Uncertainty optimization method for building disaster damage assessment coupled with multi-source data and random distribution

Abstract

The application discloses a kind of building disaster loss evaluation uncertainty optimization method of multi-source data and random distribution coupling, comprising the following steps: collecting the historical flood data of different cities, different types of buildings;The evaluation area is divided into grid;High-resolution water depth data of each grid unit is simulated and generated using HiPIMS model, and disaster loss data of each grid is generated by combining remote sensing data verification;The cost of different types of buildings is generalized using building cost data in city statistical yearbook;Building flood vulnerability is evaluated based on the gridding method;The evaluation uncertainty of building flood vulnerability is quantified using confidence level analysis method.The application solves the problem of multi-source data fusion and building accurate classification, unifies the scale difference of flood characteristics and socio-economic data, introduces confidence level analysis method to quantify uncertainty, significantly improves the evaluation accuracy and reliability, and provides scientific support for urban disaster prevention and mitigation, risk management and planning.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of urban building disaster damage assessment, in particular to a building disaster damage assessment uncertainty optimization method coupled with multi-source data and random distribution. BACKGROUND

[0002] The existing urban building disaster damage assessment technology has the following problems: first, the data source is single, lacking multi-source data fusion, resulting in inaccurate building classification; second, the loss rate estimation method is too simple, without considering the uncertainty of the water depth-loss rate curve, making it difficult to accurately reflect the actual loss; third, the spatial resolution is insufficient, unable to accurately quantify the spatial distribution of building disaster damage; in addition, high-performance hydrodynamic models are not applied enough, with low calculation efficiency, making it difficult to obtain accurate real-time submerged water depth information; at the same time, the assessment is highly dependent on historical disaster data, limiting the applicability of the method in data-scarce cities; finally, the building classification method is simple, without fully considering the loss differences of different functional buildings, and the model lacks a reliable verification mechanism, making it difficult to guarantee the reliability and applicability of the assessment results. These problems restrict the precision and decision support capability of disaster damage assessment, and need to be optimized and improved. SUMMARY

[0003] In view of the above technical problems in the related art, the present application provides a building disaster damage assessment uncertainty optimization method coupled with multi-source data and random distribution, which can solve the above problems.

[0004] To achieve the above technical purposes, the technical scheme of the present application is as follows:

[0005] A building disaster damage assessment uncertainty optimization method coupled with multi-source data and random distribution, comprising the following steps:

[0006] S100, collecting historical flood disaster data of different types of buildings in different cities, including water depth, flood loss rate and attribute feature information of buildings, and cleaning, deduplicating and standardizing the data, and in the process of collecting building disaster damage data, combining electronic map data, POI data, remote sensing data and social and economic survey data information, using multi-source data fusion technology to classify the functions of buildings;

[0007] S200, multi-level spatial grid division is performed on the region;

[0008] S300, water depth data of each grid cell is obtained by using a HiPIMS model, and disaster damage data of each grid is generated by combining remote sensing data verification;

[0009] S400, building cost estimation is performed using building cost data of different types of buildings in the urban statistical yearbook;

[0010] S500, performing building flood vulnerability assessment based on a gridding method;

[0011] S510, determining a loss rate function of each type of building under different water depths;

[0012] S520, estimating the cost value of each type of building and estimating the area proportion of each type of building in the grid;

[0013] S530, counting the total loss of buildings in the grid and estimating the flood loss of the entire region;

[0014] S600, using a confidence level analysis method to quantify the uncertainty of building flood vulnerability and improve the reliability of disaster loss assessment.

[0015] Further, the attribute feature information of the building in S100 includes the location, type, area and height of the building.

[0016] Further, the building function classification in S100 specifically includes the following steps:

[0017] S110, using high-resolution remote sensing images and GIS tools, combining image recognition algorithms and visual interpretation methods, extracting buildings in the selected area, and assigning each building a unique identification number for spatial distribution analysis;

[0018] S120, classifying based on building attribute information of the electronic map;

[0019] S130, screening and deduplicating the collected POI data, and according to the land use type and function classification standard, using the POI data frequency density analysis method to assign preliminary function attributes to buildings that cannot be determined by the electronic map label;

[0020] S140, for the buildings that cannot be classified above, using kernel density estimation method for function classification, the kernel density estimation method uses Gaussian distribution kernel function to smooth the POI data around the building, and estimates the function distribution probability of the building.

