Rapid evaluation method of seismic performance of masonry structures based on image recognition

Abstract

The present invention proposes a method for rapidly assessing the seismic performance of masonry structures based on image recognition, comprising the following steps: identifying the outlines, openings, and markers in an image; using OpenCV to perform perspective correction on the image using the markers and calculate the actual size of the target; using the MobileViTV3 model to classify the image by the number of storeys, thereby automatically identifying building dimension parameters and rapidly collecting house information; parametrically establishing a house structure model, calculating basic parameters such as storey height, area, and representative values ​​of structural gravity loads; calculating the equivalent cross-sectional area, axial compression ratio, and equivalent moment of inertia of each floor wall using a simplified method; rapidly calculating the lateral stiffness of the masonry structure wall using a simplified shear layer model; calculating the seismic action and effects of the structure using the bottom shear method; and rating the seismic performance of the structure using the inter-story displacement angle as an indicator. The present invention, based on image recognition methods, can rapidly collect dimension information of old houses and rapidly assess the performance of masonry structures.

Description

Technical Field

[0001] The present invention relates to the field of house performance evaluation, and in particular to a method for quickly evaluating the seismic performance of masonry structures based on image recognition. Background Art

[0002] Older buildings in my country are primarily brick structures, with the majority being single-story. Due to economic constraints and limited construction standards, most houses are not designed according to standards. Instead, they are constructed by local construction teams based on experience, using untested building materials. This results in generally poor safety and earthquake resistance, directly endangering residents' lives and property. The government has introduced policies to conduct a nationwide housing risk survey.

[0003] However, the current screening method has the following shortcomings:

[0004] 1. Risk inspections of old houses are not completed by professional house inspection agencies. Due to cost issues, the safety inspections of many old houses are carried out by relevant personnel from government departments. They need to manually check the corresponding structural parts according to requirements and manually input the data into the old house inspection software based on personal judgment, which leads to low data collection efficiency.

[0005] 2. Most of the existing old house inspection software collects qualitative data such as whether the house is professionally designed, whether it has been seismically reinforced, and whether there are obvious tilting cracks. It also relies on non-professionals to judge whether there are safety hazards. It has a low level of intelligence and the accuracy of the assessment results is difficult to guarantee.

[0006] 3. Existing structural seismic assessment calculation software requires manual establishment of structural models based on drawings and adjustment of calculation parameters according to actual site conditions. The seismic performance assessment of a single building requires a large amount of modeling and calculation time, making it difficult to promote and apply it to the inspection and assessment of a large number of old houses across the country. Summary of the Invention

[0007] The technical problem to be solved by the present invention is to address the deficiencies of the above-mentioned existing technologies and provide a method for rapid evaluation of the seismic performance of masonry structures based on image recognition. By collecting house facade images with markers, the SegFormer and MobileViTV3 models are used to automatically identify and extract house information in the images, a parameterized house structure model is established, and a simplified shear layer model is used to quickly calculate the lateral stiffness of the masonry structure wall. The seismic action and effect of the structure are calculated according to the bottom shear method, and the seismic performance of the structure is rated using the inter-story displacement angle as an indicator.

[0008] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0009] A rapid assessment method for seismic performance of masonry structures based on image recognition comprises the following steps:

[0010] Step 1: Paste markers on all exterior walls of the masonry structure to serve as reference for image distortion correction and actual size conversion;

[0011] Step 2, image acquisition: take photos of all facades of the masonry structure;

[0012] Step 3: Perform target recognition on the image collected in step 2 to output a binary image: Use the SegFormer model to identify the outlines of the house facade, door and window openings, and signs in the image;

[0013] Step 4: Based on the binary image output in step 3, perform parameter measurement and obtain the total number of floors of the building. The parameters include the actual size and location of the building facade, door and window openings, and landmarks. Specifically, they include:

[0014] Use OpenCV to correct the deformation of the marker, count the required pixel size of the house facade and door and window openings (pixel), and the number of pixels occupied by the side length of the marker (mark_pixel), and calculate the actual size using the following formula:

[0015] L = pixel / mark_pixel × 30cm;

[0016] The actual dimensions include the width, height, and center coordinates of the building outline, markers, and door and window openings;

[0017] Use the MobileViTV3 model to classify images by the number of floors of the house and output the total number of floors of the house;

[0018] Step 5. Calculate the floor height and opening ratio: Based on the total house height house_height and the total number of floors level obtained in step 4, calculate the floor height H of each floor as follows:

[0019] H = house_height / level;

[0020] Calculate the wall opening ratio ω i To be used to calculate the lateral stiffness K in the subsequent step 6; ω i = total area of ​​openings on each floor / wall area of ​​that floor;

[0021] Step 6: Based on the floor height H obtained in step 5, perform the structural equivalent total gravity load G eq Calculate the lateral stiffness K for use in the subsequent seismic verification and rating in step 7:

[0022] Among them, the equivalent total gravity load of the structure G eq :

[0023]

[0024] G eq,i =A·(q d +q l );

[0025]

[0026] q l =2.0kN / m 2 ;

[0027] A=L x ·L y ;

[0028] Among them, L x - The width of the house in the horizontal direction;

[0029] L y - the longitudinal width of the house;

[0030] B-wall thickness; q m - bulk density of wall materials; q f - Bulk density of floor slab material; h f - Floor slab thickness;

[0031] Among them, the calculation of lateral stiffness K: determine the masonry elastic modulus E and shear modulus G based on local statistical data, and calculate the equivalent cross-sectional area A of the i-th layer wall w,i , axial pressure ratio v i , the equivalent section inertia moment I of the wall i , combined with the floor height H and opening rate ω obtained in step 5 i Calculate the elastic stiffness K of each layer of wall according to the following formula: i , the elastic stiffness K of the two parallel walls i The cumulative value is the lateral stiffness K of the wall in that direction;

[0032] in,

[0033]

[0034] The specific steps include:

[0035] Step 6B1. Calculate the equivalent cross-sectional area A of each wall layer w,i :

[0036]

[0037] n=edge_y_num-1;

[0038] Among them, in the i-th layer, the width l of the j-th columnij The number of i,size , the height h of the kth row ik The number of h i,size Construct a two-dimensional list isHole, with the default element being 1. When the area at the jth column and the kth row of the wall is a hole, set the corresponding element of isHole to 0.

