A VR panoramic method for collecting earthquake damage information of single buildings

Abstract

The embodiments of the present application relate to the field of image processing technology, and specifically, to a method for collecting earthquake damage information of single buildings with a VR panoramic view. The method can comprehensively collect data by acquiring a panoramic image of a target single building and a reference panoramic image with earthquake damage annotations. Vector mining is performed on the reference panoramic image with earthquake damage annotations to obtain corresponding vectors, which can accurately reflect the special state of the earthquake damage scene. Vector mining of the panoramic image of the target single building can characterize its own characteristics in detail. Finally, a panoramic image of a new state is generated based on the two vectors to achieve an update from an initial state to a second state related to earthquake damage. This helps to accurately display the condition of the building after it may suffer earthquake damage, and provides a comprehensive and accurate image basis for the evaluation of the seismic performance of the building and the analysis of earthquake damage conditions, thereby improving the accuracy and reliability of research on earthquake damage characteristics of buildings.

Description

Technical Field

[0001] The present application relates to the field of image processing technology, and in particular to a method for collecting earthquake damage information of single buildings in a VR panoramic view. Background Art

[0002] In the research field of earthquake emergency response and earthquake disaster prevention, it is crucial to accurately grasp the seismic performance and damage of buildings. Traditional information collection and display methods are difficult to fully and accurately reflect the state of buildings after earthquake damage. The evaluation of buildings often lacks the basis for complete image conversion from the initial state to the earthquake damage state, resulting in inaccurate seismic performance evaluation and lack of intuitiveness and accuracy in earthquake damage analysis. Existing technologies make it difficult to integrate building information in different states and display it in an intuitive and effective way. Summary of the invention

[0003] In order to at least overcome the above-mentioned deficiencies in the prior art, one of the objectives of the present application is to provide a method for collecting earthquake damage information of single buildings with a VR panoramic view.

[0004] The embodiment of the present application provides a method for collecting earthquake damage information of a single building in a VR panoramic view, the method being applied to an information collection system, the method comprising:

[0005] Acquire a panoramic image of a target single building in a first VR modeling and annotation state, and acquire a reference panoramic image of earthquake damage annotation of a target earthquake damage annotation scene;

[0006] Performing VR modeling annotation state vector mining on the earthquake damage annotation reference panoramic image to obtain a VR modeling annotation state vector of the earthquake damage annotation reference panoramic image, wherein the VR modeling annotation state vector is used to reflect a second VR modeling annotation state of the target earthquake damage annotation scene that is different from the first VR modeling annotation state;

[0007] Performing single-building panoramic image vector mining on the target single-building panoramic image to obtain the single-building panoramic image vector of the target single-building panoramic image;

[0008] Based on the VR modeling annotation state vector and the single building panoramic image vector, a target single building panoramic image in the second VR modeling annotation state is generated to update the VR modeling annotation state of the target single building panoramic image from the first VR modeling annotation state to the second VR modeling annotation state.

[0009] Optionally, the performing VR modeling annotation state vector mining on the earthquake damage annotated reference panoramic image to obtain the VR modeling annotation state vector of the earthquake damage annotated reference panoramic image includes:

[0010] Performing rendering state vector mining on the image rendering information of the earthquake damage annotated reference panoramic image to obtain a rendering state vector;

[0011] Performing VR modeling annotation state vector mining on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain a target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image;

[0012] A vector mapping operation is performed on the target VR modeling annotation state vector to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

[0013] Optionally, the rendering state vector includes a plurality of regional rendering state vectors, and the VR modeling annotation state vector mining is performed on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain a target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image, including:

[0014] For each of the regional rendering state vectors, VR modeling annotation state vector mining is performed on the regional rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain a local VR modeling annotation state vector;

[0015] A local VR modeling annotation state vector chain is constructed based on the multiple local VR modeling annotation state vectors, and the local VR modeling annotation state vector chain is used as the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

[0016] Optionally, the performing VR modeling annotation state vector mining on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain a target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image includes:

[0017] Performing a first downsampling process on the rendering state vector to obtain a downsampled rendering state vector;

[0018] Performing a second downsampling process on the structural state vector of the earthquake damage annotated reference panoramic image to obtain a downsampled structural state vector;

[0019] Combining the down-sampling rendering state vector and the down-sampling structure state vector to obtain a panoramic image combination state vector;

[0020] VR modeling annotation state vector mining is performed on the panoramic image combination state vector to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

[0021] Optionally, the first downsampling process includes an X-level downsampling process, and the second downsampling process includes a Y-level downsampling process;

[0022] Before combining the down-sampling rendering state vector and the down-sampling structure state vector to obtain a panoramic image combination state vector, the method further includes: obtaining a transitional down-sampling rendering state vector obtained by a u-th level down-sampling process in the first down-sampling process, and obtaining a transitional down-sampling structure state vector obtained by a v-th level down-sampling process in the second down-sampling process; combining the transitional down-sampling rendering state vector and the transitional down-sampling structure state vector to obtain a transitional panoramic image combination state vector;

[0023] The combining the down-sampled rendering state vector and the down-sampled structure state vector to obtain a panoramic image combination state vector comprises: combining the down-sampled rendering state vector, the down-sampled structure state vector, and the transition panoramic image combination state vector to obtain the panoramic image combination state vector;

[0024] Wherein, X, Y, u and v are all integers greater than 1, u is smaller than X, and v is smaller than Y.

[0025] Optionally, the performing VR modeling annotation state vector mining on the panoramic image combination state vector to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image includes:

[0026] Downsampling the panoramic image combination state vector to obtain a downsampled combination vector;

[0027] Determine first quantization knowledge and second quantization knowledge of the down-sampled combination vector, and determine global quantization knowledge of the first quantization knowledge and the second quantization knowledge;

[0028] VR modeling annotation state vector mining is performed on the global quantitative knowledge to obtain a target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

[0029] Optionally, performing a vector mapping operation on the target VR modeling annotation state vector to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image includes:

[0030] Performing feature optimization on the target VR modeling annotation state vector to obtain a VR modeling annotation state optimization vector;

[0031] Performing spatial feature mapping on the VR modeling annotation state optimization vector, to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

[0032] Optionally, the performing single-building panoramic image vector mining on the target single-building panoramic image to obtain the single-building panoramic image vector of the target single-building panoramic image includes:

[0033] Acquire image rendering information of the panoramic image of the target single building;

[0034] The rendering state vector of the image rendering information is mined to obtain a single building panoramic image vector of the target single building panoramic image.

[0035] Optionally, generating the target single building panoramic image in the second VR modeling annotation state according to the VR modeling annotation state vector and the single building panoramic image vector includes:

[0036] The VR modeling annotation state vector and the single building panoramic image vector are combined to obtain a first combined vector, and the first combined vector is downsampled to obtain a downsampled vector;

[0037] The down-sampling vector and the VR modeling annotation state vector are combined to obtain a second combination vector, and image recognition is performed on the second combination vector to obtain a panoramic image of the target single building in the second VR modeling annotation state.

