Comprehensive density index-based flood inundation road disaster assessment method and system

Abstract

The invention discloses a flood inundation road disaster situation assessment method and system based on a comprehensive density index. The method comprises the following steps: firstly, performing radiation and geometric correction on high-resolution remote sensing images of a pre-disaster road and a disaster flood; secondly, respectively extracting binary road and flood masks by adopting improved U-Net and improved DeepLabv3 +, vectorizing the binary road and flood masks, and accurately obtaining a submerged road area through pixel-level AND operation; based on the submerged length and width of each section of road, a comprehensive density index is constructed; then, primarily dividing high, middle and low submerging levels by using K-means clustering, and optimizing a classification boundary through an SVM model to improve the evaluation precision; and finally, automatic data loading, deep learning reasoning, spatial analysis and visualization are realized. The method can comprehensively reflect the influence of road geometric features on flood inundation disasters, and provides scientific and accurate support for road emergency rescue decisions after flood disasters.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of road disaster assessment, and relates to a flood submerged road disaster condition assessment method and system based on a comprehensive density index, which is suitable for rapid assessment of damage degree of road infrastructure in flood disasters and decision support for emergency rescue. BACKGROUND

[0002] With the continuous expansion of the scale of highway network in China, the safety and emergency response capability of highway traffic are facing severe challenges, especially under the background of frequent extreme weather and intensified geological disasters such as landslides and debris flows, the number of road sudden disaster events has increased significantly. At present, the disaster condition assessment of flood submerged roads at home and abroad mainly relies on the following technical methods: one is a visual interpretation method based on remote sensing images, which compares and analyzes the remote sensing images before and after the disaster by professional personnel to identify the submerged area and damaged road; this method has the disadvantages of great subjectivity, low efficiency and precision depending on the experience of interpreters, and it is difficult to meet the needs of rapid assessment of large-scale disasters. Another is a statistical index-based assessment method, which mainly uses the following two density indexes: length density index: , which only considers the ratio of submerged roads to total road length; and area density index: , which only considers the ratio of submerged roads to total road area; the traditional road flood disaster condition assessment method has single limitation and lacks comprehensive consideration of road geometric characteristics; the density index description is insufficient and cannot reflect the actual submerged influence under the joint action of road width and length; the weight distribution is unreasonable and lacks objective grading standards based on historical data, and the division of disaster severity often depends on expert experience and lacks scientific basis; and the calculation efficiency is low.

[0003] Therefore, it is urgent to establish a flood submerged road disaster condition assessment method that can comprehensively consider the influence of submerged road length, width and other factors and accurately reflect the rescue difficulty and disaster severity. SUMMARY

[0004] In view of the deficiencies in the prior art, the purpose of the present application is to provide a flood submerged road disaster condition assessment method and system based on a comprehensive density index, which realizes scientific grading of regional road submerged state by establishing a density index model that comprehensively considers submerged length and width, and accurately reflects the rescue difficulty and disaster severity.

[0005] To solve the above technical problems, the present application adopts the following technical solutions: A flood submerged road disaster condition assessment method based on a comprehensive density index, comprising the following steps: Step 1, Data Acquisition: Collect pre-disaster road datasets and flood datasets for the study area, and perform radiometric and geometric correction preprocessing; the pre-disaster road datasets include pre-disaster road remote sensing images and pre-disaster road masking data; the flood datasets include flood remote sensing images and flood masking data. Step 2, Feature Extraction: An improved U-Net network is used to extract pre-disaster road features from the pre-processed pre-disaster road remote sensing images, obtaining the spatial distribution and geometric features of the pre-disaster roads; an improved DeepLabv3+ network is used to extract disaster-time flood features from the pre-processed disaster-time flood remote sensing images, obtaining the spatial distribution and geometric features of the disaster-time roads; the spatial distribution and geometric features include the length and width information of the pre-disaster and disaster-time roads; Step 3, Data Fusion: Based on the spatial distribution and geometric features obtained in Step 2, the pre-disaster road masking data and the flood masking data during the disaster are overlaid using a raster overlay analysis, and logical AND operations are performed. The raster data is then converted into vector data to obtain the flooded road mask matrix, where A is the road mask matrix, B is the flood mask matrix, and C is the flooded road mask matrix; i and j represent the row index and column index in the matrix, respectively. Step 4, Density Index Calculation: Using the flooded road mask matrix obtained in Step 3, establish the comprehensive density index:

[0006] Where α is the length weight parameter, β is the width weight parameter, and α and β are determined by fitting historical data using a nonlinear least squares method; n is the total number of submerged road sections; l i Let l be the length of the i-th submerged road segment; (n+i) w is the length of the i-th segment of the road that was not submerged; i Let be the width of the i-th road segment; Step 5, graded assessment: The K-means clustering algorithm is used to cluster the comprehensive density index into 3 classes. Combined with the SVM algorithm, the classification boundary is established to divide the degree of road flooding in the area into three levels: high, medium and low. Step 6, Disaster Early Warning: Automated data loading, image analysis and extraction, data fusion, spatial data extraction, and flooding degree assessment are carried out to achieve rapid extraction and assessment of flood-inundated road information, comprehensively reflecting the impact of road geometric features on flood disasters, and providing scientific and accurate support for road emergency rescue decision-making after flood disasters.

[0007] The present invention also includes the following technical features: Specifically, the improved U-Net network uses ResNet-50 as the encoder backbone, adds an attention gating mechanism at skip connections, employs a dense connection structure in the decoder, and uses a combination of Dice loss and cross-entropy loss as the loss function. , where λ1 and λ2 are weighting coefficients.

[0008] Specifically, the backbone network of the improved DeepLabv3+ network adopts an improved ResNet, in which the ASPP module uses multi-scale dilated convolution and the decoder adopts a feature fusion mechanism.

[0009] Specifically, the raster overlay analysis involves performing a logical AND operation on each pixel location. When the pixel values ​​of both the road mask and the flood mask are 1, the corresponding position of the flooded road mask is 1; otherwise, it is 0. The process of converting the raster data into vector data includes: region segmentation, boundary extraction, and topology reconstruction.

[0010] Specifically, in step 4, the road area and length are calculated using the polygon area calculation formula and the Euclidean distance formula, respectively, and α and β are determined by fitting historical data using the nonlinear least squares method; wherein, the nonlinear least squares parameter optimization adopts the Levenberg-Marquardt algorithm.