[0039] The edge_y_i list is composed of the boundary coordinates edge_y_i of the door and window openings, and edge_y_num is the number of elements in the edge_y_i list;

[0040] The edge_x_i list is composed of the boundary coordinates edge_x_i of the door and window openings;

[0041] Subtract two adjacent elements from the edge_x_i list to get the width of each column l ij Data list;

[0042] Subtract two adjacent elements from the edge_y_i list to get the height h of each row ik Data list;

[0043] Step 6B2: Calculate the axial compression ratio v of each layer i ;

[0044]

[0045] Among them, f m - masonry compressive strength;

[0046] Step 6B3: Calculate the equivalent section inertia moment I of each wall layer i :

[0047]

[0048]

[0049] D ij -The horizontal distance from the centroid of the j-th column of area divided by the wall to the central axis of the wall;

[0050] Step 7: Take the interlayer displacement angle Δθ e,i Verify and rate the seismic performance of the structure based on the following indicators:

[0051] F Ek =α1G eq ;

[0052]

[0053] Finally, according to the inter-story displacement angle Δθ e,i Perform structural rating.

[0054] Preferably, the marker is a KT board with a size of 30cm×30cm; one KT board is pasted on one wall; when the facades are not in the same plane, the KT board is pasted on the facade closer to the shooting position and temporarily fixed with tape.

[0055] Preferably, in step 2, the four boundary lines of the photographed house facade are clearly visible, the main house is complete and occupies most of the picture; frontal shooting is adopted, and when frontal shooting is not possible, the angle of oblique shooting is not too large, and there are no people or other obstructions in the picture.

[0056] Preferably, in step 3, the training method of the SegFormer model is as follows:

[0057] (1) Dataset preparation: Take photos of the house facades with markers, use Photoshop to accurately annotate the houses, door and window openings, and markers, and divide the obtained dataset into training set, validation set, and test set at a ratio of 9:1:1;

[0058] (2) Model training: Use the weights pre-trained on Image-1K to initialize the SegFormer model and train it. Take the optimal weight on the validation set, that is, the weight with the highest average Dice coefficient on the validation set. Test it on the test set to verify the model training effect.

[0059] Preferably, in step 4, the layer number identification method is as follows:

[0060] Step 4B1. Dataset preparation: Divide the binary image processed in step 3 into four categories according to the number of layers: 1 layer, 2 layers, 3 layers, and more than 3 layers. Construct a layer classification dataset. Divide the dataset of each category into training set, validation set, and test set at a ratio of 9:1:1.

[0061] Step 4B2, model training: Initialize the MobileViTV3 model using the weights pre-trained on Image-1K and train it. Take the weight that achieves the highest accuracy on the validation set and test it on the test set to verify the model training effect;

[0062] Step 4B3, layer number identification: the binary image processed in step 3 is passed through the MobileViTV3 model for layer number identification.

[0063] Preferably, in step 6, the equivalent total gravity load G eq The calculation method is as follows:

[0064] Step 6A1. Obtain the house_width of different facades from step 4, take the average of the widths of the two parallel walls, and determine the horizontal width L of the house. x and the vertical width L y , calculate the area A of each layer according to the following formula:

[0065] A=L x ·L y ;

[0066] Step 6A2: Determine the wall thickness B and the wall material bulk density q based on local statistical data. m , floor slab material bulk density q f , floor thickness h f , combined with the standard value and the storey height H obtained in step 5 and the horizontal width L of the house obtained in step 6A1 x , vertical width L y , the area A of each floor, calculate the dead load q of each floor d and live load q l ;

[0067] Step 6A3: Combine the area A of each layer obtained in step 6A1 and the dead load q obtained in step 6A2. d and live load q l , calculate the representative value of gravity load G of each layer eq,i .

[0068] Preferably, in step 6B1, the equivalent cross-sectional area A of each wall layer is w,i The calculation steps are as follows:

[0069] Step 6B11. Count the number of openings on each floor: Calculate the upper and lower boundaries of the house outline using the y-axis center coordinate house_center_y and the height house_height obtained in Step 4. Combined with the floor height H, determine the upper and lower boundaries of the i-th floor using the following formula:

[0070] upLine=house_center_y+house_height / 2-i×H;

[0071] downLine=house_center_y+house_height / 2-(i-1)×H;

[0072] When the value of the hole's vertical coordinate holes_center_y is between the upper and lower dividing lines, the hole is judged to be located in the i-th layer;

[0073] Traverse the vertical coordinates of each hole, determine the layer number of each hole, and count the number of holes on each layer;

[0074] Step 6B12, obtaining the number of cross sections n;

[0075] Based on the number of door and window openings on the i-th layer obtained in step 6B11, divide the wall of this layer according to the door and window opening boundaries and their extension lines, calculate the coordinates of the opening boundaries according to the following formula, and store them in a list;

[0076] edge_x_i=holes_center_x±holes_width / 2;

[0077] edge_y_i=holes_center_y±holes_height / 2;

