 5 Use the trained model to calculate the entire watershed to obtain the probability value of the flood risk sensitivity of the entire watershed.
  • the probability map needs to be classified into different regions.
  • various methods exist in research such as equal interval, quantile, standard deviation.
  • the best output is generally obtained by using the quantile method for the flood basin, and the flood risk sensitivity map is obtained, and the flood disaster risk research area is divided into five types of flood sensitivity: low risk area, low risk area, medium Risk areas, high risk and very high risk areas.
  • Sanmenxia to Huayuankou in the Yellow River Basin is selected as the study area, and the MODIS remote sensing image is obtained by using historical flood data recorded in hydrology books, thereby obtaining the maximum submerged range of the study area, and randomly sampling from it.
  • a total of 300 submerged sample points and 300 non-submerged sample points were selected in the study area, 70% of which were used as training set and 30% as test set.
  • the study area selected a total of 10 flood impact factors including elevation, slope, aspect, curvature, SPI, TWI, distance from the river, soil, vegetation and rainfall, and calculated the Laplace scores of each flood impact factor. The results are shown in Table 1.
  • the present invention selects LightGBM and XGBoost, the mainstream integrated learning method on the market, to conduct comparative tests. After comparative tests, it is found that the accuracy rate of XGBoost is 80.97%, and the accuracy rate of LightGBM is 81.29%, and the speed of operation must be Much higher than XGBoost.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computing Systems (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Mathematical Physics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Multimedia (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Quality & Reliability (AREA)
  • Computational Linguistics (AREA)
  • Operations Research (AREA)
  • General Business, Economics & Management (AREA)
  • Molecular Biology (AREA)
  • Tourism & Hospitality (AREA)
  • Game Theory and Decision Science (AREA)
  • Biophysics (AREA)
  • Marketing (AREA)
  • Biomedical Technology (AREA)
  • Astronomy & Astrophysics (AREA)
  • Remote Sensing (AREA)

Abstract

本发明公开了一种基于集成学习的洪水敏感性风险评估方法，包括：收集研究区的地形地貌、水文气象、土壤植被等数据作为特征数据，并将特征数据进行标准化处理；根据历史水位数据和遥感数据提取研究流域历史淹没点与非淹没点；利用拉普拉斯得分选择最优的特征子集；将样本点分为训练集和测试集并对集成学习模型进行训练；利用训练好的模型对整个流域进行洪水风险敏感性计算，生成流域洪水敏感性风险等级分布图。本发明使用研究区各特征数据作为输入，采用了新颖的集成学习模型，提高了流域洪水风险评估的准确性，最后生成流域洪水风险映射图，可直观地展现研究区的洪水风险状况。