[0038] Optionally, generating a panoramic image of a target single building in the second VR modeling annotation state according to the VR modeling annotation state vector and the single building panoramic image vector includes: generating a panoramic image of a target single building in the second VR modeling annotation state through a VR modeling annotation processing network according to the VR modeling annotation state vector and the single building panoramic image vector;

[0039] The VR modeling annotation processing network includes a P-level downsampling branch and a P-level image recognition branch. The VR modeling annotation processing network generates a target single building panoramic image in the second VR modeling annotation state according to the VR modeling annotation state vector and the single building panoramic image vector, including:

[0040] Determine the first-level downsampling vector generated by the first-level downsampling branch in the P-level downsampling branches according to the VR modeling annotation state vector and the single building panoramic image vector, and combine the VR modeling annotation state vector and the first-level downsampling vector to obtain a first-level combined vector;

[0041] Performing image recognition on the first-level combined vector through the first-level image recognition branch in the P-level image recognition branch to obtain a first-level image descriptor;

[0042] Obtaining a k-th downsampling vector generated by a k-th downsampling branch in the P-th downsampling branch, and combining the VR modeling annotation state vector, the k-th downsampling vector, and the k-1-th image descriptor to obtain a k-th combined vector;

[0043] Performing image recognition on the k-th level combined vector through the k-th level image recognition branch in the P-level image recognition branches to obtain a k-th level image descriptor, wherein P and k are integers greater than 1, and k is less than or equal to P;

[0044] The k is walked to obtain a P-th level image descriptor, and the P-th level image descriptor is used as the target single building panoramic image with the second VR modeling annotation state.

[0045] Optionally, determining the first-level downsampling vector generated by the first-level downsampling branch in the P-level downsampling branches according to the VR modeling annotation state vector and the single building panoramic image vector includes:

[0046] The VR modeling annotation state vector and the single building panoramic image vector are combined to obtain a P-th level combined vector, and the P-th level combined vector is downsampled by a P-th level downsampling branch in the P-th level downsampling branches to obtain a P-th level downsampling vector;

[0047] The VR modeling annotation state vector and the k+1th level downsampling vector are combined to obtain a kth level combination vector, and the kth level downsampling branch in the P-level downsampling branch is used to downsample the kth level combination vector to obtain a kth level downsampling vector;

[0048] The k is walked to obtain a first-level downsampling vector generated by a first-level downsampling branch in the P-level downsampling branches.

[0049] An embodiment of the present application also provides a computer device, including a processor, a memory and a bus connected to the processor; wherein the processor and the memory communicate with each other through the bus; the processor is used to call program instructions in the memory to execute the above-mentioned VR panoramic single building earthquake damage information collection method.

[0050] The embodiment of the present application also provides a computer-readable storage medium on which a program is stored. When the program is executed by a processor, the above-mentioned VR panoramic single-building earthquake damage information collection method is implemented.

[0051] Beneficial Effects

[0052] The embodiment of the present application provides a VR panoramic method for collecting earthquake damage information of single buildings, which can comprehensively collect data by acquiring a panoramic image of the target single building and a reference panoramic image with earthquake damage annotations. Vector mining of the earthquake damage annotated reference panoramic image obtains corresponding vectors, which can accurately reflect the special state of the earthquake damage scene. Vector mining of the panoramic image of the target single building can characterize its own characteristics in detail. Finally, a panoramic image of a new state is generated based on the two vectors, and an update from the initial state to the second state related to the earthquake damage is achieved, which helps to accurately display the condition of the building after it may suffer earthquake damage, and provides a comprehensive and accurate image basis for the evaluation of the seismic performance of the building and the analysis of the earthquake damage situation, thereby improving the accuracy and reliability of earthquake resistance research. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0054] Figure 1 This is a flow chart of a method for collecting earthquake damage information of a single building in a VR panoramic view provided in an embodiment of the present application.

[0055] Figure 2 A block diagram of a computer device provided in an embodiment of the present application.

[0056] icon:

[0057] 100-Computer equipment;

[0058] 101 - processor; 102 - memory; 103 - bus. DETAILED DESCRIPTION

[0059] The exemplary embodiments disclosed in the present application will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present application and to fully convey the scope of the present application to those skilled in the art.

[0060] In order to better understand the above technical scheme, the technical scheme of the present application is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present application and the specific features in the embodiments are detailed descriptions of the technical scheme of the present application, rather than limitations on the technical scheme of the present application. In the absence of conflict, the embodiments of the present application and the technical features in the embodiments can be combined with each other.

[0061] Figure 1 The flowchart of a method for collecting earthquake damage information of a single building in a VR panoramic view provided according to an embodiment of the present application is applied to a computer device and includes steps 101 to 104.

[0062] Step 101: Acquire a panoramic image of a target single building in a first VR modeling and annotation state, and acquire a earthquake damage annotation reference panoramic image of a target earthquake damage annotation scene.

[0063] Step 102: VR modeling annotation state vector mining is performed on the earthquake damage annotation reference panoramic image to obtain a VR modeling annotation state vector of the earthquake damage annotation reference panoramic image, wherein the VR modeling annotation state vector is used to reflect a second VR modeling annotation state of the target earthquake damage annotation scene that is different from the first VR modeling annotation state.

[0064] Step 103: mining the single-building panoramic image vector of the target single-building panoramic image to obtain the single-building panoramic image vector of the target single-building panoramic image.

[0065] Step 104: Generate a target single building panoramic image in the second VR modeling annotation state based on the VR modeling annotation state vector and the single building panoramic image vector, so as to update the VR modeling annotation state of the target single building panoramic image from the first VR modeling annotation state to the second VR modeling annotation state.

[0066] In earthquake work, the information collection of typical single buildings is very important. This application uses the information collection system as the execution body, and through a series of steps, realizes VR panoramic related operations of typical single buildings, thereby achieving all-round information collection and display of the building's seismic performance or earthquake damage.

[0067] First, in step 101, the information collection system needs to obtain a panoramic image of the target single building in the first VR modeling and annotation state, and obtain a reference panoramic image of the earthquake damage annotation of the target earthquake damage annotation scene. Obtaining a panoramic image of the target single building is the basis of the entire process. In actual operation, drones and professional panoramic cameras can be used to collect panoramic photos of the interior and exterior of the building. For example, the drone can fly around the building according to a preset flight path, and the lens angle of the professional panoramic camera can be set to 360 degrees in the horizontal direction and 180 degrees in the vertical direction to ensure that the appearance of the building can be fully covered. For the indoor part, the operator can hold a professional panoramic camera and shoot in a certain order in each room inside the building, such as from the bottom floor to the top floor, and each room starts from the corner and gradually covers the space of the entire room. After preliminary processing of these photographs, a panoramic image of the target single building can be obtained, and the image has the first VR modeling and annotation state, which may contain basic annotations such as the initial structure information and building material information of the building. At the same time, the reference panoramic image of the earthquake damage annotation of the target earthquake damage annotation scene also needs to be obtained. This earthquake damage annotated reference panoramic image can come from previous image records of earthquake damage to similar buildings, or representative earthquake damage scene images obtained from a database, which can provide a reference standard for subsequent analysis.

[0068] Then proceeding to step 102, the information collection system shall perform VR modeling annotation state vector mining on the earthquake damage annotation reference panoramic image to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image. This vector is used to reflect the second VR modeling annotation state of the target earthquake damage annotation scene, which is different from the first VR modeling annotation state. The so-called VR modeling annotation state vector mining is a complex analysis process. To take a specific example, there are earthquake damages such as wall cracks and structural deformations in the earthquake damage annotation reference panoramic image. During the mining process, the information collection system will perform quantitative analysis on these earthquake damage features in the image. For example, for wall cracks, quantification can be performed based on factors such as the length, width, and direction of the cracks. If the wall is regarded as a plane coordinate system, the length of the crack can be obtained by calculating the distance between the two ends of the crack in the coordinate system. For example, if the coordinates of one end of the crack are (10, 10) and the coordinates of the other end are (20, 10), the length is 10 units. The width of the crack can be estimated by comparing the pixels of the image. For example, after comparative analysis, it is found that the actual width of the crack width of 5 pixels is 0.5 cm. The direction of the crack can be obtained by calculating the angle between the crack and the coordinate axis of the coordinate system. For example, the angle with the x-axis is 0 degrees, indicating a horizontal direction. For structural deformation, the displacement of structural components such as beams and columns can be analyzed. For example, the coordinates of one end of the beam in the normal state are (30, 30), and the coordinates become (31, 30) after the earthquake damage, then the horizontal displacement is 1 unit. These quantified earthquake damage feature information constitute the VR modeling annotation state vector, which reflects the second VR modeling annotation state of the target earthquake damage annotation scene after the earthquake damage, which is significantly different from the previous first VR modeling annotation state.