[0078] Then sort the data in the edge_y_i list. When the difference between the boundary coordinates in the list is less than the threshold Δ = 0.05 × H, merge the two into the same boundary coordinate by taking the average;

[0079] Sort the data in the edge_x_i list. When the difference between the boundary coordinates in the list is less than the threshold Δ = 0.05 × H, merge the two into the same boundary coordinate by taking the average;

[0080] Finally, count the number of elements in the edge_y_i list edge_y_num;

[0081] Step 6B13: Measure the width l of the jth column after the area is divided ij and the height h of row k ik ; Subtract the adjacent elements of the edge_x_i list obtained in step 6B12 from each other to obtain the width of each column l ij Data list; subtract adjacent elements of the edge_y_i list obtained in step 6B12 to obtain the height h of each row ik Data list;

[0082] According to the number of cross sections n, the door and window opening boundaries and their extension lines, the opening conditions of each layer are calculated to construct a two-dimensional list isHole. isHole stores the opening information of the wall area. The value of isHole is 1 for the solid area and 0 for the opening area.

[0083] Step 6B14: Determine the vertical direction of l based on local statistical data. ij The wall thickness B in the direction is calculated according to formula A. ik = Calculate the cross-sectional area A of the wall in the kth row ik ;

[0084] Step 6B15: Calculate the equivalent cross-sectional area of ​​the wall using the following formula:

[0085]

[0086] Compare A in each row ik , take the smallest one as A w,min,i , the minimum cross-sectional area A of the four facade walls w,min,i Summing to get ∑A w,min,i, which is used for the axial compression ratio v of each layer in the subsequent step 6B2 i .

[0087] Preferably, in step 6B2, the axial compression ratio v of each layer is i The calculation steps are as follows:

[0088] Step 6B21: The constant load q of each layer obtained in step 6A2 d and live load q l Calculate the design value of the axial force of the i-th layer N according to the following formula i ;

[0089]

[0090] Step 6B22: Determine the masonry compressive strength f from local statistical data m , combined with the sum of the minimum cross-sectional areas ∑A obtained in step 6B15 w,min,i and the design value N of the axial force obtained in step 6B21 i , calculate the axial compression ratio of each layer according to the following formula:

[0091]

[0092] Preferably, in step 6B3, the equivalent section moment of inertia I of each layer of wall is i The calculation steps are as follows:

[0093] (1) According to the h obtained in step 6B13 ik , calculate the wall parameters of the k-th row area by the following formula

[0094]

[0095] Then the vertical direction of l is determined by local statistics. ij The wall thickness B in the direction, combined with the l obtained in step 6B13 ij and isHole ijk The coefficient is calculated by the following formula: the moment of inertia I of the wall in the jth column and the kth row to the center axis of the wall ij :

[0096]

[0097] And calculate the distance D from the centroid of the j-th column area divided by the wall to the central axis of the wall ij ;

[0098] Step 6B32: Calculate the equivalent section moment of inertia of each layer using the following formula:

[0099]

[0100] Preferably, in step 7, when the interlayer displacement angle Δθe,i <1 / 1000, the structure rating is A; when 1 / 1000≤Δθ e,i <1 / 670, the structural rating is B; when 1 / 670≤Δθ e,i <1 / 330, the structural rating is C; when Δθ e,i When the ratio is greater than 1 / 330, the structural rating is D.

[0101] The present invention has the following beneficial effects:

[0102] Compared with the prior art, the present invention has the following characteristics and advantages:

[0103] 1. Applicable to the condition safety assessment of old masonry structure houses;

[0104] 2. Rapidly collect dimensional information of old houses based on digital imaging methods;

[0105] 3. Rapid performance evaluation of old masonry houses based on image recognition methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0106] Figure 1 Flow chart of the rapid evaluation method of the present invention.

[0107] Figure 2 is the marker used in the present invention.

[0108] Figure 3(a) to Figure 3(b) Schematic diagram of the output of house outline, door and window openings and marker recognition in the present invention: FIG3(a) is a front elevation image of the house; FIG3(b) is a side elevation image of the house.

[0109] Figure 4(a) to Figure 4(c) Schematic diagram of the measurement parameters of a single wall photo after deformation correction in the present invention (unit: mm): Figure 4(a) shows the width, height, and center point coordinates of the house outline; Figure 4(b) shows the width, height, and center point coordinates of the marker; Figure 4(c) shows the width, height, and center point coordinates of each door and window opening.

[0110] FIG5( a ) is a schematic diagram of the classification effect of a first-floor house in the present invention.

[0111] FIG5( b ) is a schematic diagram of the classification effect of two-story houses in the present invention.

[0112] FIG5( c ) is a schematic diagram of the classification effect of a three-story house in the present invention.

[0113] Figure 6 This is a schematic diagram of dividing the wall into areas based on the boundaries of door and window openings in the present invention.

[0114] Figure 7 Schematic diagram for calculating the length and height of each area of ​​the wall in the present invention.

[0115] Figure 8 Schematic diagram of the isHole parameter value for calculating the hole conditions of each layer in the present invention. DETAILED DESCRIPTION

[0116] The present invention will be further described in detail below with reference to the accompanying drawings and specific preferred embodiments.

[0117] In the description of the present invention, it should be understood that the terms "left side," "right side," "upper," "lower," etc., indicating orientations or positional relationships, are based on the orientations or positional relationships shown in the accompanying drawings and are intended solely to facilitate description and simplify the present invention. They do not indicate or imply that the devices or components referred to must have, be constructed, or operate in a specific orientation. Therefore, they should not be construed as limitations on the present invention. The specific dimensions used in this embodiment are merely for illustrative purposes and do not limit the scope of protection of the present invention.

