A method for detecting a three-dimensional damage of a building

Abstract

The application discloses a kind of building three-dimensional damage detection methods, comprising: one, with unmanned aerial vehicle carrying visible light imager to target building image acquisition is carried out, obtain the multiple building images of covering entire target building, and generate building DSM model;Two, target building top elevation image acquisition;Three, target building elevation variation detection;Four, target building top damage detection;Five, target building side damage detection;Six, target building damage visualization.The method of the application is simple, reasonable in design, uses unmanned aerial vehicle multi-angle to shoot the image of target building, realizes the elevation variation, top damage and side damage detection of target building, realizes the overall visualization of target building damage.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of image recognition, and particularly relates to a building three-dimensional damage detection method. BACKGROUND

[0002] Change detection based on multi-temporal remote sensing images is to extract the changes of ground objects between different time phases through comparison and analysis of two or more remote sensing images of the same area imaged at different time phases. With the importance of building change detection in city development monitoring and damage assessment, building change detection technology based on multi-temporal remote sensing images has wide application prospects in civil and military fields. At present, there are many building change detection methods, which are applied to solve various problems.

[0003] However, the current building damage detection methods mainly focus on two-dimensional change detection, mainly using the gray difference between different time phase remote sensing images to detect the top damage information of the building or the facade damage information of the building. The main problem of this kind of method is that it can only detect the top change or side change of the building between different time phases, and cannot obtain comprehensive three-dimensional damage information, so the application in building disaster damage detection and strike effect evaluation is limited.

[0004] Therefore, a building three-dimensional damage detection method with reasonable design is needed, which uses the multi-angle image of the target building taken by the unmanned aerial vehicle to realize the detection of the height change, top damage and side damage of the target building, and realizes the overall visualization of the damage of the target building. SUMMARY

[0005] The technical problem to be solved by the application is to provide a building three-dimensional damage detection method, which has simple method steps and reasonable design, uses the multi-angle image of the target building taken by the unmanned aerial vehicle to realize the detection of the height change, top damage and side damage of the target building, and realizes the overall visualization of the damage of the target building.

[0006] To solve the above technical problems, the technical scheme adopted by the application is as follows: a building three-dimensional damage detection method, characterized in that the method comprises the following steps:

[0007] Step 1: use the unmanned aerial vehicle to carry the visible light imager to collect the image of the target building, obtain multiple building images covering the entire target building, and generate a building DSM model;

[0008] Step 2: obtain the top elevation image of the target building:

[0009] obtaining building top elevation data from the generated building DSM model by using a computer, and taking the building top elevation data as a building top elevation image;

[0010] Step three, detection of target building elevation change:

[0011] Step 301, obtaining the building top elevation images before and after damage according to the method of step one and step two, and taking them as the building top elevation image before damage and the building top elevation image after damage respectively;

[0012] Step 302, using a computer to register and perform difference operation on the building top elevation image before damage and the building top elevation image after damage, and obtaining the target building elevation change area;

[0013] Step four, detection of target building top damage:

[0014] Step 401, obtaining the image containing the target building top taken from the top view from multiple building images, and taking the images containing the target building top before and after damage as the building top image before damage and the building top image after damage respectively;

[0015] Step 402, using a computer to register the building top image before damage and the building top image after damage, and processing them through CVA algorithm and fuzzy C-means clustering algorithm, and obtaining the target building top damage area;

[0016] Step five, detection of target building side damage:

[0017] Step 501, obtaining the image containing the target building side taken from the side view from multiple building images, and taking the images containing the target building side before and after damage as the building side image before damage and the building side image after damage respectively;

[0018] Step 502, using a computer to register the building side image before damage and the building side image after damage by using SIFT feature matching method, and obtaining the registered building side image before damage and the registered building side image after damage;

[0019] Step 503, using a computer to perform empirical mode decomposition and reconstruction on the registered building side image before damage and the registered building side image after damage respectively by using two-dimensional empirical mode decomposition method, and obtaining the building side reconstructed image before damage with shadow eliminated and the building side reconstructed image after damage with shadow eliminated;

[0020] Step 504, HOG feature extraction is performed on the shadow-eliminated pre-damage building side reconstruction image and the shadow-eliminated post-damage building side reconstruction image respectively, and Euclidean distance judgment is performed, to obtain a target building side damage area;

[0021] Step six, target building damage visualization:

[0022] Step 601, obtain building images containing a target building top and side from multiple building images, and mark the building images as to-be-displayed building images;

[0023] Step 602, mark target building elevation change areas, target building top damage areas and target building side damage areas on the to-be-displayed building images using a computer, to realize overall visualization of target building damage.