[0069] Then, step 103 is performed, in which the information acquisition system mines the single-building panoramic image vector of the target single-building panoramic image to obtain the single-building panoramic image vector of the target single-building panoramic image. In this process, the information acquisition system will analyze and mine various features in the panoramic image of the target single-building. For example, for the structural features of the building, the size and position relationship of the structural components such as beams, columns, and walls will be analyzed. Taking the column as an example, the diameter, height, and distribution position of the column in the building can be determined by image analysis. For example, in the image, the diameter of the column is shown as a diameter occupying 10 pixels on the image. According to the conversion relationship between the previously calibrated pixels and the actual size (such as 10 pixels corresponding to 1 meter), the actual diameter of the column is 1 meter; the height of the column occupies 200 pixels from the bottom to the top of the image, and the actual height after conversion is 20 meters; the position of the column in the building can be determined by the relative coordinates with the building boundary, such as 10 meters from the left boundary of the building, 5 meters from the front boundary, etc. In addition to the structural features, the building material features will also be analyzed, for example, the color, texture and other features of the image can be used to determine whether the building material is concrete, brick or other materials. If the wall in the image is gray, rough and has regular block texture, it may be judged as a brick wall. These characteristic information about building structure, building materials, etc. are quantified to form a single building panoramic image vector.

[0070] Finally, in step 104, based on the VR modeling annotation state vector and the panoramic image vector of the single building, a panoramic image of the target single building in the second VR modeling annotation state is generated, so as to update the VR modeling annotation state of the panoramic image of the target single building from the first VR modeling annotation state to the second VR modeling annotation state. Specifically, the information acquisition system will fuse the earthquake damage feature information in the VR modeling annotation state vector obtained in step 102 with the initial building feature information in the panoramic image vector of the single building obtained in step 103. For example, the wall crack feature in the earthquake damage annotation reference panoramic image is fused to the wall in the panoramic image of the target single building. If the position coordinates of the wall in the panoramic image of the target single building in the panoramic image vector of the single building correspond to the wall position coordinates in the earthquake damage annotation reference panoramic image, then the same crack situation can be simulated on the wall in the panoramic image of the target single building according to the length, width, direction and other information of the wall crack in the earthquake damage annotation reference panoramic image. The same is true for structural deformation. According to the displacement of the structural components in the earthquake damage annotation reference panoramic image, the same displacement effect is simulated on the corresponding structural components in the panoramic image of the target single building. In this way, a panoramic image of the target single building in the second VR modeling annotation state can be generated. This image can intuitively show the situation of the target single building after it may suffer earthquake damage, and realize the update from the initial first VR modeling annotation state to the second VR modeling annotation state reflecting the earthquake damage situation. This updated VR panoramic image can provide all-round information for earthquake resistance performance assessment and earthquake damage analysis in earthquake work, making it convenient for relevant personnel to make more accurate judgments and studies on the earthquake resistance performance or earthquake damage of buildings.

[0071] It can be seen that the present application uses the information collection system as the execution body, and through the operations of the above steps 101 to 104, it can effectively process the VR panorama of a typical single building, and realize the all-round information collection and display of the seismic performance or earthquake damage of the building.

[0072] In an optional embodiment, the performing VR modeling annotation state vector mining on the earthquake damage labeled reference panoramic image to obtain the VR modeling annotation state vector of the earthquake damage labeled reference panoramic image includes: performing rendering state vector mining on image rendering information of the earthquake damage labeled reference panoramic image to obtain a rendering state vector; performing VR modeling annotation state vector mining on the rendering state vector and the structural state vector of the earthquake damage labeled reference panoramic image to obtain a target VR modeling annotation state vector of the earthquake damage labeled reference panoramic image; and performing a vector mapping operation on the target VR modeling annotation state vector to obtain the VR modeling annotation state vector of the earthquake damage labeled reference panoramic image.

[0073] In the present application, the process of mining the VR modeling and annotation state vector of the earthquake damage annotated reference panoramic image to obtain the corresponding vector includes multiple detailed operations. First, the rendering state vector of the image rendering information of the earthquake damage annotated reference panoramic image is mined to obtain the rendering state vector. Image rendering information includes many aspects, such as color, lighting, texture, etc. Taking color as an example, in an earthquake damage annotated reference panoramic image, there may be different color areas representing different degrees of earthquake damage or different parts of building structures. For example, in a certain area, red represents severely damaged parts, green represents undamaged parts, and blue represents slightly damaged parts. When mining the rendering state vector, these color areas will be quantitatively analyzed. For color quantification, a color space model, such as an RGB model, can be used. Red may be represented as (255, 0, 0) in the RGB model, green as (0, 255, 0), and blue as (0, 0, 255). At the same time, lighting conditions are also very important. Bright areas may indicate structural exposure or newly generated fracture surfaces, and dark areas may indicate depressions or obscured parts. By quantifying the distribution of light intensity in the image, for example, the light intensity quantification value of the brightest area in the image is 100 (which can be converted according to certain standards, such as proportional conversion based on a certain standard light intensity), and the darkest area is 10. In terms of texture, if the earthquake damage causes rough and uneven texture changes on the wall, this texture change can be quantified through image analysis technology. For example, the complexity quantification value of a normal wall texture is 10, and it becomes 20 after the earthquake damage. These quantitative information about color, light, texture, etc. are combined to form a rendering state vector.

[0074] Next, VR modeling annotation state vector mining is performed on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image. The structural state vector of the earthquake damage annotation reference panoramic image involves the performance of the building structure after the earthquake damage. For example, the beam structure of the building may be bent and deformed after the earthquake damage. The degree of bending of the beam can be determined through image analysis. For example, when the beam is not damaged, it is a straight line, and its straightness quantification value in the image is 100 (it can be calculated based on the degree of deviation between the pixel distribution of the beam in the image and the ideal straight line. The smaller the deviation, the higher the value). After the earthquake damage, it is bent, and the straightness quantification value becomes 80. The column may be tilted. The verticality quantification value of a normal column is 100 (based on the verticality to the ground, calculated by the deviation between the pixel distribution of the column in the image and the vertical direction), which becomes 90 after the earthquake damage. The verticality and integrity of the wall will also change. The verticality of the undamaged wall is quantified as 100, and the integrity is quantified as 100 (it can be quantified based on whether the wall has cracks, defects, etc., and 100 is complete). After the earthquake damage, the verticality becomes 95 and the integrity becomes 80. Combine the quantitative information such as color, lighting, and texture in the rendering state vector with the quantitative information of the beam, column, wall and other structures in the structural state vector to mine the VR modeling annotation state vector. For example, the red damaged area may correspond to the bending part of the beam and the crack part of the wall, and the area with high light intensity may be associated with the exposed part of the structural surface caused by structural deformation. Through this association and comprehensive analysis, the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image is obtained.