[0118] like Figure 1 As shown, the rapid evaluation method of seismic performance of masonry structures based on image recognition includes the following steps.

[0119] Step 1: Post markers as a reference for image distortion correction and actual size conversion: When taking photos of the facade of a masonry structure, due to differences in shooting angles and distances, it is impossible to directly obtain the actual size of the house from the image. Therefore, it is necessary to post markers on all the facade walls of the masonry structure, use the four-sided outline of the markers to correct the image distortion, and use the size of the markers to convert the actual size of the entire house and the door and window openings. Figure 2 As shown, the upper and lower edges of the sign are horizontal, and the left and right edges are vertical. The sign is made of KT board, which is lightweight and easy to carry, easy to paste and remove, and has a size of 30cm×30cm. One KT board is pasted on one wall. When the facades are not in the same plane, the KT board is pasted on the facade closer to the shooting position and temporarily fixed with tape.

[0120] Step 2: Image acquisition: Take photos of all exterior facades of the masonry structure. The four boundary lines of the facade are clearly visible, and the main building is complete and occupies most of the picture. Use frontal photography. When frontal photography is not possible, the angle of the oblique shooting should not be too large, and there should be no people or other obstructions in the picture.

[0121] Step 3: Perform target recognition on the image collected in step 2 to output a binary image: Use the SegFormer model to recognize the outlines of the house facade, door and window openings, and signs in the image, such as Figure 3(a) to Figure 3(b)As shown, the output is a binary image. The model has only about 3.7M parameters. Compared with commonly used networks such as UNet and DeepLabV3+, it has better segmentation ability, anti-interference ability, and generalization ability, and is suitable for real-time detection scenarios. The training method of the SegFormer model is as follows:

[0122] Step 31, Dataset Preparation: Take photos of the house facades with markers, use Photoshop to accurately annotate the houses, door and window openings, and markers, and divide the obtained dataset into training, validation, and test sets at a ratio of 9:1:1.

[0123] Step 32: Model training: Initialize the SegFormer model using the weights pre-trained on Image-1K and train it. Take the optimal weight on the validation set, that is, the weight with the highest average Dice coefficient on the validation set, and test it on the test set to verify the model training effect.

[0124] Step 4: Based on the binary image output in step 3, perform parameter measurement and obtain the total number of floors of the building. The parameters include the actual size and location of the building facade, door and window openings, and landmarks.

[0125] like Figure 4(a) to Figure 4(c) As shown in , calculate the actual size and position of the building facade, door and window openings and signs respectively; Figure 5(a) to Figure 5(c) As shown in the figure, the MobileViTV3 model is used to classify images according to the number of floors of the house and output the total number of floors of the house.

[0126] The steps for calculating the actual size and location of the building facade, door and window openings, and signs in Step 4 are as follows:

[0127] Step 4A: Use OpenCV to correct the deformation of the marker. Count the required pixel size of the house facade and door and window openings (pixel) and the number of pixels occupied by the side length of the marker (mark_pixel). Calculate the actual size using the following formula:

[0128] L = pixel / mark_pixel × 30cm

[0129] Where L refers to the actual dimensions of the building facade or door and window openings. Specifically, it can be replaced by the width of the building outline (house_width), height (house_height), and center coordinates (house_center_x and house_center_y); the width of the marker (mark_width), height (mark_height), and center coordinates (mark_center_x and mark_center_y); and the width of each door and window opening (holes_width), height (holes_height), and center coordinates (holes_center_x and holes_center_y). The origin of the image pixel coordinates is located in the upper-left corner of the image, with the positive x direction pointing rightward and the positive y direction pointing downward.

[0130] The MobileViTV3 model training and recognition method in step 4 is as follows:

[0131] Step 4B1. Dataset Preparation: As most old houses are low-rise masonry structures, with 1 to 3 stories being the most common, to ensure uniformity in training data across different classification categories, the binary images processed in Step 3 are divided into four categories based on the number of floors: 1 floor, 2 floors, 3 floors, and 3 or more floors. When an image is classified into the fourth category of 3 or more floors, manual intervention is performed to determine the number of floors. A data set for floor classification is constructed, and the data set for each category is divided into training, validation, and test sets at a ratio of 9:1:1.

[0132] Step 4B2, model training: Initialize the MobileViTV3 model using the weights pre-trained on Image-1K and train it. Take the weight that achieves the highest accuracy on the validation set and test it on the test set to verify the model training effect;

[0133] Step 4B3, layer number identification: the binary image processed in step 3 is input into the MobileViTV3 model after being processed in step 4 for layer number identification.

[0134] Step 5. Calculate the floor height and opening ratio: Based on the total house height (the height of the house outline) house_height and the total number of floors level obtained in step 4, calculate the floor height H of each floor as follows:

[0135] H = house_height / level

[0136] Where level is a constant, indicating the total number of floors of the house; i is a variable, ranging from 1 to level.

[0137] Calculate the wall opening ratio ω i This is used to calculate the lateral stiffness K in the subsequent step 6: it is calculated by dividing the total area of ​​the openings on each floor by the wall area of ​​that floor.

[0138] Among them, the area of ​​a single opening = door and window opening width × door and window opening height; the total area of ​​the openings = the sum of the areas of all openings on this floor.

[0139] Wall area = house outline width × floor height H.

[0140] Step 6: Based on the floor height H obtained in step 5, perform the structural equivalent total gravity load G eq Calculated with the lateral stiffness K, the equivalent total gravity load G eq Used for seismic verification and rating in the subsequent step 7.