[0024] The above-mentioned building three-dimensional damage detection method is characterized in that: step one, the specific process is as follows:

[0025] Step 101, use a UAV to carry a visible light imager to collect omnidirectional multi-view images of a target building, to obtain multiple building images covering the entire target building, and transmit the collected multiple building images and POS data corresponding to the multiple building images to a computer;

[0026] Step 102, use a computer to import the multiple building images and the POS data corresponding to the multiple building images into Agisoft PhotoScan software, and use the Agisoft PhotoScan software to process the multiple building images and the POS data corresponding to the multiple building images, to generate a building DSM model.

[0027] The above-mentioned building three-dimensional damage detection method is characterized in that: step 302, the specific process is as follows:

[0028] Step 3021, use a computer to use an image matching algorithm based on FFT phase correlation to register a pre-damage building top elevation image and a post-damage building top elevation image, to obtain a registered pre-damage building top elevation image and a registered post-damage building top elevation image;

[0029] Step 3022, use a computer to perform difference operation on the registered pre-damage building top elevation image and the registered post-damage building top elevation image, to obtain an elevation difference image;

[0030] Step 3023, use a computer to perform binaryzation processing on the elevation difference image according to a set elevation threshold, to obtain an elevation difference binaryzation image; wherein, a pixel value greater than the elevation threshold is set to 255, and a pixel value less than the elevation threshold is set to 0.

[0031] Step 3024, using a computer to record the white area in the elevation difference binary image as the building elevation change area.

[0032] The building three-dimensional damage detection method has the characteristics that step 402 has the following specific process:

[0033] Step 4021, using a computer to use the SIFT feature matching method to register the pre-damage building top image and the post-damage building top image, and obtain the registered pre-damage building top image and the registered post-damage building top image;

[0034] Step 4022, using a computer to use the CVA algorithm to process the change intensity of the R component, the G component and the B component in the registered pre-damage building top image and the registered post-damage building top image, and obtain the target building top difference image;

[0035] Step 4023, using a computer to use the fuzzy C-means clustering algorithm to cluster and segment the target building top difference image, and obtain the clustered and segmented target building top difference image; wherein the number of clusters is 2, and the clustered and segmented target building top difference image includes 2 cluster regions;

[0036] Step 4024, using a computer to obtain the pixel mean value of the 2 cluster regions, and the cluster region with a larger pixel mean value is a cluster change region, and is recorded as the target building top damage region.

[0037] The building three-dimensional damage detection method has the characteristics that step 402 has the following specific process:

[0038] Step 5031, using a computer to use the two-dimensional empirical mode decomposition method to decompose the registered pre-damage building side image to obtain N IMF components and a trend item; N is a positive integer;

[0039] Step 5032, using a computer to obtain the gray level histogram corresponding to the i-th IMF component according to the i-th IMF component; wherein i is a positive integer, and 1≤i≤N;

[0040] Step 5033, using a computer to obtain the proportion of the number of pixels less than the gray threshold to the total pixels from the gray level histogram corresponding to the i-th IMF component, and if the proportion of the number of pixels less than the gray threshold to the total pixels is greater than the proportion threshold, the IMF component is a shadow component, and the IMF component is removed;

[0041] Step 5034, using a computer to reconstruct each IMF component after removing the shadow component and a trend item to obtain a pre-damage building side reconstruction image.

[0042] Step 5035, using a computer to obtain a shadow area from the registered pre-damage building side image, and obtain a reconstructed image corresponding to the shadow area from the pre-damage building side reconstructed image, denoted as a region reconstructed image;

[0043] Step 5036, using a computer to replace the shadow area with the region reconstructed image on the registered pre-damage building side image, and then obtain a pre-damage building side reconstructed image with shadows removed;

[0044] Step 5037, according to the method of step 5031 to step 5036, performing empirical mode decomposition and reconstruction on the registered post-damage building side image to obtain a post-damage building side reconstructed image with shadows removed.