[0011] Specifically, the objective function of the K-means clustering algorithm is: Where SSE is the sum of squared errors; K is the number of clusters; C i For the i-th cluster; μ i Let x be the cluster center of the i-th cluster; x is the value belonging to cluster C. i Data points; |x - μ i |² represents the relationship between data point x and cluster center μ. i The squared Euclidean distance between them; the cluster center update formula is: , where μ i (t+1) C is the new cluster center of the i-th cluster after the (t+1)-th iteration; i (t) Let C be the set of all data points contained in the i-th cluster at the t-th iteration; i (t) | represents the number of data points in the i-th cluster at the t-th iteration; x represents the cluster C. i (t) The data points are in the middle; where K=3, corresponding to the high, medium and low flood levels respectively.

[0012] Specifically, the classification boundaries established by the SVM algorithm are x=0.320 and x=0.685, and the classification rule is: D combinedAreas with a value ≥0.685 are considered high-level flooding areas, and areas with a value ≤0.32 are considered high-level flooding areas. combined Areas with a strength <0.685 are considered moderately flooded areas. <D combined Areas with a value less than 0.32 are considered low-level flooding areas.

[0013] A flood-inundated road disaster assessment system based on a comprehensive density index, which enables the implementation of the method described above, including: The data acquisition module is used to collect pre-disaster road datasets and flood datasets for the study area. The image spatial resolution is 1m, and radiometric and geometric correction preprocessing is performed. The pre-disaster road dataset includes pre-disaster road remote sensing images and pre-disaster road masking data; the flood dataset includes flood remote sensing images and flood masking data. The feature extraction module is used to extract pre-disaster road features from pre-processed pre-disaster road remote sensing images using an improved U-Net network, obtaining the spatial distribution and geometric features of pre-disaster roads; and to extract disaster-time flood features from pre-processed disaster-time flood remote sensing images using an improved DeepLabv3+ network, obtaining the spatial distribution and geometric features of disaster-time roads; the spatial distribution and geometric features include the length and width information of pre-disaster and disaster-time roads. The data fusion module is used to perform raster overlay analysis on pre-disaster road masking data and post-disaster flood masking data based on the obtained spatial distribution and geometric features, through logical AND operations. The raster data is then converted into vector data to obtain the flooded road mask matrix, where A is the road mask matrix, B is the flood mask matrix, and C is the flooded road mask matrix; i and j represent the row index and column index in the matrix, respectively. The density index calculation module establishes a comprehensive density index using the obtained flooded road mask matrix.

[0014] Where α is the length weight parameter, β is the width weight parameter, and α and β are determined by fitting historical data using a nonlinear least squares method; n is the total number of submerged road sections; l i Let l be the length of the i-th submerged road segment; (n+i) w is the length of the i-th segment of the road that was not submerged; i Let be the width of the i-th road segment; The hierarchical assessment module is used to cluster the comprehensive density index into 3 classes using the K-means clustering algorithm, and to establish the classification boundary by combining the SVM algorithm to divide the degree of road flooding in the region into three levels: high, medium and low. The disaster early warning module is used for automated data loading, image analysis and extraction, data fusion, spatial data extraction, and flooding degree assessment. It enables rapid extraction and assessment of flood-inundated road information, comprehensively reflects the impact of road geometric features on flood disasters, and provides scientific and accurate support for road emergency rescue decision-making after flood disasters.

[0015] Specifically, the improved U-Net network in the feature extraction module includes an encoder, skip connections, and a decoder. The encoder uses ResNet-50 as the backbone network, and an attention gating mechanism is set at the skip connections. The decoder adopts a dense connection structure. The backbone network of the improved DeepLabv3+ network uses an improved ResNet, and the decoder adopts a feature fusion mechanism. The data fusion module uses raster overlay analysis, performing a logical AND operation on each pixel location. When the pixel values ​​of both the road mask and the flood mask are 1, the corresponding position value of the flooded road mask is 1; otherwise, it is 0. The process of converting raster data into vector data includes: region segmentation, boundary extraction, and topology reconstruction. The density index calculation module calculates the road area and length using the polygon area calculation formula and the Euclidean distance formula, respectively, and determines α and β by fitting historical data using the nonlinear least squares method.

[0016] Specifically, in the graded assessment module: The objective function of the K-means clustering algorithm is: Where SSE is the sum of squared errors; K is the number of clusters; C i For the i-th cluster; μ i Let x be the cluster center of the i-th cluster; x is the value belonging to cluster C. i Data points; |x - μ i |² represents the relationship between data point x and cluster center μ. i The squared Euclidean distance between them; the cluster center update formula is: , where μ i (t+1) C is the new cluster center of the i-th cluster after the (t+1)-th iteration; i (t) Let C be the set of all data points contained in the i-th cluster at the t-th iteration; i (t) | represents the number of data points in the i-th cluster at the t-th iteration; x represents the cluster C. i (t) The data points in the data; where K=3, corresponding to the high, medium and low inundation levels respectively; The SVM algorithm establishes classification boundaries of x=0.320 and x=0.685, and the classification rule is: D combined Areas with a value ≥0.685 are considered high-level flooding areas, and areas with a value ≤0.32 are considered high-level flooding areas.combined Areas with a strength <0.685 are considered moderately flooded areas. <D combined Areas with a value less than 0.32 are considered low-level flooding areas.

[0017] Compared with the prior art, the present invention has the following technical effects: (1) This invention is the first to construct a comprehensive density index that simultaneously considers the influence of road flooding length and width, overcoming the shortcomings of traditional single indicators that cannot reflect the comprehensive effect of geometric features, making the assessment results more consistent with the actual disaster characteristics.

[0018] (2) This invention combines deep learning road extraction with flood segmentation model, and uses improved U-Net and improved DeepLabv3+ to automatically obtain accurate masks of roads and floods, thereby achieving high-precision automatic identification of flooded road areas, improving evaluation efficiency and reducing manual dependence.

[0019] (3) This invention utilizes grid-vector fusion technology to accurately extract the flooded road area through logical AND operation, ensuring the geometric consistency between the road boundary and the flooded area, and significantly improving the accuracy of spatial analysis.

[0020] (4) The comprehensive density index parameters α and β proposed in this invention are obtained by nonlinear least squares optimization based on a large amount of historical data. The weight allocation is scientific and reasonable, which significantly improves the objectivity and reliability of disaster classification.

[0021] (5) This invention combines K-means clustering and SVM to establish a road flooding level classification model with clear classification boundaries and high classification accuracy, which can provide an intuitive and quantitative assessment basis for emergency rescue.