[0075] Finally, the target VR modeling annotation state vector is vector-mapped to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image. The vector mapping operation is a process of converting and adjusting the information in the target VR modeling annotation state vector according to certain rules. For example, in the target VR modeling annotation state vector, different earthquake damage features may use different quantitative units or representation methods. The vector mapping operation can unify these different representation methods for subsequent processing and analysis. For example, in the target VR modeling annotation state vector, the bending degree of the beam is expressed as a ratio of the straightness when it is not damaged, and the crack width of the wall is expressed as the actual pixel width. In the vector mapping operation, the bending degree of the beam can also be converted to a certain standard unit (such as the actual bending deformation in centimeters), and the crack width of the wall can be converted to a proportional value related to the wall size. In this way, the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image obtained after the vector mapping operation has a unified representation form and quantitative standard, which can more accurately reflect the various earthquake damage information in the earthquake damage annotation reference panoramic image, and provide a reliable basis for the subsequent combined processing with the panoramic image of the target single building.

[0076] Through such a detailed process of VR modeling and annotation state vector mining of the earthquake damage annotation reference panoramic image, the earthquake damage information in the earthquake damage annotation reference panoramic image can be extracted more accurately. In the rendering state vector mining, the quantitative analysis of image rendering information such as color, lighting and texture can fully grasp the earthquake damage situation from the visual presentation aspect. Combined with the structural state vector mining, the deformation, damage and other conditions of the building structure are integrated with the visual information, further enriching the connotation of the earthquake damage information. The final vector mapping operation ensures the unified representation form and quantification standard of the information, so that the obtained VR modeling and annotation state vector can accurately reflect the actual situation of the earthquake damage annotation reference panoramic image, thereby providing an accurate, comprehensive and standardized earthquake damage information basis for subsequent operations in the entire invention, which helps to more effectively collect and display all-round information on the seismic performance or earthquake damage of the target single building.

[0077] Further, the rendering state vector includes multiple regional rendering state vectors, and the VR modeling annotation state vector mining is performed on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image, including: for each of the regional rendering state vectors, the VR modeling annotation state vector mining is performed on the regional rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain a local VR modeling annotation state vector; based on the multiple local VR modeling annotation state vectors, a local VR modeling annotation state vector chain is constructed, and the local VR modeling annotation state vector chain is used as the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

[0078] Based on the previous work, for the case where the rendering state vector contains multiple regional rendering state vectors, there is a specific operation process in the process of VR modeling annotation state vector mining of the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain the target VR modeling annotation state vector.

[0079] First, for each region rendering state vector, the VR modeling annotation state vector mining with the structural state vector of the earthquake damage annotation reference panoramic image is carried out to obtain the local VR modeling annotation state vector. The regional rendering state vector is the quantification result of the rendering information of different regions of the earthquake damage annotation reference panoramic image. For example, in the earthquake damage annotation reference panoramic image, it can be divided according to different parts of the building or different functional areas, such as dividing the building into the ground floor, middle floor, and high floor areas, or into the living room area, bedroom area, corridor area, etc. according to the room function. For each area, its regional rendering state vector has a specific quantization value.

[0080] Taking the bottom layer as an example, the color quantification of this area, for example, there is a color gradient, from dark gray at the edge (which may be (50, 50, 50) under the RGB model) to light gray at the center (which may be (150, 150, 150) under the RGB model), which indicates that there may be a transition from less damage to more damage. In terms of lighting, the quantification value of the light intensity at the edge is 30, and that at the center is 50, indicating that the center may have structural deformation that leads to enhanced light reflection. In terms of texture, the quantification value of the texture complexity at the center is 15, and that at the edge is 10, which may imply that the earthquake damage at the center causes greater texture changes. At the same time, for the structural state vector of the bottom layer, the verticality of the column is quantified as 90 (with verticality as 100), the straightness of the beam is quantified as 85 (with straightness as 100), and the integrity of the wall is quantified as 80 (with completeness as 100). The regional rendering state vector and the structural state vector of the bottom layer are mined for VR modeling annotation state vectors. For example, darker areas may correspond to parts with better structural integrity, and parts with high light intensity may be associated with beams or columns with larger structural deformation. Through this association and analysis, the local VR modeling annotation state vector of the underlying area is obtained.

[0081] For the middle and high-rise areas or other functional areas, the above operation is also performed. The middle-rise area may have a relatively uniform color distribution, which is medium gray ((100, 100, 100) under the RGB model), an overall light intensity of 40, a texture complexity of 12, a column verticality of 92, a beam straightness of 90, and a wall integrity of 85 in the structural state vector. The high-rise area is lighter in color ((200, 200, 200) under the RGB model), with a light intensity of 60, a texture complexity of 18, and a column verticality of 95, a beam straightness of 93, and a wall integrity of 90 in the structural state vector. These areas are mined separately to obtain their respective local VR modeling and annotation state vectors.

[0082] Then, based on multiple local VR modeling annotation state vectors, a local VR modeling annotation state vector chain is constructed, and the local VR modeling annotation state vector chain is used as the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image. For example, the local VR modeling annotation state vector of the bottom layer area is used as the starting part of the chain, followed by the local VR modeling annotation state vector of the middle layer area, and the local VR modeling annotation state vector of the high layer area is placed at the end. This local VR modeling annotation state vector chain fully reflects the comprehensive situation of earthquake damage related information from the bottom layer to the high layer or from different functional areas of the entire earthquake damage annotation reference panoramic image. The target VR modeling annotation state vector constructed in this way not only contains the earthquake damage information of each area, but also reflects the relationship between the areas, such as the correlation and difference of earthquake damage between different floors or between different functional areas.

[0083] By dividing the rendering state vector into multiple regional rendering state vectors, mining each region with the structural state vector to obtain the local VR modeling annotation state vector, and then constructing a vector chain, the earthquake damage information in the earthquake damage annotation reference panoramic image can be mined more carefully and comprehensively. The precise association between the rendering information such as color, lighting, texture and structural state information in different regions can accurately reflect the differences and connections between earthquake damage conditions within and between regions. This target VR modeling annotation state vector can provide richer and more accurate earthquake damage reference information for the subsequent processing of the panoramic image of the target single building, which helps to improve the accuracy of the collection and display of all-round information on the earthquake resistance performance or earthquake damage of the building.

[0084] In another technical solution, the VR modeling annotation state vector mining is performed on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image, including: performing a first downsampling process on the rendering state vector to obtain a downsampled rendering state vector; performing a second downsampling process on the structural state vector of the earthquake damage annotation reference panoramic image to obtain a downsampled structural state vector; combining the downsampled rendering state vector and the downsampled structural state vector to obtain a panoramic image combination state vector; and performing VR modeling annotation state vector mining on the panoramic image combination state vector to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

[0085] On the basis of the technical solution, the first downsampling process includes an X-level downsampling process, and the second downsampling process includes a Y-level downsampling process; before the downsampling rendering state vector and the downsampling structure state vector are combined to obtain the panoramic image combination state vector, the method further includes: obtaining a transition downsampling rendering state vector obtained by the u-th level downsampling process in the first downsampling process, and obtaining a transition downsampling structure state vector obtained by the v-th level downsampling process in the second downsampling process; combining the transition downsampling rendering state vector and the transition downsampling structure state vector to obtain a transition panoramic image combination state vector. Based on this, the downsampling rendering state vector and the downsampling structure state vector are combined to obtain the panoramic image combination state vector, including: combining the downsampling rendering state vector, the downsampling structure state vector, and the transition panoramic image combination state vector to obtain the panoramic image combination state vector; wherein X, Y, u, and v are all integers greater than 1, u is less than X, and v is less than Y.

[0086] On the basis of this technical solution, the VR modeling and annotation state vector mining of the panoramic image combination state vector is performed to obtain the target VR modeling and annotation state vector of the earthquake damage annotation reference panoramic image, including: downsampling the panoramic image combination state vector to obtain a downsampled combination vector; determining the first quantitative knowledge and the second quantitative knowledge of the downsampled combination vector, and determining the global quantitative knowledge of the first quantitative knowledge and the second quantitative knowledge; and VR modeling and annotation state vector mining of the global quantitative knowledge to obtain the target VR modeling and annotation state vector of the earthquake damage annotation reference panoramic image.