Description

一种基于集成学习的洪水敏感性风险评估方法 技术领域
本发明属于洪水灾害风险评估技术领域，尤其涉及一种基于集成学习的洪水敏感性风险评估方法。
背景技术
洪涝灾害是一种破坏性大，突发性强且发生频率高的自然灾害。中国是洪涝灾害发生最频繁的国家之一，每年都会因洪涝灾害而造成大量经济损失和人员伤亡，故在洪水风险敏感性评估领域的研究意义重大。洪水风险敏感性评估是对区域洪水灾害自然属性和社会属性的综合评价，旨在更准确地把握洪水风险的空间分布和其发生规律。由于洪水风险敏感性评估是一个十分复杂的过程，其评估过程涉及多个评价指标，因此一直是国内外灾害研究的难点和热点之一。
随着人工智能技术的发展，将机器学习算法应用于目标评价已成为一种趋势，但仍存在一些不足。例如，在现有技术中，专利申请CN106651211A公开了一种不同尺度区域洪水灾害风险评估的方法，利用AHP层次分析法与熵权法耦合模型评估研究区内洪水灾害风险值并划分风险等级。但是这种方法需要收集大量的自然及社会数据作为输入，一旦数据量偏低或者数据质量不高都会对结果造成比较大的偏差。另一方面，这种方法对操作人员的专业知识要求较高，当洪水影响因子的数量较多时会引起操作人员的判断混乱，从而对评估结果产生影响。
而由赖成光等人于2015年1月在《水利学报》第46卷第一期58页提出的基于随机森林的洪灾风险评估方法，则简化了风险评估过程，但是具有运行时间相对较长，精度不高的问题。
综上所述，现有的洪水敏感性风险评估方法具有以下缺陷：(1)需要大量的自然以及社会数据，数据收集工作量大。(2)对操作人员的专业知识要求较高。(3)操作运行时间长，精度相对不高。
发明内容
本发明的目的在于克服现有技术的缺陷，提供一种基于集成学习的洪水敏感性风险评估方法，可有效建立洪水灾害风险评估模型，为气象部门及相关地方政府解决洪水灾害防灾减灾措施。该方法避免了大量的人工数据收集，效率高，便于操作，且操作运行时间短，精度高。
为了解决上述技术问题，本申请采用以下技术方案。
一种基于集成学习的洪水敏感性风险评估方法，包括以下步骤：
步骤一、样本点初始数据的收集与整理：通过使用文献资料和实地调查绘制流域的洪水位置图并创建与洪水有关的空间数据库；并通过从文献中获得的数据以及现场调查，选择调节因素；选择数个洪水调节因子进行敏感性分析，并建立这些因子的空间数据库；
步骤二、对所收集的初始数据进行清洗、标准化处理并赋值给每个评价单元，转换成栅格数据存储格式，所有的数据都经过投影转换与重采样操作；对于每一个研究区域，从它对应的水文站获取历史流量数据，找出每年洪水流量峰值日期，并选择对应日期的MODIS影像来反映该洪水过程的淹没状况；将流量峰值对应的数个影像反映的淹没范围叠置，生成一个合并的最大的淹没范围图，作为流量峰值所对应的淹没范围图，即最大淹没范围；在最大淹没范围内随机选取数量为N的洪水淹没样本点，在非最大洪水淹没范围内随机选取数量为N的非洪水淹没样本点，共同构成总数为2N的样本点；将上述样本点分为训练集和测试集，其中70％的样本点作为训练集，30％的样本点作为测试集；
步骤三、计算拉普拉斯得分确定最后的特征子集：利用拉普拉斯得分对步骤二中所述的训练集样本的特征进行打分，得到每一个特征的分数，最后再取分数最高的k个特征作为选择的特征子集；对步骤二中总数为2N的样本点进行特征子集的抽取，形成新的训练集和测试集；
步骤四、利用步骤三中新的训练集对集成学习LightGBM模型进行训练；得到集成学习LightGBM模型在新的训练集和测试集的准确率；
步骤五、利用训练好的模型对整个流域进行计算，得到整个流域洪水风险敏感性的概率值；
进一步地，步骤一中所述的数个因子包括：大气、蒸发、地形、河网；可从该4项因子中提出10项洪水风险敏感性评估指标包括高程、坡度、曲率、TWI、SPI、距河流距离、土壤、植被、坡向和降雨；根据流域洪水的机理；将这些因素均基于ArcGIS软件进行计算和处理，其中SPI和TWI使用以下公式计算：
TWI＝Ln(α/tanβ)       (1)
SPI＝A s tanβ         (2)
式中α是通过一个点的累计坡面排水量，A s为特定流域面积，tanβ是该点处的坡度角。
进一步地，所述步骤二中初始数据标准化处理，包括：
对样本数据集合S进行数据清洗，去除有缺失和不需要的数据并进行关联性验证；
所有尺度条件因子均使用流行的分位数方法进行分类；准备好数据集后，将每个条件因子转换为m*n大小的网格空间数据库，并构造流域地区的网格图。
进一步地，步骤三中所述的计算拉普拉斯得分确定最后特征子集的过程包括：