[0022] (6) The present invention constructs an automated assessment system that integrates data loading, deep learning inference, spatial analysis and visualization, which greatly improves the processing speed and automation level of flood disaster assessment and has good engineering application value. Attached Figure Description

[0023] Figure 1 This is an overall flowchart of the method of the present invention.

[0024] Figure 2 This is a result diagram of a typical scenario in the data fusion embodiment of the present invention.

[0025] Figure 3 This is a schematic diagram of the K-means clustering and SVM process.

[0026] Figure 4 This is a schematic diagram of the flooding status determination results.

[0027] Figure 5 This is a schematic diagram of the main interface of the intelligent assessment system for road flooding disasters. Detailed Implementation

[0028] This invention provides a method for assessing the disaster situation of flooded roads based on a comprehensive density index, such as... Figure 1 This includes the following steps: Step 1, Data Acquisition: Collect pre-disaster road datasets and flood datasets for the study area. The image spatial resolution is 1m, and radiometric and geometric correction preprocessing is performed. The pre-disaster road dataset includes pre-disaster road remote sensing images and pre-disaster road masking data. The flood dataset includes flood remote sensing images and flood masking data.

[0029] Step 2, Feature Extraction: An improved U-Net (ResNet-50 backbone, attention-gated, and densely connected) network is used to extract pre-disaster road features from the pre-processed pre-disaster road remote sensing images, obtaining the spatial distribution and geometric features of the pre-disaster roads; an improved DeepLabv3+ (multi-scale dilated convolutional ASPP and feature fusion decoder) network is used to extract disaster-time flood features from the pre-processed disaster-time flood remote sensing images, obtaining the spatial distribution and geometric features of disaster-time roads; the spatial distribution and geometric features include information such as the length and width of the pre-disaster and disaster-time roads.

[0030] Specifically, the improved U-Net network uses ResNet-50 as the encoder backbone, adds an attention gating mechanism at skip connections, adopts a dense connection structure for the decoder, and uses a combination of Dice loss and cross-entropy loss as the loss function. , where λ1=0.6 and λ2=0.4 are weighting coefficients.

[0031] Specifically, the backbone network of the improved DeepLabv3+ network adopts an improved ResNet, in which the ASPP module uses multi-scale dilated convolution with different dilation rates set to {6, 12, 18}. The decoder adopts a feature fusion mechanism to improve the segmentation accuracy in complex water environments.

[0032] Step 3, Data Fusion: Based on the spatial distribution and geometric features obtained in Step 2, the pre-disaster road masking data and the flood masking data during the disaster are subjected to raster overlay analysis using logical AND operations. The raster data is then converted into vector data to obtain the flooded road mask matrix, where A is the road mask matrix, B is the flood mask matrix, and C is the flooded road mask matrix; i and j represent the row index and column index in the matrix, respectively.

[0033] Specifically, a raster overlay analysis is performed, and a logical AND operation is executed on each pixel position. When the pixel values ​​of both the road mask and the flood mask are 1, the corresponding position of the submerged road mask has a value of 1; otherwise, it is 0. The AND operation can filter areas that meet specific conditions to obtain areas that simultaneously possess flood and road characteristics; this area is the flooded road area. The submerged road information extraction adopts polygon calculation theory and line segment length calculation theory.

[0034]

[0035] The meanings of each character in the above formula are as follows: A: Represents the area of ​​the road zone; pi: (xi, yi) represents the coordinates of the i-th vertex of the polygonal region; n: Number of vertices in the polygon; L: Indicates the length of the road segment; (x1, y1) and (x2, y2) are the coordinates of the two endpoints of the road segment.

[0036] The above formulas are used to calculate the polygonal area of ​​the road and the Euclidean distance of the road centerline, respectively.

[0037] Based on the above calculation theory, key indicator information of roads, floods, and flooded road areas within the study area is extracted. The results for typical scenarios in the data fusion implementation example are as follows: Figure 2 As shown.

[0038] Specifically, raster data is converted into vector data: discrete pixel information is transformed into a vector form with continuous geometric boundaries, thereby enabling refined extraction of information features in road flood areas. This is a crucial step in data processing. The process of converting raster data into vector data includes three steps: region segmentation, boundary extraction, and topology reconstruction. Region segmentation is based on homogeneity judgment, defining a region R that satisfies: ∀(x1,y1),(x2,y2)∈R, |f(x1,y1)-f(x2,y2)|<ε, where R is the defined region; (x1,y1) and (x2,y2) are the coordinates of any two pixels within region R; f(x1,y1) is the attribute value at pixel (x1,y1), and f(x2,y2) is the attribute value at pixel (x2,y2), where attribute values ​​include grayscale or color values; ε is the allowed attribute difference threshold; ∀ means "for any"; ∈ means "belongs to". Region growing, clustering, and other region segmentation methods are based on this theory, grouping pixels with similar attributes together and extracting the boundaries of each region to construct vector geometry. The theoretical basis of boundary extraction comes from image edge detection and discrete gradient calculation. Let... For the segmented region, for any pixel within the region According to the formula: .

[0039] The gradient is calculated. In the discrete case, the gradient is estimated by the difference between adjacent pixels. If the gradient magnitude exceeds a certain threshold, it can be determined that the pixel is located at the edge of the region. The final step of raster to vector conversion is to connect the extracted boundary points into a continuous geometric shape. The topological structure of the region needs to be reconstructed to ensure that the boundaries of the converted vector data are closed and do not self-intersect.

[0040] Step 4, Density Index Calculation: Using the flooded road mask matrix obtained in Step 3, establish a comprehensive density index model:

[0041] Where α is the length weight parameter and β is the width weight parameter, determined by fitting historical data using the nonlinear least squares method: α = 1.244, β = 0.802; where n is the total number of submerged road sections; l i Let l be the length of the i-th submerged road segment; (n+i) w is the length of the i-th segment of the road that was not submerged; i Let be the width of the i-th road segment; Specifically, in step 4, the road area and length are calculated using the polygon area calculation formula and the Euclidean distance formula, respectively, and α and β are determined by fitting historical data using the nonlinear least squares method; wherein, the nonlinear least squares parameter optimization adopts the Levenberg-Marquardt algorithm, and the optimization objective function is: Where S(θ) is the objective function (sum of squared residuals); θ is the parameter vector to be optimized (including α, β, etc.); N is the total number of historical data samples; z i f(x) represents the actual observed value of the i-th sample group (e.g., rescue time); i ;y i ;θ) represents the model's predicted value; x i y i The input features for the i-th sample group include the flooding length and width; the parameter update equation is: Where J is the Jacobian matrix, representing the partial derivatives of the objective function with respect to the parameters; J T is the transpose of the Jacobian matrix; λ is the damping factor used to control the step size; I is the identity matrix; Δθ is the parameter increment; y is the vector of actual observed values; f(θ) is the vector of model predicted values; when or When ||Δθ|| is reached, parameter updates stop, where ||Δθ|| represents the norm of the parameter increment (i.e., the magnitude of parameter change); ε is the set convergence threshold; θ k Let θ be the parameter value for the k-th iteration; k+1S(θ) represents the parameter value for the (k+1)th iteration. k ) and S(θ k+1 ) are the objective function values ​​for the k-th and k+1-th iterations, respectively.