[0087] In this technical solution, the process of performing VR modeling annotation state vector mining on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image is relatively complex and includes multiple sub-processes.

[0088] First, the rendering state vector is subjected to the first downsampling process to obtain a downsampled rendering state vector. Downsampling is a data processing method that aims to reduce the amount of data for subsequent operations. Here, the first downsampling process includes X-level downsampling processes. For example, the rendering state vector initially has 1000 quantized values ​​in the color dimension to represent the distribution of colors in the image. After the first level of downsampling, it may be reduced to 500 quantized values. After several levels of downsampling, the final downsampled rendering state vector may have only 100 quantized values. This is a gradual simplification process of color information at different resolutions. Similarly, other information contained in the rendering state vector, such as lighting and texture, is processed in a similar manner.

[0089] Next, the structural state vector of the earthquake damage annotated reference panoramic image is subjected to a second downsampling process to obtain a downsampled structural state vector. For example, the verticality of the column in the structural state vector is initially quantized at a height interval of 1 cm, with a total of 2,000 quantized values. After the second downsampling process (Y-level downsampling process), for example, after the first level of downsampling, it is quantized at a height interval of 5 cm, and the quantized values ​​are reduced to 400. The final downsampled structural state vector may have only 100 quantized values, which effectively reduces the amount of data in the structural state vector while retaining key structural state information.

[0090] Before combining the downsampled rendering state vector and the downsampled structural state vector to obtain the panoramic image combination state vector, it is also necessary to obtain the transitional downsampled rendering state vector obtained by the u-th level downsampled processing in the first downsampled processing, and obtain the transitional downsampled structural state vector obtained by the v-th level downsampled processing in the second downsampled processing, and then combine these two transitional downsampled state vectors to obtain the transitional panoramic image combination state vector. For example, in the first downsampled processing, for example, X=5, u=2, and in the second downsampled processing, for example, Y=4, v=2. At this stage, the transitional downsampled rendering state vector obtained by the second level downsampled processing may have 300 quantized values ​​in the color dimension, and the verticality of the column in the transitional downsampled structural state vector has 200 quantized values, which are combined to form the transitional panoramic image combination state vector.

[0091] Then, the downsampled rendering state vector, the downsampled structural state vector and the transitional panoramic image combination state vector are combined to obtain a panoramic image combination state vector. This panoramic image combination state vector contains the rendering state information and structural state information after downsampling and combination at different stages, and is a comprehensive state representation.

[0092] Afterwards, the panoramic image combination state vector is downsampled to obtain a downsampled combination vector. For example, the panoramic image combination state vector may have 500 quantized values, and after downsampling, the downsampled combination vector has only 100 quantized values. Then, the first quantized knowledge and the second quantized knowledge of the downsampled combination vector are determined, and the global quantized knowledge of the first quantized knowledge and the second quantized knowledge is determined. The first quantized knowledge and the second quantized knowledge can be quantized features of different aspects in the downsampled combination vector. For example, the first quantized knowledge can be quantized features related to color and illumination, and the second quantized knowledge can be quantized features related to structural deformation. Determining the global quantized knowledge is to comprehensively consider the quantized features of these two aspects, for example, to obtain a comprehensive quantized knowledge through weighted averaging or other fusion methods. This global quantized knowledge comprehensively reflects various information in the downsampled combination vector. Finally, the global quantized knowledge is mined for VR modeling and annotation state vectors to obtain the target VR modeling and annotation state vector of the earthquake damage annotation reference panoramic image.

[0093] Downsampling can effectively reduce the amount of data and improve computing efficiency without losing key information. The transition downsampling state vectors are obtained at different stages and combined so that the panoramic image combination state vector contains more levels of information. After determining the multiple quantitative knowledge of the downsampling combination vectors and fusing them into global quantitative knowledge, the target VR modeling annotation state vector is obtained through mining, which can more comprehensively and accurately reflect the earthquake damage information in the earthquake damage annotation reference panoramic image, providing a more accurate basis for subsequent operations, and helping to improve the accuracy and efficiency of the entire information collection and display of the earthquake resistance performance or earthquake damage of the building.

[0094] In a preferred embodiment, performing a vector mapping operation on the target VR modeling annotation state vector to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image includes: performing feature optimization on the target VR modeling annotation state vector to obtain a VR modeling annotation state optimization vector; performing spatial feature mapping on the VR modeling annotation state optimization vector to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

[0095] In this preferred embodiment, there is a specific operation flow for the process of performing a vector mapping operation on the target VR modeling annotation state vector to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

[0096] First, the target VR modeling annotation state vector is feature optimized to obtain the VR modeling annotation state optimization vector. The target VR modeling annotation state vector contains various information about the earthquake damage annotation reference panoramic image obtained from a series of previous operations, such as vectors that may be mined and integrated from multiple aspects such as color, lighting, texture, and structural state in previous operations. When performing feature optimization, the features in these vectors will be screened, adjusted, and enhanced.

[0097] Taking color-related features as an example, there may be multiple quantized values ​​representing color information in the target VR modeling annotation state vector. If there are certain errors or redundancies in the color information in some areas, they will be corrected and simplified during the feature optimization process. For example, in a local area, due to some interference in the previous acquisition or processing process, the color quantization value may fluctuate. For example, in an area representing the color of the wall, the original color quantization value has a certain degree of discreteness under the RGB model, such as some pixels are represented as (120, 120, 120), and some are (130, 125, 120). During the feature optimization process, it will be optimized to a more stable value, such as (125, 123, 120) based on the surrounding pixels and the overall color trend.

[0098] The same is true for features related to the structural state. For example, in the target VR modeling annotation state vector, the quantified value of the bending degree of the beam has some inaccuracy. The original quantified value may not reflect the actual bending condition well due to problems with the measurement range or calculation method. During feature optimization, this quantified value will be adjusted based on more image information and relevant structural knowledge. For example, the original quantified value of the bending degree of the beam was expressed as 80 (100 according to some standards). After feature optimization, it was adjusted to 83 based on a more comprehensive analysis. This adjusted value can more accurately reflect the actual bending state of the beam.

[0099] Similar operations are performed on lighting and texture. By optimizing various features in these target VR modeling annotation state vectors, removing errors and redundancies, and enhancing the expressiveness of key features, the VR modeling annotation state optimization vector is obtained. Compared with the target VR modeling annotation state vector, this optimized vector is more accurate and refined in representing various features of the earthquake damage annotation reference panoramic image.

[0100] Next, the VR modeling annotation state optimization vector is subjected to spatial feature mapping to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image. Spatial feature mapping is an operation that remaps the features in the optimized vector according to the spatial relationship. In the earthquake damage annotation reference panoramic image, each feature has a specific distribution and mutual relationship in space.

[0101] For example, for different structural parts of a building, such as columns, beams, and walls, they are connected and supported in space. In the VR modeling annotation state optimization vector, the features of these structural parts are quantified and represented separately. In the spatial feature mapping process, the features of these structural parts will be recombined and mapped according to their spatial relationships in the actual building. For example, at the connection between the column and the beam, the respective features of the column and the beam in the optimization vector need to be integrated according to their spatial connection relationship. If the verticality quantization value of the column is 90, and the support angle quantization value of the beam at the connection is 85 (quantization values ​​based on vertical or horizontal standards), in the spatial feature mapping process, these two quantization values ​​will be remapped according to their spatial association to reflect the overall structural state at this connection.