针对步骤二中训练集样本，构建一张邻接矩阵G：当type(i)＝type(j)时，G ij＝1，否则G ij＝0)，然后对于矩阵中G ij＝1的点，令
Figure PCTCN2022078765-appb-000001
其中t为合适的常数；
由此得到的矩阵就是该训练集的权重矩阵S，其中
Figure PCTCN2022078765-appb-000002
计算拉普拉斯得分的公式为：
Figure PCTCN2022078765-appb-000003
其中，L r为第r个特征的拉普拉斯得分；f ri-f rj为第i个样本和第j个样本的第r个特 征的差值；S ij为权重矩阵中的对应的值；Var(f r)为第r个特征在所有样本上的方差。
进一步地，在所述步骤五中，将洪水灾害风险研究区域等级分为五级:低风险区、较低风险区、中等风险区、高风险和极高风险地区。
与现有技术相比，本发明具有以下优点和有益效果：
(1)采用历史遥感技术提取每年洪水流量峰值日期的MODIS影像来反映该洪水过程的淹没状况，生成最大淹没范围图，具有直观性好，准确性高的优点，同时也避免了大量的人工数据收集工作，大大提高了效率。
(2)利用拉普拉斯得分法可以直观的看出各个洪水影响因子对评估结果影响的重要程度，在整体进行了洪水风险评估之后，操作人员可以直接对对结果影响程度较高的影响因子进行预防，相比于传统的人工判断，使可操作性大大提高。
(3)与传统的集成学习方法相比，本发明采用的LightGBM占用的内存更小，运算的时间更少，精度也越高。
附图说明
图1是本发明的一种实施例的方法流程图。
图2是本发明的一种实施例的计算拉普拉斯得分流程图。
图3是本发明的一种实施验证方法结果图。
具体实施方式
本发明提供了一种基于集成学习的洪水敏感性风险评估方法，包括：收集研究区的地形地貌、水文气象、土壤植被等数据作为特征数据，并将特征数据进行标准化处理；根据历史水位数据和遥感数据提取研究流域历史淹没点与非淹没点；利用拉普拉斯得分选择最优的特征子集；将样本点分为训练集和测试集并对集成学习模型进行训练；利用训练好的模型对整个流域进行洪水风险敏感性计算，生成流域洪水敏感性风险等级分布图。本发明使用研究区各特征数据作为输入，采用了新颖的集成学习模型，提高了流域洪水风险评估的准确性，最后生成流域洪水风险映射图，可直观地展现研究区的洪水风险状况。
下面结合附图对本发明做进一步详细说明。
图1是本发明提供的一种基于集成学习的洪水敏感性风险评估方法流程图。
步骤一、样本点数据的收集与整理。为了估算某个地区未来的洪灾事件，分析其过去的记录非常重要。首先，通过使用文献资料和实地调查绘制流域的洪水位置图并创建与洪水有关的空间数据库。其次通过从文献中获得的数据以及现场调查，选择调节因素。最后选择数个洪水调节因子进行敏感性分析，并建立这些因子的空间数据库。
其中选择用历史遥感抽取历史发生洪水的样本点信息并选取和洪水发生有关的数项因子包括：大气，蒸发，地形，河网，并从该4项因子中提出10项洪水风险敏感性评估指标包括高程，坡度，曲率，TWI，SPI，距河流距离，土壤，植被，坡向和降雨。根据流域洪水的机理，将这些因素均基于ArcGIS软件进行计算和处理。其中SPI和TWI使用以下公式计算：
TWI＝Ln(α/tanβ)       (1)
SPI＝A s tanβ        (2)
式中α是通过一个点的累计坡面排水量，A s为特定流域面积，tanβ是该点处的坡度 角。
步骤二、对所收集的初始数据进行清洗并标准化处理，统一坐标系；对步骤一所述原始样本数据进行标准化处理并赋值给每个评价单元，并转换成栅格数据存储格式，所有的数据都经过投影转换与重采样操作。由于流量洪峰是引发洪涝灾害最主要的原因，所以对于每一个研究区域，从它对应的水文站获取历史流量数据后，找出每年的洪水流量峰值日期，并选择对应日期的MODIS影像来反映该洪水过程的淹没状况。利用ENVI5.3提取洪水淹没范围，并将流量峰值对应的数个影像反映的淹没范围叠置，生成一个合并的最大的淹没范围图，作为流量峰值所对应的淹没范围图，即得到最大淹没范围。在最大淹没范围内随机选取数量为N的洪水淹没样本点，在非最大洪水淹没范围内随机选取数量为N的非洪水淹没样本点，共同构成总数为2N的样本点。将上述样本点分为训练集和测试集，其中70％的样本点作为训练集，30％的样本点作为测试集。该样本点选取方法利用了历史遥感技术提取最大淹没范围图，具有直观性好，准确性高的优点，同时也避免了大量的人工数据收集工作，大大提高了效率。
步骤三、计算拉普拉斯得分确定最后的特征子集：利用拉普拉斯得分对步骤二中所述的训练集样本的特征进行打分，得到每一个特征的分数，最后再取分数最高的k个特征作为选择的特征子集；对步骤二中总数为2N的样本点进行特征子集的抽取，形成新的训练集和测试集。如图2所示，是本发明的一种实施例的计算拉普拉斯得分流程图。