[0042] More specifically, assuming L1, L2, and L represent the length of the flooded area, the length of the unflooded area, and the total length of roads in the region, respectively, and S1, S2, and S represent the area of ​​the flooded area, the area of ​​the unflooded area, and the total area of ​​roads in the region, respectively, the flood-inundated road length density index and area density index are calculated as follows:

[0043]

[0044] definition , ,right and Perform analysis and comparison, and construct:

[0045] Construct the difference expression:

[0046] After expanding the molecule and combining like terms, we get:

[0047] Setting the numerator to be greater than zero, we can solve the inequality to obtain: or At that time, B .

[0048] at this time Conversely, when w and l have other relative sizes, .

[0049] The meanings of each character in the above formula are as follows: D: Represents the road width matrix, where each element Dij represents the width contribution of the cell at that location; A: Represents the road mask matrix; B: Represents the flood mask matrix; i, j: represent the pixel row and column indices, respectively; li: represents the actual length of the i-th submerged road segment; wi: represents the actual width of the i-th road segment.

[0050] Based on this, a comprehensive density index for flooded roads, taking into account the effects of flood length and width, was constructed:

[0051] It reflects the actual difficulty of rescue and the severity of the disaster.

[0052] in, For length weight parameters ( This reflects the contribution of length to the difficulty of rescue; Width weight parameter ( This reflects the contribution of width to the difficulty of rescue operations. The formula... and Hyperparameters are determined by combining historical data.

[0053] The rescue time was fitted using the nonlinear least squares method. T With comprehensive density index The relationship, the formula to be fitted The optimal parameters are determined by minimizing the sum of squared residuals between the observed data and the model predictions. The optimization objective function is as follows: .

[0054] Measure the overall deviation between model predictions and actual observations. This is achieved by adjusting parameters. make Minimize to obtain the optimal parameter estimate .

[0055] The Levenberg-Marquardt optimization algorithm is used to solve the nonlinear least squares problem, and the parameter update equation is... Select an initial parameter value. Calculate the residual Calculate the Jacobian matrix J and solve the normal equation to obtain the parameter increment. Using formula Update parameters.

[0056] when or At that point, parameter updates stop.

[0057] Parameter estimation results are obtained through the coefficient of determination ( R 2 The mean squared error (MSE) is used for evaluation.

[0058] Using the Python programming language, the Levenberg-Marquardt optimization algorithm is employed to solve the nonlinear least squares problem. The data is fitted, and the optimized parameters are as follows: , , , b =0.666.

[0059] Therefore, select , Using length and width as weighting parameters respectively, the formula for the comprehensive density index of submerged roads is obtained:

[0060] Since the road area is in matrix form, the resulting exponential density formula will be represented in matrix form for ease of calculation and display:

[0061] Where matrices A and B are the aggregated matrices C and D, respectively. Formation.

[0062]

[0063]

[0064] C and D are calculated using the following formulas:

[0065]

[0066] The meanings of each character in the above formula are as follows: C: Represents the road flooding length matrix, where the element Cij represents the flooding length contribution of the road cell at the i-th row and j-th column position; D: Represents the road width matrix, where each element Dij represents the width contribution of the cell at that location; A: Represents the road mask matrix; B: Represents the flood mask matrix; i, j: represent the pixel row and column indices, respectively; li: represents the actual length of the i-th submerged road segment; wi: represents the actual width of the i-th road segment; α: Road length weighting parameter; β: Road width weighting parameter.

[0067] The C and D matrices are used to transform road lengths and widths into matrix space for rasterized calculation of the comprehensive density index.

[0068] Step 5, hierarchical assessment: The K-means clustering algorithm is used to cluster the comprehensive density index into 3 classes. Combined with the SVM algorithm, the classification boundary is established to divide the degree of road flooding in the area into three levels: "high", "medium" and "low".

[0069] Specifically, the objective function of the K-means clustering algorithm is: Where SSE is the sum of squared errors; K is the number of clusters (K=3 in this invention); C i For the i-th cluster; μ i Let x be the cluster center of the i-th cluster; x is the value belonging to cluster C. i Data points; |x - μ i |² represents the relationship between data point x and cluster center μ. i The squared Euclidean distance between them; the cluster center update formula is: , where μ i (t+1) C is the new cluster center of the i-th cluster after the (t+1)-th iteration; i (t) Let C be the set of all data points contained in the i-th cluster at the t-th iteration; i (t) | represents the number of data points in the i-th cluster at the t-th iteration; x represents the cluster C. i (t) The data points in the cluster; Σ represents the summation of all data points within the cluster; where K=3, corresponding to the three flooding levels of "high", "medium" and "low".

[0070] Specifically, the SVM algorithm establishes classification boundaries of x=0.320 and x=0.685, and the classification rule is: D combined Areas with a value ≥0.685 are considered "high" levels of flooding, and areas with a value ≤0.32 are considered "low" levels of flooding. combined Areas with a strength <0.685 are considered "moderately" flooded areas. <D combined Areas with a value less than 0.32 are considered "low" flooding areas.

[0071] Step 6, Disaster Early Warning: This step involves automated data loading, image analysis and extraction, data fusion, spatial data extraction, and inundation level assessment. It enables rapid extraction and evaluation of flood-inundated road information, comprehensively reflecting the impact of road geometry on flood disasters, and providing scientific and accurate support for post-flood emergency rescue decision-making. Specifically, automated data loading, deep learning inference, spatial analysis, and visualization functions are implemented on ArcGIS Engine, PyTorch, and C# / .NET platforms. Verification results show that the model has R... 2 With excellent performance of 0.986 and MSE=2.965, it can comprehensively reflect the impact of road geometry on flood inundation disasters, and provide scientific and accurate support for road emergency rescue decision-making after flood disasters.