[0102] The same is true for features such as color, lighting, and texture in different areas. For example, in a corner area of ​​a building, color changes may be spatially correlated with the intensity of light and the texture changes of the wall. In the spatial feature mapping process, these relatively independent color, lighting, and texture features in the VR modeling annotation state optimization vector will be mapped according to their spatial relationship in this corner area, so that the final VR modeling annotation state vector of the earthquake damage annotation reference panoramic image can accurately reflect the spatial relationship between each feature, thereby more comprehensively and realistically reflecting the earthquake damage situation in the earthquake damage annotation reference panoramic image.

[0103] By optimizing the features of the target VR modeling annotation state vector, the accuracy and refinement of the feature representation in the vector can be improved, errors and redundant information can be removed, and subsequent operations can be based on more reliable data. The spatial feature mapping operation remaps the features in the optimized vector according to the spatial relationship, which helps to integrate the originally relatively independent features according to the actual spatial relationship, thereby more comprehensively and accurately reflecting the earthquake damage in the earthquake damage annotation reference panoramic image, providing more accurate earthquake damage reference information for the related operations with the panoramic image of the target single building in the entire invention, and improving the quality of the collection and display of information on the earthquake resistance performance or earthquake damage of the building.

[0104] In an exemplary embodiment, the performing single-building panoramic image vector mining on the target single-building panoramic image to obtain the single-building panoramic image vector of the target single-building panoramic image includes: acquiring image rendering information of the target single-building panoramic image; performing rendering state vector mining on the image rendering information to obtain the single-building panoramic image vector of the target single-building panoramic image.

[0105] In this exemplary embodiment, the process of mining the single-building panoramic image vector of the target single-building panoramic image to obtain the single-building panoramic image vector has a clear operation flow.

[0106] First, the image rendering information of the panoramic image of the target single building is obtained. The panoramic image of the target single building is an image that fully displays the single building, and its image rendering information includes multiple aspects. For example, in terms of color, different structural parts of the building may have different colors, the wall may be grayish white, the roof may be red, etc. The specific values ​​of these colors have specific representations under a certain color model. For example, the grayish white wall may be represented as (200, 200, 200) under the RGB color model, and the red roof may be (255, 0, 0). Lighting information is also an important part of image rendering information. Walls with different orientations may have different light intensities. The light intensity on the sunny side may be higher, with a quantization value of 80 (measured by a certain standard maximum light intensity of 100), and the light intensity on the shady side may be lower, with a quantization value of 30. Texture cannot be ignored either. If the wall is a brick structure, it will have the texture characteristics of bricks, and the complexity of its texture can be quantified, for example, the quantization value is 10 (the quantization standard can be determined based on factors such as the repetition pattern and fineness of the texture).

[0107] Next, the rendering state vector of the image rendering information is mined to obtain the single building panoramic image vector of the target single building panoramic image. When mining the rendering state vector of the color, in addition to recording the color values ​​of each part, the distribution law of the color is also analyzed. For example, in the entire panoramic image, the proportion of different color areas, the gray-white wall area may account for 60% of the total image area, and the red roof area accounts for 20%. These color values, distribution and other information together constitute the color part in the rendering state vector. For lighting information, in addition to the intensity of light, the uniformity of light is also considered. For example, the standard deviation of the light intensity in the entire image is 10 (calculated based on the light intensity of each point in the image). This standard deviation reflects the uniformity of the light. High light intensity and small standard deviation indicate uniform and bright light. These information such as light intensity and uniformity constitute the part of light in the rendering state vector. In terms of texture, in addition to texture complexity, texture direction and other characteristics are also analyzed. For example, the texture direction of brick walls is mainly horizontal. This direction information and complexity information together constitute the part of texture in the rendering state vector. By combining the color, lighting, texture and other aspects in the rendering state vector, we can get the single building panoramic image vector of the target single building panoramic image. This vector comprehensively reflects the image rendering information of the target single building panoramic image, and provides a basis for subsequent combination or analysis with other vectors.

[0108] By acquiring the image rendering information of the panoramic image of the target single building and mining the rendering state vector to obtain the panoramic image vector of the single building, the panoramic image of the target single building can be analyzed comprehensively and meticulously from the perspective of image rendering. The quantification and integration of color value, distribution, light intensity, uniformity, texture complexity, direction and other aspects enable the obtained panoramic image vector of the single building to accurately reflect the characteristics of the panoramic image of the target single building, providing an accurate data basis for the processing of the panoramic image of the target single building involved in the entire invention, and helping to improve the accuracy of the collection and display of information on the seismic performance or earthquake damage of the building.

[0109] In another exemplary embodiment, generating a target single building panoramic image in the second VR modeling annotation state based on the VR modeling annotation state vector and the single building panoramic image vector includes: combining the VR modeling annotation state vector and the single building panoramic image vector to obtain a first combination vector, and downsampling the first combination vector to obtain a downsampled vector; combining the downsampled vector and the VR modeling annotation state vector to obtain a second combination vector, and performing image recognition on the second combination vector to obtain a target single building panoramic image in the second VR modeling annotation state.

[0110] In this exemplary embodiment, the process of generating a target single building panoramic image in a second VR modeling annotation state according to the VR modeling annotation state vector and the single building panoramic image vector is relatively complex and involves a series of specific operations.

[0111] First, the VR modeling annotation state vector and the single building panoramic image vector are combined to obtain the first combination vector. The VR modeling annotation state vector contains various quantitative information about the earthquake damage obtained by a series of processing of the earthquake damage annotation reference panoramic image, such as structural deformation information after the earthquake damage, information on the degree of damage in different areas, etc. The single building panoramic image vector is information about the building's own characteristics mined from the target single building panoramic image, such as quantitative information on the building's color, lighting, texture, etc. When these two vectors are combined, the different information they contain is integrated together according to certain rules. For example, the structural deformation degree information of a certain area in the VR modeling annotation state vector is combined with the color information of the corresponding area in the single building panoramic image vector. For example, the structural deformation quantification value of a certain wall area in the VR modeling annotation state vector is 80 (normal state is 100), and the color of the wall area in the single building panoramic image vector is (200, 200, 200) under the RGB model. These information are combined into the first combination vector.

[0112] Then, the first combination vector is downsampled to obtain a downsampled vector. The purpose of downsampling is to reduce the amount of data for subsequent operations. For example, the first combination vector may contain a large amount of quantization information. In terms of color, there may be color quantization values ​​corresponding to thousands of pixels, and there may also be a large number of detail quantization values ​​in terms of structural deformation. After downsampling, the color quantization values ​​may be simplified according to a certain area or ratio. For example, each small area originally has an independent color quantization value. After downsampling, it is averaged and quantized according to a larger area, reducing from thousands of quantization values ​​to hundreds. The quantization values ​​in terms of structural deformation are also similarly simplified, from many detail quantization values ​​to quantization values ​​of some key positions or areas, thereby obtaining a downsampled vector.

[0113] Then, the downsampled vector and the VR modeling annotation state vector are combined to obtain a second combined vector. This step again integrates the simplified information in the downsampled vector with the information in the VR modeling annotation state vector. The color simplified quantization value in the downsampled vector is combined again with the structural deformation information in the VR modeling annotation state vector. For example, if the average color quantization value of a certain area in the downsampled vector is (180, 180, 180), the structural deformation quantization value of the area in the VR modeling annotation state vector is adjusted to 70, and this information is combined into the second combined vector.