[0045] The above-mentioned building three-dimensional damage detection method is characterized in that step 504 has the following specific process:

[0046] Step 5041, using a computer to perform HOG feature extraction on the pre-damage building side reconstructed image with shadows removed according to a window size of 3x3 and a sliding step size of 1, to obtain a HOG feature vector descriptor of a center pixel of an Mth window; wherein M is a positive integer, and M represents the total number of windows in HOG feature extraction;

[0047] Step 5042, using a computer to normalize the HOG feature vector descriptor of the center pixel of the jth window to obtain a HOG normalized feature vector of the center pixel of the jth window; wherein j is a positive integer, and 1≤j≤M;

[0048] Step 5043, using a computer to perform HOG feature extraction and normalization on the post-damage building side reconstructed image with shadows removed to obtain a HOG normalized feature vector of the center pixel of the jth window;

[0049] Step 5044, obtaining the Euclidean distance dj between the HOG normalized feature vector of the center pixel of the jth window before damage and the HOG normalized feature vector of the center pixel of the jth window after damage;

[0050] Step 5045, using a computer to judge dj and the Euclidean distance threshold value, if dj is greater than the Euclidean distance threshold value, then the center pixel of the jth window is a changed pixel;

[0051] Step 5046, using a computer to take the connected region formed by the changed pixel as the target building side damage area.

[0052] The above-mentioned method for detecting three-dimensional damage to buildings is characterized in that: in step 5035, a computer is used to obtain the shadow area from the registered side image of the building before damage, as detailed below:

[0053] Step A: Use a computer to transform the RGB color space corresponding to the registered side image of the building before damage to the HSI color space to obtain the corresponding hue components. saturation component and brightness component ;

[0054] Step B: Use a computer to calculate the formula. Obtain the ratio image The Otsu's method was used to contrast the images. Processing to obtain a single threshold And obtain the ratio image. Medium pixel value greater than a single threshold The pixels;

[0055] Step C: Using a computer, identify elements in the registered pre-damage side images of the building that are larger than a single threshold. The region corresponding to each pixel is used as the shadow segmentation image. ;

[0056] Step D: Obtain shadow segmentation images using a computer. The corresponding B-channel component image is processed using the Otsu's method to obtain a single threshold. And obtain the pixel values ​​in the B-channel component image that are less than a single threshold. The pixels;

[0057] Step E: Use a computer to segment the image in the shadows. Lieutenant General less than a single threshold The region corresponding to each pixel is used as the shadow segmentation image. ;

[0058] Step F: Obtain shadow segmentation images using a computer. The corresponding saturation component, and the saturation value component that is greater than the set saturation threshold. The pixels;

[0059] Step G: Use a computer to segment the image in the shadows. Lieutenant General greater than the threshold The regions corresponding to the pixels are processed as connected regions and used as shadow regions.

[0060] Compared with the prior art, the present invention has the following advantages:

[0061] 1. The method of the present application has simple steps and reasonable design, and solves the problem that the current building damage detection method cannot obtain comprehensive three-dimensional damage information of the building.

[0062] 2. The present application registers the pre-damage building top image and the post-damage building top image, and processes them through CVA algorithm and fuzzy C-means clustering algorithm to obtain the target building top damage area. The CVA algorithm can quickly detect the changes of the registered two-phase images, improving the detection speed.

[0063] 3. The present application uses two-dimensional empirical mode decomposition method to perform empirical mode decomposition and reconstruction on the registered pre-damage building side image and post-damage building side image, and then performs HOG feature extraction and Euclidean distance judgment. The shadows of the target building caused by different imaging conditions before and after damage are eliminated as much as possible through empirical mode decomposition and reconstruction, so as to avoid the influence of shadows on damage detection. Through HOG feature extraction and normalization of the reconstructed image after eliminating shadow interference, the shape and structure changes of the target building before and after damage can be more accurately represented.

[0064] 4. The present application uses the target building top elevation image to realize the target building elevation change detection, combines with the target building top damage area detection, adapts to different degrees of damage of the target building top, and realizes the overall visualization of the target building damage, which is convenient for intuitive viewing.