[0072] This invention also provides a flood-inundated road disaster assessment system based on a comprehensive density index. This system can implement the above-mentioned methods and includes: a data acquisition module, a feature extraction module, a data fusion module, a density index calculation module, a hierarchical assessment module, and a disaster early warning module; specifically: The data acquisition module is used to collect pre-disaster road datasets and flood datasets for the study area. The image spatial resolution is 1m, and radiometric and geometric correction preprocessing is performed. The pre-disaster road dataset includes pre-disaster road remote sensing images and pre-disaster road masking data; the flood dataset includes flood remote sensing images and flood masking data. The feature extraction module is used to extract pre-disaster road features from pre-processed pre-disaster road remote sensing images using an improved U-Net network, obtaining the spatial distribution and geometric features of pre-disaster roads; and to extract disaster-time flood features from pre-processed disaster-time flood remote sensing images using an improved DeepLabv3+ network, obtaining the spatial distribution and geometric features of disaster-time roads; the spatial distribution and geometric features include information such as the length and width of pre-disaster and disaster-time roads; Specifically, the improved U-Net network in the feature extraction module includes an encoder, skip connections, and a decoder. The encoder uses ResNet-50 as the backbone network, and an attention gating mechanism is set at the skip connections. The decoder adopts a dense connection structure. The backbone network of the improved DeepLabv3+ network uses an improved ResNet, and the decoder adopts a feature fusion mechanism.

[0073] The data fusion module is used to perform raster overlay analysis on pre-disaster road masking data and post-disaster flood masking data based on the obtained spatial distribution and geometric features, using logical AND operations. The raster data is then converted into vector data to obtain the flooded road mask matrix, where A is the road mask matrix, B is the flood mask matrix, and C is the flooded road mask matrix; i and j represent the row index and column index in the matrix, respectively. Specifically, the data fusion module uses raster layer overlay analysis technology to perform a logical "AND" operation on each pixel position. When the pixel values ​​of both the road mask and the flood mask are 1, the corresponding position of the flooded road mask is 1; otherwise, it is 0. The process of converting raster data into vector data includes: region segmentation, boundary extraction, and topology reconstruction. The density index calculation module establishes a comprehensive density index model using the obtained flooded road mask matrix.

[0074] Where α is the length weight parameter and β is the width weight parameter, determined by fitting historical data using the nonlinear least squares method: α = 1.244, β = 0.802; where n is the total number of submerged road sections; li Let l be the length of the i-th submerged road segment; (n+i) w is the length of the i-th segment of the road that was not submerged; i Let be the width of the i-th road segment; Specifically, the density index calculation module calculates the road area and length using the polygon area calculation formula and the Euclidean distance formula, respectively, and determines α and β by fitting historical data using the nonlinear least squares method; and determines the weight parameters α=1.244 and β=0.802 based on 60 sets of historical data using the Levenberg-Marquardt optimization algorithm.

[0075] The tiered assessment module, such as Figure 3 As shown, the K-means clustering algorithm is used to cluster the comprehensive density index into 3 classes, and the SVM algorithm is used to establish the classification boundary, dividing the degree of road flooding in the region into three levels: "high", "medium" and "low".

[0076] Specifically, in the tiered assessment module: The objective function of the K-means clustering algorithm is: Where SSE is the sum of squared errors; K is the number of clusters (K=3 in this invention); C i For the i-th cluster; μ i Let x be the cluster center of the i-th cluster; x is the value belonging to cluster C. i Data points; |x - μ i |² represents the relationship between data point x and cluster center μ. i The squared Euclidean distance between them; the cluster center update formula is: , where μ i (t+1) C is the new cluster center of the i-th cluster after the (t+1)-th iteration; i (t) Let C be the set of all data points contained in the i-th cluster at the t-th iteration; i (t) | represents the number of data points in the i-th cluster at the t-th iteration; x represents the cluster C. i (t) The data points in the cluster; Σ represents the summation of all data points within the cluster; where K=3, corresponding to the three flooding levels of "high", "medium" and "low" respectively; More specifically, the specific steps of the K-means clustering algorithm are as follows: (1) Randomly select K data points as initial cluster centers: , … ,in represents the initial center of the i-th cluster, and (0) represents the 0th iteration.

[0077] (2) Assign each data point x to the nearest cluster center according to the formula:

[0078] (3) Calculate the new center of each cluster (the mean of the data points within the cluster), using the formula shown below:

[0079] (4) Repeat (2) and (3) until the cluster centers no longer change or the maximum number of iterations is reached.

[0080] The classification boundaries established by the SVM algorithm are x=0.320 and x=0.685, and the classification rule is: D combined Areas with a value ≥0.685 are considered "high" levels of flooding, and areas with a value ≤0.32 are considered "low" levels of flooding. combined Areas with a strength <0.685 are considered "moderately" flooded areas. <D combined Areas with a value less than 0.32 are considered "low" flooding areas.

[0081] More specifically, after clustering, labeled data points are obtained, and the road flooding state is classified based on the SVM algorithm. The specific steps are as follows: 1. Divide the data with obtained flooding state labels into training and test sets in a 7:3 ratio.

[0082] 2. Input the training set into the SVM model, optimize the model parameters, and divide the training set by a straight line.

[0083] 3. The optimal model parameters are learned through training, accurately segmenting various flooding states, and the training effect of the model is evaluated using a test set.

[0084] 4. Output the final classification model.

[0085] The classified samples were divided into training and test sets in a 7:3 ratio. An SVM model was trained using the training set to obtain the SVM classification model. The classification results are as follows: Figure 4 As shown.

[0086] In this embodiment, the K-means clustering algorithm is set to have K=3 clusters, and the cluster centers for each flooding state are 0.164, 0.462, and 0.821, respectively. The classification accuracy of the SVM classifier reaches 96.7%.

[0087] The disaster early warning module is used for automated data loading, image analysis and extraction, data fusion, spatial data extraction, and inundation level assessment. It enables rapid extraction and evaluation of information on flood-inundated roads, comprehensively reflecting the impact of road geometry on flood disasters, and providing scientific and accurate support for emergency road rescue decisions after floods. It features automated data loading, image analysis and extraction, data fusion, spatial data extraction, and inundation level assessment functions, enabling rapid extraction and evaluation of information on flood-inundated roads and providing scientific support for emergency rescue decisions for road traffic infrastructure.