其具体方法包括：针对步骤二中训练集样本，构建一张邻接矩阵G(当type(i)＝type(j)时，G ij＝1，否则G ij＝0)，然后对于矩阵中G ij＝1的点，令
Figure PCTCN2022078765-appb-000004
(其中t为合适的常数)，这样得到的矩阵就是该训练集的权重矩阵S，其中
Figure PCTCN2022078765-appb-000005
进一步，计算拉普拉斯得分，计算公式为：
Figure PCTCN2022078765-appb-000006
其中：
L r为第r个特征的拉普拉斯得分；
f ri-f rj为第i个样本和第j个样本的第r个特征的差值；
S ij为权重矩阵中的对应的值；
Var(f r)为第r个特征在所有样本上的方差；
至此，每一个特征都会打出一个分数，最后再取分数最高的k个特征作为最后选择的特征子集。利用拉普拉斯得分法可以直观的看出各个洪水影响因子对评估结果影响的重要程度，在整体进行了洪水风险评估之后，操作人员可以直接对对结果影响程度较高的影响因子进行预防，相比于传统的人工判断，拉普拉斯得分法大大降低了操作门槛。
步骤四、利用步骤三中新的训练集对集成学习LightGBM模型进行训练；得到集成学习LightGBM模型在新的训练集和测试集的准确率。LightGBM(轻量级梯度提升树)是一个基于传统机器学习模型GBDT(梯度下降树)的一种集成学习提升方法，它有效的降低了算法运算的复杂度，与传统的集成学习方法相比，LightGBM主要采用了GOSS(单边梯度采样)方 法，根据样本采样结果计算梯度。GOSS在对待梯度较大的样本时，保留全部样本，在对待梯度较小的样本时，GOSS对样本进行随机抽样。采用的GOSS算法主要流程如下：
首先GOSS使用决策树学习获得一个将输入空间映射到梯度空间的函数。假设步骤三利用拉普拉斯得分法得到的特征子集共有n个实例，特征维度为s，每次进行梯度迭代时，LightGBM模型损失函数的负梯度方向表示为g1，…，g n，决策树通过最优切分点(最大信息增益点)将样本数据划分到各个叶结点，特征j的分割点d定义为：
Figure PCTCN2022078765-appb-000007
其中n O＝∑I[x i∈O],
Figure PCTCN2022078765-appb-000008
O表示某个固定节点的训练集。
接着GOSS根据梯度训练降序排序，保留topa个样本实例，作为数据子集A。对剩下的小梯度样本，随机采样大小为b的数据子集B，随后将数据集A和B合并。
最后通过公式(5)估计信息增益：
Figure PCTCN2022078765-appb-000009
经过一次GOSS计算，训练出一个弱分类器；接着重复GOSS算法训练多个弱分类器，直到公式(5)收敛或到达迭代步数，最后将所有训练好的弱分类器信息增益相加，得到最终的集成学习模型，并且得到集成学习LightGBM模型在新的训练集和测试集的准确率。
步骤五、利用训练好的模型对整个流域进行计算，得到整个流域洪水风险敏感性的概率值，为了对洪水易感位置进行可视化解释，需要将概率图分类为不同区域。为了进行分类，研究中存在各种方法，例如等间隔，分位数，标准差。其中针对洪水流域使用分位数方法一般可获得最佳输出，由此获得洪水风险敏感性图，并将洪水灾害风险研究区域分为五类洪水敏感性：低风险区、较低风险区、中等风险区、高风险和极高风险地区。
为了验证本发明方法的可行性，选取黄河流域的三门峡到花园口为研究区，利用从水文书籍中记载的历史洪水数据获取MODIS遥感图像，从而得到研究区的最大淹没范围，并从中随机取样。研究区总共选取了300个淹没样本点和300个非淹没样本点，其中70％作为训练集，30％作为测试集。研究区选取了高程、坡度、坡向、曲率、SPI、TWI、距河流距离、土壤、植被和降雨量总共10个洪水影响因子，并分别计算了个洪水影响因子的拉普拉斯得分，计算结果如表一所示。
在进行模型训练时，本发明选取了LightGBM和市面上主流的集成学习方法XGBoost进行对比试验，经过对比试验，发现XGBoost的准确率为80.97％，LightGBM的准确率为81.29％，并且运行的速度要远高于XGBoost。
将研究区数据都输入到LightGBM模型中，生成洪水敏感性概率图并按照分位数方法，将概率图分为极高风险、高风险、中风险、低风险和极低风险五类，其试验结果如图3所 示。
表1
Figure PCTCN2022078765-appb-000010