[0065] In summary, the method of the present application has simple steps and reasonable design, and uses the images of the target building taken by the unmanned aerial vehicle from multiple angles to realize the detection of the elevation change, top damage and side damage of the target building, and realizes the overall visualization of the target building damage.

[0066] The technical solutions of the present application will be further described in detail below with the help of the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0067] Figure 1 The present application is a method flow chart.

[0068] Figure 2 The present application is a target building damage overall visualization diagram. DETAILED DESCRIPTION

[0069] As shown in the figure, the building three-dimensional damage detection method of the present application comprises the following steps: Figure 1 Step one, use the unmanned aerial vehicle to carry the visible light imager to collect the images of the target building, obtain multiple building images covering the entire target building, and generate a building DSM model;

[0070]

[0071] ​Step two, obtaining the target building top elevation image:

[0072] Obtaining the building top elevation data from the generated building DSM model by using the computer, and taking the building top elevation data as the building top elevation image;

[0073] Step three, detecting the target building elevation change:

[0074] Step 301, obtaining the building top elevation images before and after damage according to the methods of step one and step two, and recording them as the building top elevation image before damage and the building top elevation image after damage respectively;

[0075] Step 302, using the computer to register and perform difference operation on the building top elevation image before damage and the building top elevation image after damage, and obtaining the target building elevation change area;

[0076] Step four, detecting the target building top damage:

[0077] Step 401, obtaining the image containing the target building top taken from the top view from multiple building images, and recording the images containing the target building top before and after damage as the building top image before damage and the building top image after damage respectively;

[0078] Step 402, using the computer to register the building top image before damage and the building top image after damage, and processing them through the CVA algorithm and the fuzzy C-means clustering algorithm, to obtain the target building top damage area;

[0079] Step five, detecting the target building side damage:

[0080] Step 501, obtaining the image containing the target building side taken from the side view from multiple building images, and recording the images containing the target building side before and after damage as the building side image before damage and the building side image after damage respectively;

[0081] Step 502, using the computer to register the building side image before damage and the building side image after damage by using the SIFT feature matching method, to obtain the registered building side image before damage and the registered building side image after damage;

[0082] Step 503, using the computer to perform empirical mode decomposition and reconstruction on the registered building side image before damage and the registered building side image after damage respectively by using the two-dimensional empirical mode decomposition method, to obtain the building side reconstructed image before damage with shadow eliminated and the building side reconstructed image after damage with shadow eliminated;

[0083] Step 504, HOG feature extraction is performed on the shadow-eliminated pre-damage building side reconstruction image and the shadow-eliminated post-damage building side reconstruction image respectively, and Euclidean distance judgment is performed, to obtain a target building side damage area;

[0084] Step six, target building damage visualization:

[0085] Step 601, a building image containing a target building top and side from a plurality of building images is recorded as a building image to be displayed;

[0086] Step 602, a computer is used to mark a target building elevation change area, a target building top damage area and a target building side damage area on the building image to be displayed, to realize overall visualization of target building damage.

[0087] In this embodiment, step one, the specific process is as follows:

[0088] Step 101, a visible light imager is carried by a UAV to perform omnidirectional multi-view image acquisition on a target building, to obtain a plurality of building images covering the entire target building, and the plurality of building images acquired and the POS data corresponding to the plurality of building images are transmitted to a computer;

[0089] Step 102, a computer is used to import the plurality of building images and the POS data corresponding to the plurality of building images into Agisoft PhotoScan software, and Agisoft PhotoScan software is used to process the plurality of building images and the POS data corresponding to the plurality of building images, to generate a building DSM model.

[0090] In this embodiment, step 302, the specific process is as follows:

[0091] Step 3021, a computer is used to perform registration on a pre-damage building top elevation image and a post-damage building top elevation image by using an image matching algorithm based on FFT phase correlation, to obtain a registered pre-damage building top elevation image and a registered post-damage building top elevation image;

[0092] Step 3022, a computer is used to perform difference operation on the registered pre-damage building top elevation image and the registered post-damage building top elevation image, to obtain an elevation difference image;

[0093] Step 3023, a computer is used to perform binaryzation processing on the elevation difference image according to a set elevation threshold, to obtain an elevation difference binaryzation image; wherein a pixel value greater than the elevation threshold is set to 255, and a pixel value less than the elevation threshold is set to 0;

[0094] Step 3024, using the computer to record the white region in the elevation difference binary image as the building elevation change region.