[0141] The equivalent total gravity load G of the structure in step 6 eq The calculation method is as follows:

[0142] Step 6A1. Obtain the house_width of different facades from step 4, take the average of the widths of the two parallel walls, and determine the horizontal width L of the house. x (The average width of the two parallel walls) and the longitudinal width L y (Take the average width of the two parallel walls) and calculate the area A of each floor according to the following formula:

[0143] A=L x ·L y .

[0144] Step 6A2: Determine the wall thickness B and the wall material bulk density q based on local statistical data. m , floor slab material bulk density q f , floor thickness h f , combined with the standard value and the storey height H obtained in step 5 and the horizontal width L of the house obtained in step 6A1 x , vertical width L y , the area A of each floor, calculate the dead load q of each floor d and live load q l And combine them.

[0145] For example Figure 4(a) to Figure 4(c) In the case of , the dead load is taken as the sum of the wall deadweight, floor deadweight and other permanent loads, and then converted to unit area:

[0146]

[0147] The calculation method for the deadweight of walls and floor slabs is: calculate the volume of the wall or floor slab and multiply it by the bulk density of the material; other permanent loads are calculated based on the actual situation and take 1.5kN / m 2 Live load q l Take q according to the actual situation l =2.0kN / m 2 .

[0148] Step 6A3: Combine the area A of each layer obtained in step 6A1 and the dead load q obtained in step 6A2. d and live load q l Calculate the representative value of gravity load G of each layer according to the following formula: eq,i :

[0149] G eq,i =A·(q d +q l )

[0150] Calculate the equivalent total gravity load G of the structure according to the following formula: eq :

[0151]

[0152] The calculation method of lateral stiffness K in step 6 is as follows: Determine the elastic modulus E and shear modulus G of masonry from local statistical data, and calculate the equivalent cross-sectional area A of the i-th layer wall w,i , axial pressure ratio v i , the equivalent section inertia moment I of the wall i , combined with the floor height H and opening rate ω obtained in step 5 i , calculate the elastic stiffness of each layer of wall according to the following formula:

[0153]

[0154] Where β is the boundary condition parameter, which is 12 when level = 1 and 3 when level ≥ 2. The elastic stiffness K of two parallel walls i The sum of these values ​​is the lateral stiffness K of the wall in that direction. As mentioned in the technical background, older buildings in my country are primarily brick masonry structures, with the majority being single-story buildings. Exterior walls bear the majority of earthquake loads, while interior walls account for a smaller proportion and are mostly partition walls. Therefore, we conservatively ignore the contribution of interior walls to lateral stiffness. The calculation steps for the remaining parameters are as follows:

[0155] Step 6B1: Equivalent cross-sectional area A of each wall w,i The calculation steps are as follows:

[0156] Step 6B11. Count the number of openings on each floor: Calculate the upper and lower boundaries of the house outline using the y-axis center coordinate house_center_y and the height house_height obtained in Step 4. Combined with the floor height H, determine the upper and lower boundaries of the i-th floor using the following formula:

[0157] upLine=house_center_y+house_height / 2-i×H

[0158] downLine=house_center_y+house_height / 2-(i-1)×H

[0159] When the value of the hole's ordinate, holes_center_y, is between the upper and lower dividing lines, the hole is considered to be located on the i-th layer. Traverse the ordinates of each hole, determine the layer number of each hole, and count the number of holes on each layer.

[0160] Taking Figure 4(a) as an example, the upper and lower dividing lines of the first layer are:

[0161] upLine=3438+2904 / 2-1×2904=1986

[0162] downLine=3438+2904 / 2-(1-1)×2904=4890

[0163] As shown in Figure 4(c), the y coordinate of the doorway center is 3689, which is between 1986 and 4890, so it is judged that the doorway is located on the 1st floor.

[0164] Calculate the left and right limits of the house outline using the x-direction center coordinate house_center_x and width house_width of the house facade obtained in step 4:

[0165] LeftLine=house_center_x-house_width / 2

[0166] rightLine=house_center_x+house_width / 2

[0167] Step 6B12, as Figure 6 As shown, for the number of door and window openings on the i-th layer obtained in step 6B11, the wall of this layer is divided according to the door and window opening boundaries and their extension lines, and the coordinates of the opening boundaries are calculated and stored in a list according to the following formula:

[0168] edge_x_i=holes_center_x±Holes_width / 2

[0169] edge_y_i=holes_center_y±holes_height / 2

[0170] by Figure 6For example, the upper boundary coordinates of the left window are 3194-1247 / 2=2571, the upper boundary coordinates of the right window are 3189-1245 / 2=2567, and the upper boundary coordinates of the doorway are 3689-2243 / 2=2568. These three are essentially the same boundary line and need to be merged. Add the upper and lower boundary lines upLine and downLine obtained in step 6B11 to the edge_y_i list and sort the data in the edge_y_i list. When the difference between two adjacent boundary coordinates in the list is less than a threshold Δ=0.05×H=0.05×2904=145, merge them into the same boundary coordinate by taking the average. If there are multiple similar coordinates, merge the first and second coordinates in order, then merge the resulting coordinate with the third coordinate, and so on. The final coordinates after merging the above boundary lines are ((2567+2568) / 2+2571) / 2=2570. Add the left and right boundaries of the house outline obtained in step 6B11, LeftLine and rightLine, to the edge_x_i list. The sorting and merging methods are the same as those for edge_y_i. Count the number of elements in the edge_y_i list, edge_y_num, and the number of cross sections, n, is calculated as follows:

[0171] n=edge_y_num-1

[0172] by Figure 6 For example, after the edge_y_i list coordinates are merged, there are 4 elements left, and the number of cross sections can be calculated as n=3.