Claims (4)

  1. 一种基于集成学习的洪水敏感性风险评估方法，其特征在于，包括以下步骤：
    步骤一、样本点初始数据的收集与整理：通过使用文献资料和实地调查绘制流域的洪水位置图并创建与洪水有关的空间数据库；并通过从文献中获得的数据以及现场调查，选择调节因素；选择数个洪水调节因子进行敏感性分析，并建立这些因子的空间数据库；
    步骤二、对所收集的初始数据进行清洗、标准化处理并赋值给每个评价单元，转换成栅格数据存储格式，所有的数据都经过投影转换与重采样操作；对于每一个研究区域，从它对应的水文站获取历史流量数据，找出每年洪水流量峰值日期，并选择对应日期的MODIS影像来反映该洪水过程的淹没状况；将流量峰值对应的数个影像反映的淹没范围叠置，生成一个合并的最大的淹没范围图，作为流量峰值所对应的淹没范围图，即最大淹没范围；在最大淹没范围内随机选取数量为N的洪水淹没样本点，在非最大洪水淹没范围内随机选取数量为N的非洪水淹没样本点，共同构成总数为2N的样本点；将上述样本点分为训练集和测试集，其中70％的样本点作为训练集，30％的样本点作为测试集；
    步骤三、计算拉普拉斯得分确定最后的特征子集：利用拉普拉斯得分对步骤二中所述的训练集样本的特征进行打分，得到每一个特征的分数，最后再取分数最高的k个特征作为选择的特征子集；对步骤二中总数为2N的样本点进行特征子集的抽取，形成新的训练集和测试集；
    步骤四、利用步骤三中新的训练集对集成学习LightGBM模型进行训练；得到集成学习LightGBM模型在新的训练集和测试集的准确率；
    步骤五、利用训练好的模型对整个流域进行计算，得到整个流域洪水风险敏感性的概率值；
    步骤一中所述的数个因子包括：大气、蒸发、地形、河网；可从该4项因子中提出10项洪水风险敏感性评估指标即特征包括高程、坡度、曲率、TWI、SPI、距河流距离、土壤、植被、坡向和降雨；根据流域洪水的机理；将这些因素均基于ArcGIS软件进行计算和处理，其中SPI和TWI使用以下公式计算：
    TWI＝Ln(α/tanβ)     (1)
    SPI＝A s tanβ       (2)
    式中α是通过一个点的累计坡面排水量，A s为特定流域面积，tanβ是该点处的坡度角。
  2. 根据权利要求1所述的一种基于集成学习的洪水敏感性风险评估方法，其特征在于，所述步骤二中初始数据标准化处理，包括：
    对样本数据集合S进行数据清洗，去除有缺失和不需要的数据并进行关联性验证；
    所有尺度条件因子均使用流行的分位数方法进行分类；准备好数据集后，将每个条件因子转换为m*n大小的网格空间数据库，并构造流域地区的网格图。
  3. 根据权利要求1所述的一种基于集成学习的洪水敏感性风险评估方法，其特征在于，步骤三中所述的计算拉普拉斯得分确定最后特征子集的过程包括：
    针对步骤二中训练集样本，构建一张邻接矩阵G：当type(i)＝type(j)时，G ij＝1，否则G ij＝0)，然后对于矩阵中G ij＝1的点，令
    Figure PCTCN2022078765-appb-100001
    其中t为合适的常数；
    由此得到的矩阵就是该训练集的权重矩阵S，其中
    Figure PCTCN2022078765-appb-100002
    计算拉普拉斯得分的公式为：
    Figure PCTCN2022078765-appb-100003
    其中，L r为第r个特征的拉普拉斯得分；f ri-f rj为第i个样本和第j个样本的第r个特征的差值；S ij为权重矩阵中的对应的值；Var(f r)为第r个特征在所有样本上的方差。
  4. 根据权利要求1所述的一种基于集成学习的洪水敏感性风险评估方法，其特征在于，在所述步骤五中，将洪水灾害风险研究区域等级分为五级:低风险区、较低风险区、中等风险区、高风险和极高风险地区。
PCT/CN2022/078765 2021-03-02 2022-03-02 一种基于集成学习的洪水敏感性风险评估方法 WO2022184088A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US17/995,018 US20230141886A1 (en) 2021-03-02 2022-03-02 Method for assessing hazard on flood sensitivity based on ensemble learning