[0095] In this embodiment, step 402, the specific process is as follows:

[0096] Step 4021, using the computer to use the SIFT feature matching method to register the pre-damage building top image and the post-damage building top image, and obtain the registered pre-damage building top image and the registered post-damage building top image;

[0097] Step 4022, using the computer to use the CVA algorithm to process the change intensity of the R component, the G component and the B component in the registered pre-damage building top image and the registered post-damage building top image, and obtain the target building top difference image;

[0098] Step 4023, using the computer to use the fuzzy C-means clustering algorithm to cluster and segment the target building top difference image, and obtain the clustered and segmented target building top difference image; wherein the number of clusters is 2, and the clustered and segmented target building top difference image includes two types of cluster regions;

[0099] Step 4024, using the computer to obtain the pixel mean value of the two types of cluster regions, and the cluster region with a larger pixel mean value is a cluster change region, and is recorded as the target building top damage region.

[0100] In this embodiment, step 503, the specific process is as follows:

[0101] Step 5031, using the computer to use the two-dimensional empirical mode decomposition method to decompose the registered pre-damage building side image to obtain N IMF components and a trend item; N is a positive integer;

[0102] Step 5032, using the computer to obtain the gray level histogram corresponding to the i-th IMF component according to the i-th IMF component; wherein i is a positive integer, and 1≤i≤N;

[0103] Step 5033, using the computer to obtain the proportion of the number of pixels less than the gray threshold to the total pixels from the gray level histogram corresponding to the i-th IMF component, and if the proportion of the number of pixels less than the gray threshold to the total pixels is greater than the proportion threshold, the IMF component is a shadow component, and the IMF component is removed;

[0104] Step 5034, using the computer to reconstruct each IMF component after removing the shadow component and a trend item to obtain a pre-damage building side reconstruction image;

[0105] Step 5035, using a computer to obtain a shadow area from the registered pre-damage building side image, and to obtain a reconstructed image corresponding to the shadow area from the pre-damage building side reconstructed image, denoted as a region reconstructed image;

[0106] Step 5036, using a computer to replace the shadow area with the region reconstructed image on the registered pre-damage building side image, thereby obtaining a pre-damage building side reconstructed image with shadows eliminated;

[0107] Step 5037, according to the method of steps 5031 to 5036, performing empirical mode decomposition and reconstruction on the registered post-damage building side image to obtain a post-damage building side reconstructed image with shadows eliminated.

[0108] In this embodiment, step 504, the specific process is as follows:

[0109] Step 5041, using a computer to perform HOG feature extraction on the pre-damage building side reconstructed image with shadows eliminated according to a window of 3x3 and a sliding step of 1, to obtain a HOG feature vector descriptor of a pre-damage M window center pixel; wherein M is a positive integer, and M represents the total number of windows in HOG feature extraction;

[0110] Step 5042, using a computer to normalize the HOG feature vector descriptor of the pre-damage jth window center pixel to obtain a HOG normalized feature vector of the pre-damage jth window center pixel; wherein j is a positive integer, and 1≤j≤M;

[0111] Step 5043, using a computer to perform HOG feature extraction and normalization on the post-damage building side reconstructed image with shadows eliminated to obtain a HOG normalized feature vector of the jth window center pixel;

[0112] Step 5044, obtaining the Euclidean distance dj between the HOG normalized feature vector of the pre-damage jth window center pixel and the HOG normalized feature vector of the post-damage jth window center pixel;

[0113] Step 5045, using a computer to judge dj and the Euclidean distance threshold, if dj is greater than the Euclidean distance threshold, then the center pixel of the jth window is a changed pixel;

[0114] Step 5046, using a computer to take the connected region formed by the changed pixel as the target building side damage area.