[0021] Further, the kernel function of building function classification in S140 is two-dimensional Gaussian distribution, and the expression is as follows:

[0022]

[0023] In the formula, the coefficient A represents the amplitude of the Gaussian function, x0 and y0 represent the position coordinates of the center point, and σ x , σ y respectively represent the variance in x and y directions, which affects the search / diffusion radius;

[0024] After the kernel function is determined, four kernel density layers are constructed according to the known four categories of buildings, and for the unknown category building, the sum of the kernel density layers in the region of the unknown category building is calculated, and the unknown category building is classified into the category with the highest density value, and the mathematical expression is as follows:

[0025]

[0026]

[0027]

[0028]

[0029] In the formula, K1, K2, K3, K4 represent the kernel density values of each category of buildings, N b represents the number of pixels in the building area, K 1i , K 2i , K 3i , K 4i represent the kernel density values of the corresponding pixels on the kernel density layer of each category of buildings.

[0030] Further, the unit area cost value of each category of buildings in S400 is calculated by the following formula:

[0031]

[0032] In the formula, θ i，j represents the unit area cost value of the i-th category of buildings in the j-th administrative region, N i，j represents the number of pixels of the i-th category of buildings in the j-th administrative region, A k represents the area of each pixel, G j represents the cost value of the j-th administrative region.

[0033] Further, in S510, the discrete water depth-loss rate curve points are fitted by using the curve fitting toolbox of Matlab, and the most suitable fitting function is selected according to the curve characteristics of different types of buildings.

[0034] Further, it is assumed that the building loss rate obeys Gaussian distribution, and the maximum gap of the water depth-loss rate curve of the same type of building at the same water depth is used for estimation.

[0035] Further, the Monte Carlo simulation method is adopted, a large number of experiments are simulated under each water depth condition to simulate the possible loss rate distribution, and then the maximum probability value method is used to select the loss rate value with the highest probability as the final estimated value.

[0036] Further, the total flood loss of the building in each grid in S530 is calculated by the following formula:

[0037]

[0038] C in the formula b represents the total loss estimation value of buildings in the grid under flood, that is, the flood vulnerability of the grid, θ i represents the value cost per unit area of the i-th type of building, f i (d) represents the loss rate of the i-th type of building under water depth h, N i represents the number of pixels occupied by the i-th type of building, A k represents the area of each pixel.

[0039] Further, S600 specifically comprises the following steps:

[0040] S610, calculating the vulnerability index of the building according to the loss value and water depth data of each building under different flood scenarios;

[0041] S620, performing confidence level analysis on the vulnerability index of each building through Monte Carlo simulation and confidence interval calculation;

[0042] S630, combining the loss data and vulnerability index of different types of buildings to perform loss quantification of urban buildings.

[0043] The application has the following beneficial effects: the application comprehensively utilizes multi-source data to accurately classify building functions, greatly improving the accuracy of disaster loss assessment; through the grid method, the flood characteristics and social and economic data are unified, the inconsistency problem of different scale data is solved, and accurate assessment of building disaster loss is realized. In the analysis of urban flood building vulnerability, the Matlab curve fitting tool is used to generate a water depth-loss rate mean curve, and the Monte Carlo simulation is used to simulate the loss rate distribution under different water depth conditions, and the maximum probability value method is used to determine the final estimation result. At the same time, combined with the city grid division, histogram statistical analysis and confidence level analysis of building flood vulnerability, the uncertainty in the disaster loss assessment process is effectively reduced, and the scientificity and applicability of the flood disaster loss assessment are further improved. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings described below are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0045] The present application will be further described in detail below according to the drawings.

[0046] Fig. 1It is a flow chart of a building disaster loss assessment uncertainty optimization method of multi-source data and random distribution coupling according to an embodiment of the present application.

[0047] Fig. 2 It is a principle diagram of functional classification of buildings that cannot be classified using a kernel density estimation method according to an embodiment of the present application.

[0048] Fig. 3 It is a schematic diagram of building flood vulnerability assessment based on a gridding method according to an embodiment of the present application. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application belong to the scope of protection of the present application.