[0173] Step 6B13: Measure the width l of the jth column after the area is divided ij and the height h of row k ik :like Figure 7 As shown, the adjacent elements of the edge_x_i list obtained in step 6B12 are subtracted from each other to obtain the width of each column l ij Data list; subtract adjacent elements of the edge_y_i list obtained in step 6B12 to obtain the height h of each row ik Data list. It should be noted that k is a variable and n is a constant, and the value range of k is 1 to n.

[0174] According to the number of cross sections n, the door and window opening boundaries and their extension lines, the opening conditions of each layer are calculated to construct a two-dimensional list isHole: Figure 8 As shown, according to (each layer) the i-th layer l ij The number of i,size and h ik The number of h i,sizeConstruct a two-dimensional list isHole, with the default element being 1. When the area at the jth column and the kth row of the wall is a hole area, change the corresponding element of isHole to 0.

[0175] Step 6B14: Determine the vertical direction of l based on local statistical data. ij The wall thickness B in the direction, the width of each column l ij Multiply by the wall thickness B to get the cross-sectional area of ​​the wall column. When encountering an open area, the isHole obtained in step 6B13 ijk The coefficient is 0, and after multiplication, it is no longer included in the cross-sectional area of ​​the wall section. Calculate the cross-sectional area A of the wall in the kth row according to the following formula ik :

[0176]

[0177] Step 6B15: Consider the contribution of the cross-sectional area of ​​each row of walls, and use the height h of the row obtained in step 6B13. ik As a weight, combined with the wall cross-sectional area A of the kth row region obtained in step 6B14 ik , calculate the equivalent cross-sectional area of ​​the wall by the following formula:

[0178]

[0179] Compare A in each row ik , take the smallest one as A w,min,i ,by Figure 7 For example, when k=2, the cross-sectional area is the smallest, and the A of this row is taken. i2 As A w,min,i The minimum cross-sectional area A of the four facade walls w,min,i Summing to get ∑A w,min,i , which is used for the axial compression ratio v of each layer in the subsequent step 6B2 i .

[0180] Step 6B2, axial compression ratio v of each layer i The calculation steps are as follows:

[0181] Step 6B21: The constant load q of each layer obtained in step 6A2 d and live load q l Calculate the design value of the axial force of the i-th layer N according to the following formula: i :

[0182]

[0183] When calculating the axial force for floor i, i and level are constants, representing the current floor and the total number of floors, respectively. t is a variable, ranging from i to level. For example, if a building has 5 floors and level = 5, when calculating the axial force for floor i = 3, the design loads for floors t = 3 through t = 5 are accumulated in the formula.

[0184] Step 6B22: Determine the masonry compressive strength f from local statistical data m , combined with the sum of the minimum cross-sectional areas ∑A obtained in step 6B15 w,min,i and the design value N of the axial force obtained in step 6B21 i , calculate the axial compression ratio of each layer according to the following formula:

[0185]

[0186] Step 6B3, equivalent section moment of inertia I of each layer of wall i The calculation steps are as follows:

[0187] Step 6B31: h obtained in step 6B13 ik , calculate the wall parameters of the k-th row area by the following formula

[0188]

[0189] Then the vertical direction of l is determined by local statistics. ij The wall thickness B in the direction, combined with the l obtained in step 6B13 ij and isHole ijk The coefficient is calculated by the following formula: the moment of inertia I of the wall in the jth column and the kth row to the center axis of the wall i k:

[0190]

[0191] Among them, D ij is the horizontal distance from the centroid of the j-th column area divided by the wall to the central axis of the wall, Figure 6 For example, the x-coordinate of the second column area centroid is 1418, and the x-coordinate of the wall center is 3758, then D i2 =3758-1418=2340.

[0192] Step 6B32: Use the wall parameters obtained in step 6B31 Consider the moment of inertia of each row of walls I for weight ik The contribution of , the equivalent section moment of inertia of each layer is calculated by the following formula:

[0193]

[0194] Step 7: Verify and rate the seismic performance of the structure using the inter-story displacement angle as an indicator: Determine the maximum value of the horizontal earthquake influence coefficient α based on the seismic fortification intensity of the area where the house is located and according to Table 5.1.4-1 of the Code for Seismic Design of Buildings GB 50011-2010 (2016 Edition) max , and directly used as the horizontal seismic influence coefficient value α1, combined with the structural equivalent total gravity load G obtained in step 6A4 eq , use the bottom shear method to calculate the standard value of the horizontal earthquake action of the structure F Ek :

[0195] F Ek =α1G eq

[0196] Combined with the representative value of gravity load G of each layer obtained in step 6A3 eq,i , calculate the standard value of horizontal earthquake action F of the i-th layer i :

[0197]

[0198] The interlayer shear force V of layer i is calculated by the following formula: i :

[0199]

[0200] Combined with the lateral stiffness K obtained in step 6, calculate the inter-story displacement angle Δθ of each layer under multiple earthquakes e,i :

[0201] Δθ e,i =V i / K

[0202] When the inter-story displacement angle Δθ e,i <1 / 1000, the structure rating is A; when 1 / 1000≤Δθ e,i <1 / 670, the structural rating is B; when 1 / 670≤Δθ e,i <1 / 330, the structural rating is C; when Δθ e,i When it is >1 / 330, the structural rating is D.

[0203] The preferred embodiments of the present invention are described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the scope of protection of the present invention.