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110231150.3 2021-03-02
CN202110231150.3A CN112966926B (zh) 2021-03-02 2021-03-02 一种基于集成学习的洪水敏感性风险评估方法

Publications (1)

Publication Number Publication Date
WO2022184088A1 true WO2022184088A1 (zh) 2022-09-09

Family

ID=76276460

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/078765 WO2022184088A1 (zh) 2021-03-02 2022-03-02 一种基于集成学习的洪水敏感性风险评估方法

Country Status (3)

Country Link
US (1) US20230141886A1 (zh)
CN (1) CN112966926B (zh)
WO (1) WO2022184088A1 (zh)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109598310A (zh) * 2018-12-25 2019-04-09 核工业北京地质研究院 一种多因子敏感设施识别方法
CN113408776A (zh) * 2020-12-21 2021-09-17 电子科技大学 一种基于时间维特征增强的川西野火风险预警方法
CN115730829A (zh) * 2022-12-05 2023-03-03 中国水利水电科学研究院 一种罕遇洪水洪峰流量计算方法
CN115953281A (zh) * 2022-12-19 2023-04-11 贵州大学 一种城市地下空间的内涝灾害动态评估方法及系统
CN116776238A (zh) * 2023-08-25 2023-09-19 汇杰设计集团股份有限公司 一种基于多源信息水旱灾害动态风险评估方法和系统
CN117057253A (zh) * 2023-09-28 2023-11-14 中国水利水电科学研究院 基于空间离散洗牌复形进化算法的水文模型参数率定方法
CN117540830A (zh) * 2024-01-05 2024-02-09 中国地质科学院探矿工艺研究所 基于断层分布指数的泥石流易发性预测方法、装置及介质
CN117556628A (zh) * 2023-11-23 2024-02-13 郑州大学 一种智慧城市洪涝风险评估系统
CN118091657A (zh) * 2024-04-28 2024-05-28 水利部交通运输部国家能源局南京水利科学研究院 基于分类三元搭配的流域洪涝淹没范围集成识别方法及系统