[0088] Specifically, based on the ArcGIS Engine+PyTorch platform, C# & Python languages, and the .NET framework, the model provides a user interface and results visualization functions. Verification results show that the model has R... 2 With excellent performance of 0.986 and MSE of 2.965, it can comprehensively reflect the impact of road geometry on flood inundation, providing scientific and accurate support for emergency road rescue decisions after floods. More specifically, a simplified interface with automated data loading, image analysis and extraction, data fusion, spatial data extraction, and inundation degree assessment functions has been designed. This enables rapid extraction and assessment of flood-inundated road information, providing an operable user-interactive platform for pre-disaster road extraction, flood extraction during disasters, and inundation road information extraction and assessment. It provides scientific support for emergency rescue decisions for road traffic infrastructure. The main interface of the intelligent road flood inundation disaster assessment system is shown below. Figure 5 As shown, it provides a user interface and results visualization functions.

[0089] The following are specific embodiments of the present invention. It should be noted that the present invention is not limited to the following specific embodiments. All equivalent modifications made based on the technical solutions of this application fall within the protection scope of the present invention.

[0090] Example 1: This embodiment provides a method and approach for assessing flood-inundated road disasters based on a comprehensive density index, including: 1. Construct a data acquisition module to acquire and preprocess remote sensing image data. The specific steps are as follows: 1.1 Data Collection and Preprocessing Pre-disaster road remote sensing images and flood remote sensing images during the disaster were collected for the study area, with a spatial resolution of 1m. Radiometric and geometric corrections were performed on the acquired remote sensing images to ensure data quality met the requirements for subsequent processing.

[0091] Radiation correction uses the following formula: Where L is the radiance value, DN is the digital quantization value, Gain is the gain coefficient, and Offset is the offset. Geometric correction employs a polynomial transform method, with the root mean square error of the control points controlled within 0.5 pixels to ensure registration accuracy.

[0092] 1.2 Data format standardization The preprocessed remote sensing images were uniformly converted to TIFF format, with a pixel depth set to 8 bits to ensure data format consistency. A unified coordinate reference system was established, adopting the WGS 1984 Web Mercator projection coordinate system.

[0093] 2. Construct a feature extraction module to extract features from roads and flooded areas respectively. The specific steps are as follows: 2.1 Construction of Road Feature Extraction Network An improved U-Net network is used as the backbone network for road extraction. The network structure consists of three parts: an encoder, skip connections, and a decoder.

[0094] The encoder uses ResNet-50 as the backbone network, and the number of network layers is set as shown in the table below: Table 1 Network Layer Settings

[0095] An attention gating mechanism is added at the jump connection, and the attention weight calculation formula is as follows: Where g is the gating signal, x is the feature map, Wg and Wx are the weight matrices, bg is the bias term, and σ is the Sigmoid activation function.

[0096] The decoder employs a densely connected structure, with each decoding block including upsampling, convolution, batch normalization, and ReLU activation functions.

[0097] The loss function uses a combination of Dice loss and cross-entropy loss: Where λ1=0.6 and λ2=0.4 are weighting coefficients.

[0098] 2.2 Construction of Flood Feature Extraction Network An improved DeepLabv3+ network was used to extract features of flooded areas. The backbone network adopted an improved ResNet, and the ASPP module used multi-scale dilated convolution with different dilation rates set to {6, 12, 18}.

[0099] The decoder employs a feature fusion mechanism, specifically as follows: The low-level features are reduced to 48 channels using 1×1 convolution; the ASPP output features are upsampled by 4 times. The two features are concatenated and then processed through a 3×3 convolution to obtain the final output. Network training parameters are set as follows: learning rate 0.001, batch size 16, training epochs 100, and the Adam optimizer is used.

[0100] 2.3 Raster to Vector Data Processing The extracted binary mask image is then vectorized, specifically including: (1) Region segmentation: Define a region R that satisfies the homogeneity condition. Where ε is the allowed attribute difference threshold, which is set to 0.1.

[0101] (2) Boundary extraction: Calculate the gradient operator ∇f(x,y) = (∂f / ∂x, ∂f / ∂y). When the gradient magnitude exceeds the threshold, the pixel is determined to be located at the edge of the region.

[0102] (3) Topological reconstruction: Connect discrete boundary points into a continuous geometric shape to ensure that the boundary is closed and does not self-intersect.

[0103] 3. Construct a data fusion module to realize the overlay analysis of road flood data. The specific steps are as follows: 3.1 Spatial Registration Unify pre-disaster road data and flood data during the disaster into the WGS 1984 Web Mercator coordinate system to ensure accurate spatial correspondence.

[0104] 3.2 Raster Overlay Operation Use logical AND operations to overlay grids: Where A is the road mask matrix, B is the flood mask matrix, and C is the flooded road mask matrix.

[0105] Specific implementation code logic:

[0106] 4. Construct a density index calculation module to calculate the comprehensive density index. The specific steps are as follows: 4.1 Extraction of basic indicators Road area extraction based on polygon area calculation formula:

[0107] Calculating road length based on Euclidean distance:

[0108] 4.2 Calculation of Comprehensive Density Index Establish a comprehensive density index model:

[0109] 4.3 Parameter Optimization Process The parameters are fitted using the nonlinear least squares method. The objective function is: The Levenberg-Marquardt algorithm is used to solve this problem, with the parameter update equation as follows: Where S(θ) is the objective function; N=60 is the total number of historical data samples; θ=(α, β, k, b) is the parameter vector to be optimized; z_i is the actual rescue time of the i-th sample; f(x_i;y_i;θ) is the model prediction value; x_i and y_i are the flooding length and road width of the i-th sample, respectively. J is the Jacobian matrix (N×4 dimensions), with each element J_ij = ∂f_i / ∂θ_j; J^T is the transpose of the Jacobian matrix; λ is the adaptive damping factor; I is the 4×4 identity matrix; Δθ is the parameter increment vector; y is the N-dimensional actual observation vector; and f(θ) is the N-dimensional model prediction vector.

[0110] The optimized parameters were obtained through training with 60 sets of historical data: k = 1.24 × 10⁻ 4 , α=1.244, β=0.802, b=0.666.

[0111] The model has a good fit, with a coefficient of determination R² = 0.986 and a mean squared error MSE = 2.965.