Claims

1. A method for rapid assessment of seismic performance of masonry structures based on image recognition, characterized by: The following steps are involved: Step 1: Paste markers on all exterior walls of the masonry structure to serve as reference for image distortion correction and actual size conversion; Step 2, image acquisition: take photos of all facades of the masonry structure; Step 3: Perform target recognition on the image collected in step 2 to output a binary image: Use the SegFormer model to identify the outlines of the house facade, door and window openings, and signs in the image; Step 4: Based on the binary image output in step 3, perform parameter measurement and obtain the total number of floors of the building. The parameters include the actual size and location of the building facade, door and window openings, and landmarks. Specifically, they include: Use OpenCV to correct the deformation of the marker, count the required pixel size of the house facade and door and window openings (pixel), and the number of pixels occupied by the side length of the marker (mark_pixel), and calculate the actual size using the following formula: L = pixel / mark_pixel × 30cm; The actual dimensions include the width, height, and center coordinates of the building outline, markers, and door and window openings; Use the MobileViTV3 model to classify images by the number of floors of the house and output the total number of floors of the house; Step 5. Calculate the floor height and opening ratio: Based on the total house height house_height and the total number of floors level obtained in step 4, calculate the floor height H of each floor as follows: H = house_height / level; Calculate the wall opening ratio ω i To be used to calculate the lateral stiffness K in the subsequent step 6; ω i = total area of ​​openings on each floor / wall area of ​​that floor; Step 6: Based on the floor height H obtained in step 5, perform the structural equivalent total gravity load G eq Calculate the lateral stiffness K for use in the subsequent seismic verification and rating in step 7: Among them, the equivalent total gravity load of the structure G eq : Among them, L x - The width of the house in the horizontal direction; L y - the longitudinal width of the house; B-wall thickness; q m - bulk density of wall materials; q f - Bulk density of floor slab material; h f - Floor slab thickness; Among them, the calculation of lateral stiffness K: determine the masonry elastic modulus E and shear modulus G based on local statistical data, and calculate the equivalent cross-sectional area A of the i-th layer wall w,i , axial pressure ratio v i , the equivalent section inertia moment I of the wall i , combined with the floor height H and opening rate ω obtained in step 5 i Calculate the elastic stiffness K of each layer of wall according to the following formula: i , the elastic stiffness K of the two parallel walls i The cumulative value is the lateral stiffness K of the wall in this direction; q d is the dead load; q l For live load; in, Among them, β is the boundary condition parameter; The specific steps include: Step 6B1. Calculate the equivalent cross-sectional area A of each wall layer w,i : Among them, A ik is the cross-sectional area of ​​the wall; in the i-th layer, the width l of the j-th column ij The number of ,size , the height h of the kth row ik The number of h i,size Construct a two-dimensional list isHole, with the default element being 1. When the area at the jth column and the kth row of the wall is a hole, set the corresponding element of isHole to 0. The edge_y_i list is composed of the boundary coordinates edge_y_i of the door and window openings, and edge_y_num is the number of elements in the edge_y_i list; The edge_x_i list is composed of the boundary coordinates edge_x_i of the door and window openings; Subtract two adjacent elements from the edge_x_i list to get the width of each column l ij Data list; Subtract two adjacent elements from the edge_y_i list to get the height h of each row ik Data list; Step 6B2: Calculate the axial compression ratio v of each layer i ; Among them, f m - Masonry compressive strength; A w,min,i is the minimum cross-sectional area of ​​the wall; N i is the design value of the axial force of the i-th layer; Step 6B3: Calculate the equivalent section inertia moment I of each wall layer i : D ij -The horizontal distance from the centroid of the j-th column of area divided by the wall to the central axis of the wall; is the wall parameter; I ik The moment of inertia of the wall about its neutral axis; Step 7: Take the interlayer displacement angle Δθ e,i Verify and rate the seismic performance of the structure based on the following indicators: F Ek =α1G eq ; Among them, F Ek is the standard value of the horizontal earthquake action on the structure; α1 is the horizontal earthquake influence coefficient value; F i is the standard value of horizontal earthquake action on the i-th floor; V i is the interlayer shear force of the i-th layer; Finally, according to the inter-story displacement angle Δθ e,i Perform structural rating.

2. The method for rapid assessment of seismic performance of masonry structures based on image recognition according to claim 1, characterized in that: The sign is made of KT board with a size of 30cm×30cm; one KT board is pasted on one wall; when the facades are not in the same plane, the KT board is pasted on the facade closer to the shooting position and temporarily fixed with tape.

3. The method for rapid assessment of seismic performance of masonry structures based on image recognition according to claim 1, characterized in that: In step 2, the four boundary lines of the house facade are clearly visible, the main house is complete and occupies most of the picture; the front view is used for shooting. When the front view cannot be shot, the angle of the oblique shot is not too large, and there are no people or other obstructions in the picture.

4. The method for rapid assessment of seismic performance of masonry structures based on image recognition according to claim 1, characterized in that: In step 3, the training method of the SegFormer model is as follows: (1) Dataset preparation: Take photos of the house facades with markers, use Photoshop to accurately annotate the houses, door and window openings, and markers, and divide the obtained dataset into training set, validation set, and test set at a ratio of 9:1:1; (2) Model training: Use the weights pre-trained on Image-1K to initialize the SegFormer model and train it. Take the optimal weight on the validation set, that is, the weight with the highest average Dice coefficient on the validation set. Test it on the test set to verify the model training effect.

5. The method for rapid assessment of seismic performance of masonry structures based on image recognition according to claim 1, characterized in that: In step 4, the layer identification method is as follows: Step 4B1. Dataset preparation: Divide the binary image processed in step 3 into four categories according to the number of layers: 1 layer, 2 layers, 3 layers, and more than 3 layers. Construct a layer classification dataset. Divide the dataset of each category into training set, validation set, and test set at a ratio of 9:1:

1. Step 4B2, model training: Initialize the MobileViTV3 model using the weights pre-trained on Image-1K and train it. Take the weight that achieves the highest accuracy on the validation set and test it on the test set to verify the model training effect; Step 4B3, layer number identification: the binary image processed in step 3 is passed through the MobileViTV3 model for layer number identification.