[0114] Finally, the second combination vector is subjected to image recognition to obtain a panoramic image of the target single building in the second VR modeling and annotation state. The image recognition process is to construct a new state of the panoramic image of the target single building based on various quantitative information in the second combination vector. In this process, the color distribution in the image is determined according to the color quantization value in the second combination vector, such as setting the color of the corresponding area according to the previous (180, 180, 180). The morphology of the building structure in the image is adjusted according to the structural deformation quantization value. For example, in the area where the structural deformation quantization value is 70, the corresponding wall or beam column and other structures are drawn in the image according to certain deformation rules, thereby constructing a panoramic image of the target single building in the second VR modeling and annotation state reflecting the earthquake damage.

[0115] By combining and downsampling the VR modeling annotation state vector and the single building panoramic image vector multiple times, the information is gradually integrated and simplified, so that the panoramic image of the target single building in the second VR modeling annotation state can be accurately constructed based on this information. Downsampling reduces the amount of data without losing key information and improves processing efficiency. The new panoramic image constructed by image recognition can intuitively show the state of the target single building after it may suffer earthquake damage, providing an effective image basis for the analysis and research of the seismic performance or earthquake damage of the building.

[0116] In another exemplary embodiment, the method of generating a target single building panoramic image in the second VR modeling annotation state based on the VR modeling annotation state vector and the single building panoramic image vector comprises: generating a target single building panoramic image in the second VR modeling annotation state through a VR modeling annotation processing network based on the VR modeling annotation state vector and the single building panoramic image vector; the VR modeling annotation processing network comprises a P-level downsampling branch and a P-level image recognition branch, and the method of generating a target single building panoramic image in the second VR modeling annotation state through a VR modeling annotation processing network based on the VR modeling annotation state vector and the single building panoramic image vector comprises: determining a first-level downsampling vector generated by a first-level downsampling branch in the P-level downsampling branch based on the VR modeling annotation state vector and the single building panoramic image vector, and The VR modeling annotation state vector and the first-level downsampling vector are combined to obtain a first-level combination vector; the first-level image recognition branch in the P-level image recognition branch is used to perform image recognition on the first-level combination vector to obtain a first-level image descriptor; the k-th level downsampling vector generated by the k-th level downsampling branch in the P-level downsampling branch is obtained, and the VR modeling annotation state vector, the k-th level downsampling vector, and the k-1-th level image descriptor are combined to obtain a k-th level combination vector; the k-th level image recognition branch in the P-level image recognition branch is used to perform image recognition on the k-th level combination vector to obtain a k-th level image descriptor, wherein P and k are integers greater than 1, and k is less than or equal to P; k is walked to obtain a P-th level image descriptor, and the P-th level image descriptor is used as a panoramic image of a target single building with the second VR modeling annotation state.

[0117] In the next step, the first-level downsampling vector generated by the first-level downsampling branch in the P-level downsampling branches is determined based on the VR modeling annotation state vector and the single building panoramic image vector, including: combining the VR modeling annotation state vector and the single building panoramic image vector to obtain a P-th level combination vector, and downsampling the P-th level combination vector through the P-th level downsampling branch in the P-level downsampling branch to obtain a P-th level downsampling vector; combining the VR modeling annotation state vector and the k+1-th level downsampling vector to obtain a k-th level combination vector, and downsampling the k-th level combination vector through the k-th level downsampling branch in the P-level downsampling branch to obtain a k-th level downsampling vector; walking the k to obtain the first-level downsampling vector generated by the first-level downsampling branch in the P-level downsampling branch.

[0118] In this yet another exemplary embodiment, generating a target single building panoramic image in a second VR modeling annotation state based on a VR modeling annotation state vector and a single building panoramic image vector is achieved through a VR modeling annotation processing network, and this process includes a series of complex and orderly operations.

[0119] First, the VR modeling annotation processing network includes a P-level downsampling branch and a P-level image recognition branch. According to the VR modeling annotation state vector and the single building panoramic image vector, the first-level downsampling vector generated by the first-level downsampling branch in the P-level downsampling branch should be determined. In this process, the VR modeling annotation state vector and the single building panoramic image vector are first combined to obtain the P-th level combination vector. For example, the VR modeling annotation state vector contains the structural deformation quantification information in the earthquake damage annotation reference panoramic image, such as the quantification value of the deformation of a column structure is 80 (with the normal structure as 100), and the single building panoramic image vector contains the color quantification information of the target single building panoramic image, such as the color of a wall is (200, 200, 200) under the RGB model. These different types of quantitative information are combined into the P-th level combination vector. Then, the P-th level combination vector is downsampled by the P-th level downsampling branch in the P-level downsampling branch to obtain the P-th level downsampling vector. The downsampling process will simplify the quantization information in the combined vector. For example, in terms of color quantization information, each pixel or small area was originally quantized in detail, but after downsampling, it may be averaged and quantized according to a larger area. The structural deformation quantization information will also be similarly simplified from multiple detail quantization values ​​to quantization values ​​of key positions or areas.

[0120] Next, the VR modeling annotation state vector and the k+1th level downsampling vector are combined to obtain the kth level combination vector, and the kth level combination vector is downsampled through the kth level downsampling branch in the P-level downsampling branch to obtain the kth level downsampling vector. Here, a specific numerical example is used to illustrate. For example, at a certain stage, the illumination quantization value of a certain area in the k+1th level downsampling vector is 60 (based on a certain illumination intensity standard of 100), and the structural association quantization value of the area in the VR modeling annotation state vector is 70. They are combined into the kth level combination vector, and then downsampled to obtain the kth level downsampling vector. By walking k, the calculation is gradually performed from a larger k value to 1, and finally the first level downsampling vector generated by the first level downsampling branch in the P-level downsampling branch is obtained.

[0121] After obtaining the first-level downsampling vector, the VR modeling annotation state vector and the first-level downsampling vector are combined to obtain the first-level combined vector. Then, the first-level image recognition branch in the P-level image recognition branch performs image recognition on the first-level combined vector to obtain the first-level image descriptor. The image recognition process constructs a preliminary image description based on the quantitative information in the combined vector, such as determining the color distribution area in the image based on the color quantization value, and determining the preliminary morphology of the structure in the image based on the structural deformation quantization value.

[0122] After that, obtain the k-th downsampling vector generated by the k-th downsampling branch in the P-level downsampling branch, and combine the VR modeling annotation state vector, the k-th downsampling vector, and the k-1-th image descriptor to obtain the k-th combined vector. For example, the k-th downsampling vector contains a texture quantization value of 10 (based on a certain texture complexity standard) in a certain area after downsampling, and the related structure quantization value of the area in the VR modeling annotation state vector is 85. The k-1-th image descriptor provides some structure and color information obtained by the previous image recognition, and they are combined to form the k-th combined vector. Then, the k-th combined vector is image recognized by the k-th image recognition branch in the P-level image recognition branch to obtain the k-th image descriptor.

[0123] As k gradually increases, the above combination and image recognition process is repeated, k is wandered, and finally the P-th level image descriptor is obtained, and the P-th level image descriptor is used as the target single building panoramic image with the second VR modeling annotation state. In this process, each level of downsampling vectors and image descriptors are constantly improving and refining the construction of the target single building panoramic image, from the initial combination of VR modeling annotation state vectors and single building panoramic image vectors, through the processing of multiple layers of downsampling branches and image recognition branches, a target single building panoramic image that can accurately reflect the second VR modeling annotation state is gradually constructed.

[0124] Through the collaborative processing of the P-level downsampling branch and the P-level image recognition branch of the VR modeling and annotation processing network, the panoramic image of the target single building in the second VR modeling and annotation state can be constructed step by step and meticulously based on the VR modeling and annotation state vector and the panoramic image vector of the single building. The downsampling branch reduces the amount of data by gradually simplifying the quantitative information in the vector, improves the processing efficiency, and retains key information. The image recognition branch constructs an image descriptor based on the combined vectors of each level, gradually improves the image content, and the final panoramic image can accurately show the situation of the target single building in a specific state, providing an accurate image basis for the analysis of the seismic performance or earthquake damage of the building.