[0115] In this embodiment, in step 5035, the computer obtains a shadow area from the registered pre-damage building side image, specifically as follows:

[0116] Step A, using a computer to transform the registered RGB color space of the pre-damage building side image into HSI color space, to obtain the corresponding hue component , saturation component and brightness component ;

[0117] Step B, using a computer to obtain the ratio image according to the formula , using the maximum inter-class variance method to process the ratio image , to obtain a single threshold , and obtaining the pixel points in the ratio image with pixel values greater than the single threshold ;

[0118] Step C, using a computer to take the region corresponding to the pixel points greater than the single threshold in the registered pre-damage building side image as the shadow segmentation image ;

[0119] Step D, using a computer to obtain the B channel component image corresponding to the shadow segmentation image , using the maximum inter-class variance method to process the B channel component image, to obtain a single threshold , and obtaining the pixel points in the B channel component image with pixel values less than the single threshold ;

[0120] Step E, using a computer to take the region corresponding to the pixel points less than the single threshold in the shadow segmentation image as the shadow segmentation image ;

[0121] Step F, using a computer to obtain the saturation component corresponding to the shadow segmentation image , and obtaining the pixel points in the saturation component greater than a set saturation threshold ;

[0122] Step G, using a computer to perform connected region processing on the region corresponding to the pixel points greater than the threshold in the shadow segmentation image , as the shadow region.

[0123] In this embodiment, when specifically implemented, the heading overlap and the lateral overlap during image acquisition in step 101 are both not less than 30%.

[0124] In this embodiment, when specifically implemented, the elevation threshold in step 3023 is 3 meters, which can be adjusted according to actual needs.

[0125] ​In the embodiment, in the implementation, the step 4022 is specifically: performing difference value operation on the R component, the G component and the B component in the registered pre-damage building top image and the registered post-damage building top image respectively, to obtain an R component change value, a G component change value and a B component change value, and taking the R component change value, the G component change value and the B component change value as a three-component vector, and then calculating the amplitude of the three-component vector as the change intensity, so that the change intensity at each pixel point is the target building top difference map.

[0126] In the embodiment, in the implementation, the saturation threshold value is set to 30, which can be adjusted according to actual requirements. The value is 30, which can be adjusted according to actual requirements.

[0127] In the embodiment, in the implementation, the value of N in the step 5031 is 5, which can be adjusted according to actual requirements in actual use.

[0128] In the embodiment, in the implementation, the gray threshold value in the step 5033 is 20, and the proportion threshold value is 10%, which can be adjusted according to actual requirements in actual use.

[0129] In the embodiment, in the implementation, the HOG feature is a gradient direction histogram feature.

[0130] In the embodiment, in the implementation, the number of pixel points of the connected region formed in the step 5046 can be judged, and the connected region greater than a preset value is taken as a damage region to avoid the influence of isolated pixel points. The preset value can be 50, and can also be adjusted according to requirements.

[0131] In the embodiment, in the implementation, in the HOG feature extraction in the step 5041, the size of each window is 3*3, so that the HOG feature vector descriptor of the center pixel of each window is a one-dimensional feature vector of 8*1.

[0132] In the embodiment, in the implementation, the Euclidean distance threshold value in the step 5045 is 0.1, which can be adjusted according to actual requirements in actual use.

[0133] In the embodiment, the CVA algorithm is a change vector detection method, which can simply and quickly detect the changes of the registered pre-damage building top image and the registered post-damage building top image.

[0134] In the embodiment, by using the method, the detection of the height change, the top damage and the side damage of the target building is realized, and the overall visualization diagram of the damage of the target building as shown in FIG. 6 is obtained. Figure 2

[0135] ​In summary, the method of the present application has simple steps and reasonable design, uses the unmanned aerial vehicle to shoot the image of the target building from multiple angles, realizes the detection of the height change, top damage and side damage of the target building, and realizes the overall visualization of the damage of the target building.

[0136] The above is only the preferred embodiment of the present application, and does not limit the present application. Any simple modification, change and equivalent structural change of the above embodiment according to the technical essence of the present application are still within the protection scope of the technical solution of the present application.