[0050] As shown in Figs. 1-3 According to the present application, a building disaster loss assessment uncertainty optimization method of multi-source data and random distribution coupling is disclosed, which comprises the following steps: S100, collecting historical flood disaster data of different types of buildings in different cities, including water depth, flood loss rate and attribute characteristic information of the buildings, and cleaning, deduplicating and standardizing the data, and in the process of collecting building disaster loss data, combining electronic map data, POI data, remote sensing data and social and economic survey data, using multi-source data fusion technology to classify the functions of the buildings; S200, dividing the region into multiple levels of grids; S300, using a HiPIMS model to simulate and obtain water depth data of each grid unit, and combining remote sensing data to verify and generate disaster loss data of each grid; S400, using building cost data of different types of buildings in a city statistical yearbook to estimate the cost of the buildings; S500, performing building flood vulnerability assessment based on a gridding method; S510, determining the loss rate function of each type of building under different water depths; S520, estimating the cost value of each type of building and estimating the area proportion of each type of building in the grid; S530, after counting the total loss of the buildings in the grid, estimating the flood loss of the entire region; S600, using a confidence level analysis method to quantify the uncertainty of the building flood vulnerability, and improving the reliability of the disaster loss assessment.

[0051] Embodiment one:

[0052] When using multi-source data fusion technology to classify the functions of the buildings, the following technical means are specifically included:

[0053] (1) Spatial distribution information extraction

[0054] Using 0.5-meter resolution remote sensing images and GIS tools, combined with convolutional neural networks (CNN) and visual interpretation, automatically extract buildings in the disaster area, and through artificial correction of misjudgment and omission, ensure the accuracy and completeness of the results. Each building is assigned a unique number to facilitate spatial distribution analysis.

[0055] (2) Electronic map label classification

[0056] Using web crawlers and map data parsing tools, extract building attribute labels from electronic maps, and classify them according to pre-set classification rules and keyword libraries.

[0057] (3) POI data classification

[0058] Screen and de-duplicate the collected POI data, according to the land use type and function classification standard, combined with the number of surrounding POI data, to assign the corresponding function to the buildings that cannot be determined by the electronic map label.

[0059] (4) Kernel density estimation classification

[0060] For buildings that cannot be classified, kernel density estimation (KDE) can be used to classify functions based on surrounding POI data. Using a Gaussian kernel function to smooth local features and extend to the global, by setting the appropriate diffusion radius to calculate the POI density, to infer the building function, such as residential areas densely distributed residential POIs, industrial areas concentrated distribution of industrial POIs, to improve the classification accuracy and reduce data uncertainty.

[0061] The principle formula of the kernel density estimation method is as follows:

[0062]

[0063] In the formula, K(x) represents the kernel function (Kernel), which is a non-negative function, R represents the neighborhood radius or search radius, x i is an element with known attributes, N k represents the number of known elements in the kernel density estimation.

[0064] According to the above conditions, set the kernel function of building function classification as a two-dimensional Gaussian distribution, the expression is as follows:

[0065]

[0066] In the formula, the coefficient A represents the amplitude of the Gaussian function, x0, y0 represents the position coordinates of the center point, σ x , σ y respectively represent the variance in the x and y directions, which affects the search / diffusion radius.

[0067] According to the kernel function, four kernel density layers of residential, commercial, industrial and public buildings are constructed, and machine learning algorithm (decision tree) is used to classify the buildings by function. Through K-means clustering, data selection and sample expansion are optimized to ensure that more building types and functional attributes are covered. For unknown classification, according to the sum of each kernel density layer within the building area, it is classified as the one with the largest sum, and the mathematical expression is as follows:

[0068]

[0069]

[0070]

[0071]

[0072] In the formula, K1, K2, K3, K4 represent the kernel density values of each type of building, N b represents the number of pixels in the building area, K 1i , K 2i , K 3i , K 4i represent the kernel density values of the corresponding pixels on each type of building kernel density layer.

[0073] To solve the problem of inconsistent scales between flood feature grids and social and economic data, a gridding method is used to divide the region into multiple levels of regular grids, and different types of spatial data (flood features, social and economic attributes) are unified for comprehensive analysis. For flood feature data, the water depth data of each grid cell is obtained by using the HiPIMS model (high resolution flood simulation model) and multi-source remote sensing data.