[0112] The final formula for the comprehensive density index is determined as follows:

[0113] 5. Construct a hierarchical assessment module to classify flooding states. The specific steps are as follows: 5.1 Implementation of K-means Clustering Algorithm Objective function: Where SSE is the total sum of squared errors; K=3 is the number of clusters, corresponding to the three flooding levels of "high", "medium" and "low"; i is the cluster index; Ci is the set of data points in the i-th cluster; x is the single comprehensive density index value in cluster Ci; μi is the center value of the i-th cluster; |x - μi|² is the squared distance from the data point to the cluster center.

[0114] Algorithm steps: Step 1: Randomly select K=3 initial cluster centers μ1 （0） μ2 （0） μ3 （0） Step 2: Calculate the distance from each data point to the cluster center and assign it to the nearest cluster. Step 3: Update cluster centers: ; where $\mu_i^{(t + 1)}$ is the updated $i$-th cluster center; $t$ is the current iteration number; $C_i^{(t)}$ is the set of data points in the $i$-th cluster at the $t$-th iteration; $|C_i^{(t)}|$ is the number of data points in this cluster; $x$ is a single data point within the cluster; $\Sigma$ represents the arithmetic mean over all data points within the cluster.

[0115] Step4: Repeat Step2 and Step3 until convergence 5.2. SVM Classification Model Training Divide the clustering results into a training set and a test set in a 7:3 ratio, and train the SVM classifier.

[0116] The kernel function uses the RBF kernel: Parameter settings: $C = 1.0$, $\gamma = 0.1$ Optimize the parameters through grid search, and finally obtain the classification boundaries: $x = 0.320$ and $x = 0.685$ Classification rules: $D_{combined} \geq 0.685$: High inundation degree $0.32 \leq D_{combined} \lt 0.685$: Medium inundation degree $0 \lt D_{combined} \lt 0.32$: Low inundation degree $D_{combined} = 0$: Non-inundated area Model evaluation results: Overall precision is 96.7%, and the precision and recall rates for each category exceed 94%.

[0117] 6. Build a disaster warning module and develop an intelligent evaluation system. The specific steps are as follows: 6.1. System Architecture Design Adopt a hybrid architecture of ArcGIS Engine + PyTorch: Front-end interface: Developed based on C# and the.NET framework; Deep learning module: Implemented based on PyTorch; GIS analysis module: Implemented based on ArcGIS Engine; Data processing: Implemented based on Python.

[0118] 6.2. Module Function Implementation (1) Pre-disaster Road Extraction Module Data loading: Supports TIFF, JPEG, PNG formats; Model selection: Provides options such as U-Net, improved U-Net, etc.; Parameter settings: Learning rate, optimizer, number of training epochs; Result display: Real-time display of the extraction progress and results.

[0119] (2) In-disaster Flood Extraction Module Network architecture: Improved DeepLabv3+; Training parameters: batch size 16, learning rate 0.001; Data augmentation: random flipping, rotation, scaling.

[0120] (3) Flooding Information Assessment Module Data fusion: raster logic operations; Information extraction: Length, area, and density calculations; State classification: K-means + SVM classification; Results visualization: hierarchical coloring display.

[0121] 6.3 System Verification Experiment Four areas with different levels of flood inundation were selected for verification: The classification results are as follows: 1 high-level area, 2 medium-level areas, and 4 low-level areas.

[0122] System processing time: The average processing time for a single region is 5-8 minutes, including the entire process of data loading, model inference, and result analysis.

[0123] The above embodiments demonstrate that this invention can accurately extract information about flooded roads, scientifically assess the degree of flooding, and provide effective support for emergency rescue decision-making. The system has a user-friendly interface, is easy to operate, and possesses good practicality and promotional value.

[0124] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0125] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

[0126] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed by the present invention.

Claims

1. A method for assessing the disaster situation of flooded roads based on a comprehensive density index, characterized in that, Includes the following steps: Step 1, Data Acquisition: Collect pre-disaster road datasets and flood datasets for the study area, and perform radiometric and geometric correction preprocessing; the pre-disaster road datasets include pre-disaster road remote sensing images and pre-disaster road masking data; the flood datasets include flood remote sensing images and flood masking data. Step 2, Feature Extraction: An improved U-Net network is used to extract pre-disaster road features from the pre-processed pre-disaster road remote sensing images to obtain the spatial distribution and geometric features of the pre-disaster roads; An improved DeepLabv3+ network was used to extract flood features from preprocessed flood-related remote sensing images, yielding spatial distribution and geometric features of roads during the disaster; spatial... The distribution and geometric features include information on the length and width of roads before and during the disaster. Step 3, Data Fusion: Based on the spatial distribution and geometric features obtained in Step 2, the pre-disaster road masking data and the flood masking data during the disaster are overlaid using a raster overlay analysis, and logical AND operations are performed. The raster data is then converted into vector data to obtain the flooded road mask matrix, where A is the road mask matrix, B is the flood mask matrix, and C is the flooded road mask matrix. i and j represent the row index and column index in the matrix, respectively; Step 4, Density Index Calculation: Using the flooded road mask matrix obtained in Step 3, establish the comprehensive density index: Where α is the length weight parameter, β is the width weight parameter, and α and β are determined by fitting historical data using a nonlinear least squares method; n is the total number of submerged road sections; l i Let l be the length of the i-th submerged road segment; (n+i) w is the length of the i-th segment of the road that was not submerged; i Let be the width of the i-th road segment; Step 5, graded assessment: The K-means clustering algorithm is used to cluster the comprehensive density index into 3 classes. Combined with the SVM algorithm, the classification boundary is established to divide the degree of road flooding in the area into three levels: high, medium and low. Step 6, Disaster Early Warning: Automated data loading, image analysis and extraction, data fusion, spatial data extraction, and flooding degree assessment are carried out to achieve rapid extraction and assessment of flood-inundated road information, comprehensively reflecting the impact of road geometric features on flood disasters, and providing scientific and accurate support for road emergency rescue decision-making after flood disasters.

2. The flood-inundated road disaster assessment method based on comprehensive density index as described in claim 1, characterized in that, The improved U-Net network uses ResNet-50 as the encoder backbone, adds an attention gating mechanism at skip connections, employs a dense connection structure in the decoder, and uses a combination of Dice loss and cross-entropy loss as the loss function. , where λ1 and λ2 are weighting coefficients.

3. The flood-inundated road disaster assessment method based on comprehensive density index as described in claim 1, characterized in that, The backbone of the improved DeepLabv3+ network adopts an improved ResNet, in which the ASPP module uses multi-scale dilated convolution and the decoder uses a feature fusion mechanism.