Claims

1. A method for detecting a stereoscopic damage of a building, characterized by, The method comprises the following steps: Step one, use a UAV to carry a visible light imager to collect images of a target building, obtain multiple building images covering the entire target building, and generate a building DSM model; Step two, obtaining a target building top elevation image: Using a computer to obtain building top elevation data from the generated building DSM model, and taking the building top elevation data as the building top elevation image; Step three, detecting the height change of the target building: Step 301, according to the method of step one and step two, obtain the building top elevation images before and after damage, and mark them as the building top elevation image before damage and the building top elevation image after damage respectively; Step 302, using a computer to register and perform difference operation on the building top elevation image before damage and the building top elevation image after damage, and obtain the target building height change area; Step four, detecting the damage of the top of the target building: Step 401, obtain the image containing the top of the target building from the multiple building images, and mark the images containing the top of the target building before and after damage as the building top image before damage and the building top image after damage respectively; Step 402, using a computer to register the building top image before damage and the building top image after damage, and process them through CVA algorithm and fuzzy C-means clustering algorithm to obtain the target building top damage area; Step five, detecting the damage of the side of the target building: Step 501, obtain the image containing the side of the target building from the multiple building images, and mark the images containing the side of the target building before and after damage as the building side image before damage and the building side image after damage respectively; Step 502, using a computer to register the building side image before damage and the building side image after damage by SIFT feature matching method, and obtain the registered building side image before damage and the registered building side image after damage; Step 503, using a computer to perform empirical mode decomposition and reconstruction on the registered building side image before damage and the registered building side image after damage respectively by two-dimensional empirical mode decomposition method, and obtain the shadow area from the registered building side image before damage and after damage respectively, and obtain the reconstructed image corresponding to the shadow area from the reconstructed image of the building side before damage and after damage, and mark it as the area reconstructed image; Using a computer to replace the shadow area with the area reconstructed image on the registered building side image before damage and after damage, and obtain the shadow-eliminated building side reconstructed image before damage and after damage; Step 504, perform HOG feature extraction on the shadow-eliminated building side reconstructed image before damage and the shadow-eliminated building side reconstructed image after damage respectively, and perform Euclidean distance judgment to obtain the target building side damage area; Step six, visualizing the damage of the target building: Step 601, obtain the building image containing the top and side of the target building from the multiple building images, and mark it as the building image to be displayed; Step 602, mark the target building elevation change area, the target building top damage area and the target building side damage area on the building image to be displayed by using the computer, and realize the overall visualization of the target building damage; The method for obtaining the shadow area from the registered pre-damage building side image by using the computer is the same as that for obtaining the shadow area from the registered post-damage building side image, and the method for obtaining the shadow area from the registered pre-damage building side image by using the computer is as follows: Step A, transform the registered RGB color space of the pre-damage building side image pair to HSI color space by computer to obtain the corresponding hue component , saturation component and brightness component ; Step B: Use a computer to calculate the formula. Obtain the ratio image The Otsu's method was used to contrast the images. Processing to obtain a single threshold And obtain the ratio image. Medium pixel value greater than a single threshold The pixels; Step C. Using a computer to segment the shadow image from the registered pre- damage building side image by identifying regions of pixels greater than a single threshold value ;​ Step D, obtaining a shadow segmented image by computer corresponding to the B channel component image, processing the B channel component image by using the maximum inter-class variance method to obtain a single threshold ; and obtaining pixel points with pixel values less than the single threshold in the B channel component image; Step E, using a computer to segment the image into a shadow segmentation image corresponding to pixels less than a single threshold ;​ Step F: Obtain shadow segmentation images using a computer. The corresponding saturation component, and the saturation value component that is greater than the set saturation threshold. The pixels; Step G, using a computer to perform connected region processing on regions corresponding to pixels greater than a threshold in the shadow segmented image, as a shadow region. Step G, using a computer to perform connected region processing on regions corresponding to pixels greater than a threshold in the shadow segmented image, as a shadow region.​ 2. A method of volumetric damage detection of a building according to claim 1, wherein: Step one, the specific process is as follows: Step 101, use a UAV to carry a visible light imager to collect multi-view images of the target building in all directions, obtain multiple building images covering the entire target building, and transmit the collected multiple building images and the POS data corresponding to the multiple building images to a computer; Step 102, import the multiple building images and the POS data corresponding to the multiple building images into Agisoft PhotoScan software by using the computer, process the multiple building images and the POS data corresponding to the multiple building images by using Agisoft PhotoScan software, and generate a building DSM model.