[0074] Example two:

[0075] When selecting vulnerability samples, in order to reduce the error of building flood loss estimation, according to the central limit theorem, cities with similar size and economic development level should be selected. Using the building cost data of different types of buildings in the city statistical yearbook, the building cost is generalized, and the accuracy of building cost estimation can be improved by using big data technology combined with other related information. The unit area cost value of each type of building is calculated according to the following formula:

[0076]

[0077] In the formula, θ i，j represents the unit area cost value of the i-th type of building in the j-th administrative region, N i，j represents the number of pixels of the i-th type of building in the j-th administrative region, G j represents the cost value of the j-th administrative region.

[0078] In the building flood vulnerability assessment based on the gridding method, the region is divided into regular grids (square or rectangular), and the building type, area, cost value and water depth information are integrated in each grid. The building loss rate is calculated through the water depth-loss rate curve, and the loss value is calculated by combining the cost value and area weighting. Finally, the loss values of each grid are added up to estimate the flood loss of the whole region.

[0079] The discrete water depth-loss rate data points of different building types are fitted using the Matlab curve fitting tool, and the appropriate function is selected to reduce the influence of curve discreteness and provide an accurate basis for loss rate estimation.

[0080] Further, due to the uncertainty of the water depth-loss rate curve, it is assumed that the building loss rate follows a Gaussian distribution at a certain water depth, with the mean vulnerability as the center, and the variance is estimated according to the maximum gap of the curve. Monte Carlo simulation is used to simulate the loss rate, and the distribution rule of all possible loss rate values is analyzed through histogram statistical analysis to generate the probability density distribution, and the highest probability value is selected as the final estimated value.

[0081] The building loss rate function at different water depths is determined, and the cost value of the building and the area proportion in the grid are combined to calculate the total loss of the building in each grid. Based on the minimum spatial scale grid unit, the total loss of the region is calculated by the grid tool. According to the loss rate and unit area cost value of each type of building calculated by the derivation, the total flood loss of the building in each grid is calculated according to the following formula:

[0082]

[0083] Where C b represents the total loss estimate of the building flood in the grid, i.e. the grid flood vulnerability, Ω i represents the value cost per unit area of the i-th building, f i (h) represents the loss rate of the i-th building at water depth h, ψ i represents the area proportion of the i-th building in the grid.

[0084] In order to facilitate the grid loss calculation, the formula is optimized as:

[0085]

[0086] Where θ i represents the value cost per unit area of the i-th building, f i (d) represents the loss rate of the i-th building at water depth h, N i represents the number of pixels occupied by the i-th building, A k represents the area of each pixel.

[0087] In order to further improve the reliability of disaster damage assessment, the confidence level analysis method is used to quantify the uncertainty of building flood vulnerability. The specific steps are as follows: generate loss estimate value through Monte Carlo simulation, and combine histogram statistics to construct probability density function to determine the confidence within a certain confidence interval. The specific steps include determining the simulation method of random events, selecting the confidence interval (95%), distributing the statistical results and calculating the confidence, and finally obtaining the threshold range (maximum and minimum) of the loss. This method can quantify the uncertainty of the model, intuitively describe the reliability of the loss estimate, provide a scientific basis for disaster prevention and mitigation policy, and be suitable for different building types and water depth conditions, achieving flexible and accurate risk prediction.

[0088] In summary, through urban area grid division, building flood loss rate estimation based on histogram statistics and confidence level analysis of building flood vulnerability, the uncertainty of urban disaster damage assessment is reduced, and its accuracy and applicability are improved.