6. The method for rapid assessment of seismic performance of masonry structures based on image recognition according to claim 1, characterized in that: In step 6, the equivalent total gravity load G eq The calculation method is as follows: Step 6A1. Obtain the house_width of different facades from step 4, take the average of the widths of the two parallel walls, and determine the horizontal width L of the house. x and the vertical width L y , calculate the area A of each layer according to the following formula: A=L x ·L y ; Step 6A2: Determine the wall thickness B and the wall material bulk density q based on local statistical data. m , floor slab material bulk density q f , floor thickness h f , combined with the standard value and the storey height H obtained in step 5 and the horizontal width L of the house obtained in step 6A1 x , vertical width L y , the area A of each floor, calculate the dead load q of each floor d and live load q l ; Step 6A3: Combine the area A of each layer obtained in step 6A1 and the dead load q obtained in step 6A2. d and live load q l , calculate the representative value of gravity load G of each layer eq,i .

7. The method for rapid assessment of seismic performance of masonry structures based on image recognition according to claim 6, characterized in that: In step 6B1, the equivalent cross-sectional area A of each layer of wall w,i The calculation steps are as follows: Step 6B11. Count the number of openings on each floor: Calculate the upper and lower boundaries of the house outline using the y-axis center coordinate house_center_y and the height house_height obtained in Step 4. Combined with the floor height H, determine the upper and lower boundaries of the i-th floor using the following formula: upLine=house_center_y+house_height / 2-i×H; downLine=house_center_y+house_height / 2-(i-1)×H; When the value of the hole's vertical coordinate holes_center_y is between the upper and lower dividing lines, the hole is judged to be located in the i-th layer; Traverse the vertical coordinates of each hole, determine the layer number of each hole, and count the number of holes on each layer; Step 6B12, obtaining the number of cross sections n; Based on the number of door and window openings on the i-th layer obtained in step 6B11, divide the wall of this layer according to the door and window opening boundaries and their extension lines, calculate the coordinates of the opening boundaries according to the following formula, and store them in a list; edge_x_i=holes_center_x±holes_width / 2; edge_y_i=holes_center_y±holes_height / 2; Then sort the data in the edge_y_i list. When the difference between the boundary coordinates in the list is less than the threshold Δ = 0.05 × H, merge the two into the same boundary coordinate by taking the average; Sort the data in the edge_x_i list. When the difference between the boundary coordinates in the list is less than the threshold Δ = 0.05 × H, merge the two into the same boundary coordinate by taking the average; Finally, count the number of elements in the edge_y_i list edge_y_num; Step 6B13: Measure the width l of the jth column after the area is divided ij and the height h of row k ik ; Subtract the adjacent elements of the edge_x_i list obtained in step 6B12 from each other to obtain the width of each column l ij Data list; subtract adjacent elements of the edge_y_i list obtained in step 6B12 to obtain the height h of each row ik Data list; According to the number of cross sections n, the door and window opening boundaries and their extension lines, the opening conditions of each layer are calculated to construct a two-dimensional list isHole. isHole stores the opening information of the wall area. The value of isHole is 1 for the solid area and 0 for the opening area. Step 6B14: Determine the vertical direction of l based on local statistical data. ij The wall thickness B in the direction is calculated by the formula Calculate the cross-sectional area A of the wall in the kth row ik ; Step 6B15: Calculate the equivalent cross-sectional area of ​​the wall using the following formula: Compare A in each row ik , take the smallest one as A w,min,i , the minimum cross-sectional area A of the four facade walls w,min,i Summing to get ∑A w,min,i , which is used for the axial compression ratio v of each layer in the subsequent step 6B2 i .

8. The method for rapid assessment of seismic performance of masonry structures based on image recognition according to claim 7, characterized in that: In step 6B2, the axial compression ratio v of each layer i The calculation steps are as follows: Step 6B21: The constant load q of each layer obtained in step 6A2 d and live load q l Calculate the design value of the axial force of the i-th layer N according to the following formula i ; Step 6B22: Determine the masonry compressive strength f from local statistical data m , combined with the sum of the minimum cross-sectional areas ∑A obtained in step 6B15 w,min,i and the design value N of the axial force obtained in step 6B21 i , calculate the axial compression ratio of each layer according to the following formula:

9. The method for rapid assessment of seismic performance of masonry structures based on image recognition according to claim 7, characterized in that: In step 6B3, the equivalent section moment of inertia of each wall layer I i The calculation steps are as follows: Step 6B31: h obtained in step 6B13 ik , calculate the wall parameters of the k-th row area by the following formula Then the vertical direction of l is determined by local statistics. ij The wall thickness B in the direction, combined with the l obtained in step 6B13 ij and isHole ijk The coefficient is calculated by the following formula: the moment of inertia I of the wall in the jth column and the kth row to the center axis of the wall ik : And calculate the distance D from the centroid of the j-th column area divided by the wall to the central axis of the wall ij ; Step 6B32: Calculate the equivalent section moment of inertia of each layer using the following formula:

10. The method for rapid assessment of seismic performance of masonry structures based on image recognition according to claim 1, characterized in that: In step 7, when the interlayer displacement angle Δθ e,i <1 / 1000, the structure rating is A; when 1 / 1000≤Δθ e,i <1 / 670, the structural rating is B; when 1 / 670≤Δθ e,i <1 / 330, the structural rating is C; when Δθ e,i When it is greater than 1 / 330, the structural rating is D.

Images