3. A method for volumetric damage detection of a building according to claim 1, wherein: Step 302, the specific process is as follows: Step 3021, register the pre-damage building top elevation image and the post-damage building top elevation image by using the FFT phase correlation-based image matching algorithm by using the computer, to obtain the registered pre-damage building top elevation image and the registered post-damage building top elevation image; Step 3022, perform difference operation on the registered pre-damage building top elevation image and the registered post-damage building top elevation image by using the computer, to obtain an elevation difference image; Step 3023, perform binaryzation processing on the elevation difference image according to a set elevation threshold by using the computer, to obtain an elevation difference binaryzation image; wherein, the pixel value greater than the elevation threshold is set to 255, and the pixel value less than the elevation threshold is set to 0; Step 3024, mark the white area in the elevation difference binaryzation image as the building elevation change area by using the computer.

4. A method for volumetric damage detection of a building according to claim 1, wherein: Step 402, the specific process is as follows: Step 4021, register the pre-damage building top image and the post-damage building top image by using the SIFT feature matching method by using the computer, to obtain the registered pre-damage building top image and the registered post-damage building top image; Step 4022, perform change intensity processing on the R component, the G component and the B component in the registered pre-damage building top image and the registered post-damage building top image by using the CVA algorithm by using the computer, to obtain a target building top difference map; Step 4023, perform clustering segmentation on the target building top difference map by using the fuzzy C-means clustering algorithm by using the computer, to obtain a clustered target building top difference map; wherein, the number of clusters is 2, and the clustered target building top difference map includes two types of cluster areas; Step 4024, obtain the pixel mean value of the two types of cluster areas by using the computer, then the cluster area with a larger pixel mean value is a cluster change area, and mark it as the target building top damage area.

5. A method for volumetric damage detection of a building according to claim 1, wherein: Step 503, the specific process is as follows: Step 5031, using a computer to decompose the registered pre-damage building side image into N IMF components and a trend item by using two-dimensional empirical mode decomposition method; N is a positive integer; Step 5032, using a computer to obtain the gray histogram corresponding to the i-th IMF component according to the i-th IMF component; wherein, i is a positive integer, and 1≤i≤N; Step 5033, using a computer to obtain the proportion of the number of pixels less than the gray threshold to the total pixels from the gray histogram corresponding to the i-th IMF component, if the proportion of the number of pixels less than the gray threshold to the total pixels is greater than the proportion threshold, the IMF component is shadow component, and the IMF component is removed; Step 5034, using a computer to reconstruct each IMF component after removing the shadow component and a trend item to obtain a pre-damage building side reconstruction image; Step 5035, using a computer to obtain a shadow area from the registered pre-damage building side image, and obtain a reconstruction image corresponding to the shadow area from the pre-damage building side reconstruction image, denoted as a region reconstruction image; Step 5036, using a computer to replace the shadow area with the region reconstruction image on the registered pre-damage building side image, and obtain a pre-damage building side reconstruction image with shadows removed; Step 5037, according to the method of step 5031 to step 5036, the empirical mode decomposition and reconstruction are performed on the registered post-damage building side image to obtain a post-damage building side reconstruction image with shadows removed.

6. A method for volumetric damage detection of a building according to claim 1, wherein: Step 504, the specific process is as follows: Step 5041, using a computer to perform HOG feature extraction on the pre-damage building side reconstruction image with shadows removed according to a window of 3×3 and a sliding step of 1 to obtain HOG feature vector descriptors of M window center pixels; wherein, M is a positive integer, and M represents the total number of windows in HOG feature extraction; Step 5042, using a computer to normalize the HOG feature vector descriptor of the j-th window center pixel to obtain the HOG normalized feature vector of the j-th window center pixel; wherein, j is a positive integer, and 1≤j≤M; Step 5043, using a computer to perform HOG feature extraction and normalization on the post-damage building side reconstruction image with shadows removed to obtain the HOG normalized feature vector of the j-th window center pixel; Step 5044, obtaining the Euclidean distance dj between the HOG normalized feature vector of the j-th window center pixel and the HOG normalized feature vector of the j-th window center pixel; Step 5045, using a computer to judge dj and the Euclidean distance threshold, if dj is greater than the Euclidean distance threshold, the j-th window center pixel is a changed pixel; Step 5046, using a computer to take the connected region formed by the changed pixel as the target building side damage area.

Images