[0089] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for optimizing uncertainty in building disaster damage assessment coupled with multi-source data and random distribution, characterized in that, Includes the following steps: S100. Collect historical flood data for different types of buildings in different cities, including water depth, flood loss rate, and building attribute characteristics. Clean, deduplicate, and standardize the data. Combine electronic map data, POI data, and remote sensing data, and use multi-source data fusion technology to classify building functions. Specifically, this includes: S110. Using high-resolution remote sensing imagery and GIS tools, combined with image recognition algorithms and visual interpretation methods, buildings within the selected area are extracted, and each building is assigned a unique identification number for spatial distribution analysis. S120. Classify the attribute feature information of buildings based on electronic maps; S130. The collected POI data is screened and deduplicated. Based on the land use type and functional classification standards, the frequency density analysis method of POI data is used to assign preliminary functional attributes to buildings whose functions cannot be determined by electronic map labels. S140. For buildings that cannot be classified, the kernel density estimation method is used for functional classification. The kernel density estimation method uses a Gaussian distribution kernel function to smooth the POI data around the building and estimate the functional distribution probability of the building. After determining the kernel function, four kernel density layers are constructed based on the four known building categories. For buildings of unknown category, the sum of the kernel density layers within the area of ​​the unknown category building is calculated, and the unknown category building is assigned to the category with the highest density value. The mathematical expression is as follows: ， In the formula, K1, K2, K3, and K4 represent the kernel density values ​​of various types of buildings, and N b K represents the number of pixels within the building area. 1i K 2i K 3i K 4i This represents the kernel density value of the corresponding pixel on the kernel density layer for various types of buildings; S200, Divide the region into multi-level spatial grids; S300: The HiPIMS model is used to simulate and obtain the water depth data of each grid cell, and the disaster damage data of each grid cell is generated by combining remote sensing data. S510. Using the curve fitting toolbox in Matlab, fit the discrete water depth-loss rate curve points to obtain the loss rate function at different water depths. S520. Estimate the unit area cost value of each type of building and estimate the area proportion of each type of building in the grid. S530. Calculate the total building losses within each grid and estimate the flood losses for the entire area. The formula for calculating the total building losses within each grid is as follows: ， In the formula C b θ represents the total loss of buildings within the grid, i.e., the grid's flood vulnerability. i f represents the value cost per unit area of ​​the i-th type of building. i (d) represents the loss rate of the i-th type of building at a water depth of h, N i A represents the number of pixels occupied by the i-th type of building. k This represents the area of ​​each pixel; S600. Confidence level analysis is used to quantify the uncertainty of building flood vulnerability, thereby improving the reliability of disaster damage assessment.

2. The uncertainty optimization method for building disaster damage assessment coupled with multi-source data and random distribution as described in claim 1, characterized in that, The attribute characteristics of a building in S100 include its location, type, area, and floor height.

3. The uncertainty optimization method for building disaster damage assessment coupled with multi-source data and random distribution as described in claim 1, characterized in that, The kernel function for building function classification in S140 is a two-dimensional Gaussian distribution, expressed as follows: ， In the formula, coefficient A represents the amplitude of the Gaussian function, x0 and y0 represent the coordinates of the center point, and σ x σ y These represent the variances in the x and y directions, respectively, which affect the search / diffusion radius.

4. The uncertainty optimization method for building disaster damage assessment coupled with multi-source data and random distribution as described in claim 1, characterized in that, S520 calculates the unit area cost value of various types of buildings using the following formula: ， In the formula θ i,j N represents the unit area cost value of the i-th type of building in the j-th administrative region. i,j A represents the number of pixels of the i-th type of building in the j-th administrative region. k G represents the area of ​​each pixel. j This represents the cost value of the j-th administrative region.

5. The uncertainty optimization method for building disaster damage assessment coupled with multi-source data and random distribution as described in claim 1, characterized in that, Assuming that the building loss rate follows a Gaussian distribution, it is estimated based on the maximum gap at the same water depth according to the water depth-loss rate curve of similar buildings.

6. The uncertainty optimization method for building disaster damage assessment coupled with multi-source data and random distribution as described in claim 5, characterized in that, The Monte Carlo simulation method was used to simulate the possible loss rate distribution under each water depth condition through a large number of experiments. Then, the loss rate value with the highest probability was selected as the final estimate by the maximum probability value method.

7. The uncertainty optimization method for building disaster damage assessment coupled with multi-source data and random distribution as described in claim 1, characterized in that, The S600 specifically includes the following steps: S610. Calculate the vulnerability index of each type of building based on the loss rate and water depth data under different flood scenarios. S620. Through Monte Carlo simulation and confidence interval calculation, confidence level analysis is performed on the vulnerability index of each type of building. S630. Combine loss data and vulnerability indicators of different types of buildings to quantify the loss of urban buildings.

Images