Families Citing this family (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US11501304B2 (en) * 2020-03-11 2022-11-15 Synchrony Bank Systems and methods for classifying imbalanced data
CN112966926B (zh) * 2021-03-02 2022-04-22 河海大学 一种基于集成学习的洪水敏感性风险评估方法
CN115828757B (zh) * 2022-12-12 2024-02-23 福建中锐汉鼎数字科技有限公司 一种流域水位预测的泄洪量滞后特征构造及选择方法
CN116827607A (zh) * 2023-06-02 2023-09-29 广州大学 一种集成XGBoost和LightGBM模型的车载CAN总线入侵检测算法
CN117131756B (zh) * 2023-06-07 2024-05-24 长安大学 一种基于地表时序形变和孕灾背景的地裂缝易发性评价方法
CN116882741A (zh) * 2023-07-10 2023-10-13 中国水利水电科学研究院 一种超标准洪水灾害动态定量评估方法
CN117010274B (zh) * 2023-07-11 2024-05-10 中国地质科学院水文地质环境地质研究所 一种基于集成增量学习的地下水有害元素智能预警方法
CN116626685B (zh) * 2023-07-20 2023-09-29 山东大禹水务建设集团有限公司 基于机器学习的河道底泥实时监测方法及系统
CN116778395B (zh) * 2023-08-21 2023-10-24 成都理工大学 基于深度学习的山洪漫流视频识别监测方法
CN116933217B (zh) * 2023-09-19 2024-01-30 广州华水生态科技有限公司 基于数字孪生和ai技术的中小河流洪水预报方法
CN117436619B (zh) * 2023-12-20 2024-03-15 长江水利委员会水文局 一种基于等效防洪作用的梯级水库防洪库容联合预留方法
CN117574778B (zh) * 2024-01-12 2024-03-29 河海大学 一种基于机器学习的相似场次洪水模式库构建方法
CN117634325B (zh) * 2024-01-26 2024-04-02 水利部交通运输部国家能源局南京水利科学研究院 资料受限河口区域极值事件识别和复合洪水灾害分析方法和系统
CN117709135B (zh) * 2024-02-06 2024-04-12 水利部交通运输部国家能源局南京水利科学研究院 考虑电厂温排水影响时长权重的热影响分析方法及系统
CN118015220B (zh) * 2024-04-09 2024-06-18 南京师范大学 一种基于多模态数据的城市洪涝快速制图方法

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106651211A (zh) * 2016-12-30 2017-05-10 吉林师范大学 一种不同尺度区域洪水灾害风险评估的方法
CN108280553A (zh) * 2018-02-24 2018-07-13 中山大学 基于gis-神经网络集成的山洪灾害风险区划及预测方法
EP3454285A1 (en) * 2017-09-11 2019-03-13 Fujitsu Limited An apparatus and method for assessing flooding behaviour and an apparatus and method for predicting a flood property
CN109858647A (zh) * 2018-12-21 2019-06-07 河海大学 一种耦合gis和gbdt算法的区域洪涝灾害风险评价与预估方法
CN112966926A (zh) * 2021-03-02 2021-06-15 河海大学 一种基于集成学习的洪水敏感性风险评估方法

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111047099B (zh) * 2019-12-16 2020-08-21 杭州鲁尔物联科技有限公司 一种区域性山洪风险预测方法及系统
CN111507505A (zh) * 2020-03-20 2020-08-07 苏州丰华声赫智能科技有限公司 一种水库日入量预测模型的构建方法
CN111724033A (zh) * 2020-05-14 2020-09-29 天津大学 一种基于随机集理论的洪灾风险评价与精细区划方法
CN111597758A (zh) * 2020-05-14 2020-08-28 河海大学 一种基于负相关学习的中小河流集成预报方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106651211A (zh) * 2016-12-30 2017-05-10 吉林师范大学 一种不同尺度区域洪水灾害风险评估的方法
EP3454285A1 (en) * 2017-09-11 2019-03-13 Fujitsu Limited An apparatus and method for assessing flooding behaviour and an apparatus and method for predicting a flood property
CN108280553A (zh) * 2018-02-24 2018-07-13 中山大学 基于gis-神经网络集成的山洪灾害风险区划及预测方法
CN109858647A (zh) * 2018-12-21 2019-06-07 河海大学 一种耦合gis和gbdt算法的区域洪涝灾害风险评价与预估方法
CN112966926A (zh) * 2021-03-02 2021-06-15 河海大学 一种基于集成学习的洪水敏感性风险评估方法