4. The flood-inundated road disaster assessment method based on comprehensive density index as described in claim 1, characterized in that, The raster overlay analysis involves performing a logical AND operation on each pixel location. When the pixel values ​​of both the road mask and the flood mask are 1, the corresponding position of the flooded road mask is 1; otherwise, it is 0. The process of converting the raster data into vector data includes: region segmentation, boundary extraction, and topology reconstruction.

5. The flood-inundated road disaster assessment method based on comprehensive density index as described in claim 1, characterized in that, In step 4, the road area and length are calculated using the polygon area calculation formula and the Euclidean distance formula, respectively, and α and β are determined by fitting historical data using the nonlinear least squares method; wherein, the nonlinear least squares parameter optimization adopts the Levenberg-Marquardt algorithm.

6. The flood-inundated road disaster assessment method based on comprehensive density index as described in claim 1, characterized in that, The objective function of the K-means clustering algorithm is: Where SSE is the sum of squared errors; K is the number of clusters; C i For the i-th cluster; μ i Let x be the cluster center of the i-th cluster; x is the value belonging to cluster C. i Data points; |x - μ i |² represents the relationship between data point x and cluster center μ. i The squared Euclidean distance between them; the cluster center update formula is: , where μ i (t+1) The new cluster center of the i-th cluster after the (t+1)-th iteration; C i (t) Let C be the set of all data points contained in the i-th cluster at the t-th iteration; i (t) | represents the number of data points in the i-th cluster at the t-th iteration; x represents the cluster C. i (t) The data points are in the middle; where K=3, corresponding to the high, medium and low flood levels respectively.

7. The flood-inundated road disaster assessment method based on comprehensive density index as described in claim 1, characterized in that, The SVM algorithm establishes classification boundaries of x=0.320 and x=0.685, and the classification rule is: D combined Areas with a value ≥0.685 are considered high-level flooding areas, and areas with a value ≤0.32 are considered high-level flooding areas. combined Areas with a strength <0.685 are considered moderately flooded areas. <D combined Areas with a value less than 0.32 are considered low-level flooding areas.

8. A flood-inundated road disaster assessment system based on a comprehensive density index, characterized in that, The system can implement the method described in any one of claims 1 to 7, including: The data acquisition module is used to collect pre-disaster road datasets and flood datasets for the study area. The image spatial resolution is 1m, and radiometric and geometric correction preprocessing is performed. The pre-disaster road dataset includes pre-disaster road remote sensing images and pre-disaster road masking data; the flood dataset includes flood remote sensing images and flood masking data. The feature extraction module is used to extract pre-disaster road features from pre-processed pre-disaster road remote sensing images using an improved U-Net network, obtaining the spatial distribution and geometric features of pre-disaster roads; and to extract disaster-time flood features from pre-processed disaster-time flood remote sensing images using an improved DeepLabv3+ network, obtaining the spatial distribution and geometric features of disaster-time roads; the spatial distribution and geometric features include the length and width information of pre-disaster and disaster-time roads. The data fusion module is used to perform raster overlay analysis on pre-disaster road masking data and post-disaster flood masking data based on the obtained spatial distribution and geometric features, through logical AND operations. The raster data is then converted into vector data to obtain the flooded road mask matrix, where A is the road mask matrix, B is the flood mask matrix, and C is the flooded road mask matrix; i and j represent the row index and column index in the matrix, respectively. The density index calculation module establishes a comprehensive density index using the obtained flooded road mask matrix. Where α is the length weight parameter, β is the width weight parameter, and α and β are determined by fitting historical data using a nonlinear least squares method; n is the total number of submerged road sections; l i Let l be the length of the i-th submerged road segment; (n+i) w is the length of the i-th segment of the road that was not submerged; i Let be the width of the i-th road segment; The hierarchical assessment module is used to cluster the comprehensive density index into 3 classes using the K-means clustering algorithm, and to establish the classification boundary by combining the SVM algorithm to divide the degree of road flooding in the region into three levels: high, medium and low. The disaster early warning module is used for automated data loading, image analysis and extraction, data fusion, spatial data extraction, and flooding degree assessment. It enables rapid extraction and assessment of flood-inundated road information, comprehensively reflects the impact of road geometric features on flood disasters, and provides scientific and accurate support for road emergency rescue decision-making after flood disasters.

9. The flood-inundated road disaster assessment system based on comprehensive density index as described in claim 8, characterized in that, The improved U-Net network in the feature extraction module includes an encoder, skip connections, and a decoder. The encoder uses ResNet-50 as the backbone network, an attention gating mechanism is set at the skip connections, and the decoder adopts a dense connection structure. The backbone of the improved DeepLabv3+ network adopts an improved ResNet, and the decoder uses a feature fusion mechanism. The data fusion module uses raster overlay analysis, performing a logical AND operation on each pixel location. When the pixel values ​​of both the road mask and the flood mask are 1, the corresponding position value of the flooded road mask is 1; otherwise, it is 0. The process of converting raster data into vector data includes: region segmentation, boundary extraction, and topology reconstruction. The density index calculation module calculates the road area and length using the polygon area calculation formula and the Euclidean distance formula, respectively, and determines α and β by fitting historical data using the nonlinear least squares method.

10. The flood-inundated road disaster assessment system based on comprehensive density index as described in claim 8, characterized in that, In the graded assessment module: The objective function of the K-means clustering algorithm is: Where SSE is the sum of squared errors; K is the number of clusters; C i For the i-th cluster; μ i Let x be the cluster center of the i-th cluster; x is the value belonging to cluster C. i Data points; |x - μ i |² represents the relationship between data point x and cluster center μ. i The squared Euclidean distance between them; the cluster center update formula is: , where μ i (t+1) The new cluster center of the i-th cluster after the (t+1)-th iteration; C i (t) Let C be the set of all data points contained in the i-th cluster at the t-th iteration; i (t) | represents the number of data points in the i-th cluster at the t-th iteration; x represents the cluster C. i (t) The data points in the data; where K=3, corresponding to the high, medium and low inundation levels respectively; The SVM algorithm establishes classification boundaries of x=0.320 and x=0.685, and the classification rule is: D combined Areas with a value ≥0.685 are considered high-level flooding areas, and areas with a value ≤0.32 are considered high-level flooding areas. combined Areas with a strength <0.685 are considered moderately flooded areas. <D combined Areas with a value less than 0.32 are considered low-level flooding areas.

Images