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109598310A (zh) * 2018-12-25 2019-04-09 核工业北京地质研究院 一种多因子敏感设施识别方法
CN113408776A (zh) * 2020-12-21 2021-09-17 电子科技大学 一种基于时间维特征增强的川西野火风险预警方法
CN113408776B (zh) * 2020-12-21 2023-03-28 电子科技大学 一种基于时间维特征增强的川西野火风险预警方法
CN115730829A (zh) * 2022-12-05 2023-03-03 中国水利水电科学研究院 一种罕遇洪水洪峰流量计算方法
CN115953281A (zh) * 2022-12-19 2023-04-11 贵州大学 一种城市地下空间的内涝灾害动态评估方法及系统
CN116776238B (zh) * 2023-08-25 2023-11-03 汇杰设计集团股份有限公司 一种基于多源信息水旱灾害动态风险评估方法和系统
CN116776238A (zh) * 2023-08-25 2023-09-19 汇杰设计集团股份有限公司 一种基于多源信息水旱灾害动态风险评估方法和系统
CN117057253A (zh) * 2023-09-28 2023-11-14 中国水利水电科学研究院 基于空间离散洗牌复形进化算法的水文模型参数率定方法
CN117057253B (zh) * 2023-09-28 2023-12-08 中国水利水电科学研究院 基于空间离散洗牌复形进化算法的水文模型参数率定方法
CN117556628A (zh) * 2023-11-23 2024-02-13 郑州大学 一种智慧城市洪涝风险评估系统
CN117556628B (zh) * 2023-11-23 2024-05-28 郑州大学 一种智慧城市洪涝风险评估系统
CN117540830A (zh) * 2024-01-05 2024-02-09 中国地质科学院探矿工艺研究所 基于断层分布指数的泥石流易发性预测方法、装置及介质
CN117540830B (zh) * 2024-01-05 2024-04-12 中国地质科学院探矿工艺研究所 基于断层分布指数的泥石流易发性预测方法、装置及介质
CN118091657A (zh) * 2024-04-28 2024-05-28 水利部交通运输部国家能源局南京水利科学研究院 基于分类三元搭配的流域洪涝淹没范围集成识别方法及系统

Also Published As

Publication number Publication date
US20230141886A1 (en) 2023-05-11
CN112966926A (zh) 2021-06-15
CN112966926B (zh) 2022-04-22

Similar Documents

Publication Publication Date Title
WO2022184088A1 (zh) 一种基于集成学习的洪水敏感性风险评估方法
CN111553236A (zh) 基于道路前景图像的路面病害目标检测与实例分割方法
CN108595414B (zh) 基于源汇空间变量推理的土壤重金属企业污染源识别方法
CN111723657B (zh) 一种基于YOLOv3与自优化的河道异物检测方法及装置
CN111401419A (zh) 基于改进RetinaNet的员工着装规范检测方法
CN109063754A (zh) 一种基于OpenStreetMap的遥感影像多特征联合分类方法
CN103679675A (zh) 一种面向水质定量遥感应用的遥感影像融合方法
CN114611834B (zh) 一种基于多维特征分析的电力发电站选址评估规划方法
CN105808665A (zh) 一种新的基于手绘草图的图像检索方法
CN117495735B (zh) 一种基于结构引导的建筑物立面纹理自动修复方法及系统
CN112347927B (zh) 基于卷积神经网络概率决策融合的高分影像建筑提取方法
Elberink et al. Detection of collapsed buildings by classifying segmented airborne laser scanner data
CN106228136A (zh) 基于聚合通道特征的全景街景隐私保护方法
CN115563493A (zh) 一种基于聚类算法划分乡村景观生态单元的方法
CN116129262A (zh) 一种面向宜机化改造的耕地适宜性评价方法及系统
CN116720984A (zh) 一种基于卫星地图技术的智慧电力审计方法
CN111476197A (zh) 基于多源卫星遥感影像油棕识别及面积提取的方法和系统
CN114387261A (zh) 一种适用于铁路钢桥螺栓病害的自动检测方法
CN112347926B (zh) 基于建筑形态分布的高分辨率影像城中村检测方法
Nelson et al. Spatial statistical techniques for aggregating point objects extracted from high spatial resolution remotely sensed imagery
Engstrom et al. Evaluating the Relationship between Contextual Features Derived from Very High Spatial Resolution Imagery and Urban Attributes: A Case Study in Sri Lanka
CN114612787B (zh) 一种尺度变化策略支持的城市绿地深度学习提取方法
CN115457386A (zh) 一种村庄用地信息化生成方法
NL2030915B1 (en) Time-space dynamic monitoring method of county soil erosion based on remote sensing and gis
CN112966633B (zh) 一种语义与结构信息双约束的倾斜影像特征点过滤方法

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22762551

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22762551

Country of ref document: EP

Kind code of ref document: A1