[0125] In summary, by acquiring panoramic images of target single buildings and panoramic images of earthquake damage annotated reference images, data can be collected comprehensively. Vector mining of the panoramic images of earthquake damage annotated reference images can obtain corresponding vectors, which can accurately reflect the special state of the earthquake damage scene. Vector mining of the panoramic images of target single buildings can characterize their own characteristics in detail. Finally, a panoramic image of a new state is generated based on the two vectors, and the update from the initial state to the second state related to earthquake damage is realized. This helps to accurately display the condition of the building after it may suffer earthquake damage, and provides a comprehensive and accurate image basis for the evaluation of the seismic performance of the building and the analysis of the earthquake damage, thereby improving the accuracy and reliability of earthquake resistance research.

[0126] An embodiment of the present application provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the method for collecting earthquake damage information of a single building in the VR panorama.

[0127] An embodiment of the present application provides a processor, which is used to run a program, wherein the method for collecting earthquake damage information of a single building in a VR panorama is executed when the program is run.

[0128] In the present application embodiment, Figure 2 As shown, the computer device 100 includes at least one processor 101, and at least one memory 102 and a bus 103 connected to the processor 101; wherein the processor 101 and the memory 102 communicate with each other via the bus 103; the processor 101 is used to call the program instructions in the memory 102 to execute the above-mentioned VR panoramic single building earthquake damage information collection method.

[0129] The present application is described with reference to the flowcharts and / or block diagrams of the methods, computer devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0130] In a typical configuration, a computer device includes one or more processors (CPU), a memory, and a bus. A computer device may also include an input / output interface, a network interface, and the like.

[0131] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip. The memory is an example of a computer-readable medium.

[0132] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage computer readable storage media or any other non-transmission media that can be used to store information that can be accessed by a computer device. As defined in this article, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0133] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or computer-readable storage medium that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements inherent to such process, method, commodity or computer-readable storage medium. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the presence of other identical elements in the process, method, commodity or computer-readable storage medium that includes the elements.

[0134] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0135] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A VR panoramic single building earthquake damage information collection method, characterized in that: The method is applied to an information collection system, and the method comprises: Acquire a panoramic image of a target single building in a first VR modeling and annotation state, and acquire a reference panoramic image of earthquake damage annotation of a target earthquake damage annotation scene; Performing VR modeling annotation state vector mining on the earthquake damage annotation reference panoramic image to obtain a VR modeling annotation state vector of the earthquake damage annotation reference panoramic image, wherein the VR modeling annotation state vector is used to reflect a second VR modeling annotation state of the target earthquake damage annotation scene that is different from the first VR modeling annotation state; Performing single-building panoramic image vector mining on the target single-building panoramic image to obtain the single-building panoramic image vector of the target single-building panoramic image; Generate a target single building panoramic image in a second VR modeling annotation state according to the VR modeling annotation state vector and the single building panoramic image vector, so as to update the VR modeling annotation state of the target single building panoramic image from the first VR modeling annotation state to the second VR modeling annotation state; The step of mining the VR modeling annotation state vector of the earthquake damage annotated reference panoramic image to obtain the VR modeling annotation state vector of the earthquake damage annotated reference panoramic image includes: Performing rendering state vector mining on the image rendering information of the earthquake damage annotated reference panoramic image to obtain a rendering state vector; Performing VR modeling annotation state vector mining on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain a target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image; Performing a vector mapping operation on the target VR modeling annotation state vector to obtain a VR modeling annotation state vector of the earthquake damage annotation reference panoramic image; The step of generating a panoramic image of a target single building in the second VR modeling annotation state according to the VR modeling annotation state vector and the single building panoramic image vector comprises: The VR modeling annotation state vector and the single building panoramic image vector are combined to obtain a first combined vector, and the first combined vector is downsampled to obtain a downsampled vector; The down-sampling vector and the VR modeling annotation state vector are combined to obtain a second combination vector, and image recognition is performed on the second combination vector to obtain a panoramic image of the target single building in the second VR modeling annotation state.

2. The method according to claim 1, characterized in that The rendering state vector includes a plurality of regional rendering state vectors, and the VR modeling annotation state vector mining is performed on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image, including: For each of the regional rendering state vectors, VR modeling annotation state vector mining is performed on the regional rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain a local VR modeling annotation state vector; A local VR modeling annotation state vector chain is constructed based on the multiple local VR modeling annotation state vectors, and the local VR modeling annotation state vector chain is used as the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

3. The method according to claim 1, characterized in that The performing VR modeling annotation state vector mining on the rendering state vector and the structural state vector of the earthquake damage annotation reference panoramic image to obtain a target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image includes: Performing a first downsampling process on the rendering state vector to obtain a downsampled rendering state vector; Performing a second downsampling process on the structural state vector of the earthquake damage annotated reference panoramic image to obtain a downsampled structural state vector; Combining the down-sampling rendering state vector and the down-sampling structure state vector to obtain a panoramic image combination state vector; VR modeling annotation state vector mining is performed on the panoramic image combination state vector to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

4. The method according to claim 3, characterized in that The first downsampling process includes an X-level downsampling process, and the second downsampling process includes a Y-level downsampling process; Before combining the down-sampling rendering state vector and the down-sampling structure state vector to obtain a panoramic image combination state vector, the method further includes: obtaining a transitional down-sampling rendering state vector obtained by a u-th level down-sampling process in the first down-sampling process, and obtaining a transitional down-sampling structure state vector obtained by a v-th level down-sampling process in the second down-sampling process; combining the transitional down-sampling rendering state vector and the transitional down-sampling structure state vector to obtain a transitional panoramic image combination state vector; The combining the downsampled rendering state vector and the downsampled structure state vector to obtain a panoramic image combination state vector comprises: combining the downsampled rendering state vector, the downsampled structure state vector, and the transition panoramic image combination state vector to obtain the panoramic image combination state vector; Wherein, X, Y, u and v are all integers greater than 1, u is smaller than X, and v is smaller than Y.

5. The method according to claim 3, characterized in that The VR modeling annotation state vector mining of the panoramic image combination state vector to obtain the target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image includes: Downsampling the panoramic image combination state vector to obtain a downsampled combination vector; Determine first quantization knowledge and second quantization knowledge of the down-sampled combination vector, and determine global quantization knowledge of the first quantization knowledge and the second quantization knowledge; VR modeling annotation state vector mining is performed on the global quantitative knowledge to obtain a target VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

6. The method according to claim 1, characterized in that The performing a vector mapping operation on the target VR modeling annotation state vector to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image includes: Performing feature optimization on the target VR modeling annotation state vector to obtain a VR modeling annotation state optimization vector; Performing spatial feature mapping on the VR modeling annotation state optimization vector, to obtain the VR modeling annotation state vector of the earthquake damage annotation reference panoramic image.

7. The method according to claim 1, characterized in that The step of mining the single building panoramic image vector of the target single building panoramic image to obtain the single building panoramic image vector of the target single building panoramic image includes: Acquire image rendering information of the panoramic image of the target single building; The rendering state vector of the image rendering information is mined to obtain a single building panoramic image vector of the target single building panoramic image.

8. The method according to claim 1, characterized in that The step of generating a panoramic image of a target single building in the second VR modeling annotation state according to the VR modeling annotation state vector and the single building panoramic image vector comprises: generating a panoramic image of a target single building in the second VR modeling annotation state through a VR modeling annotation processing network according to the VR modeling annotation state vector and the single building panoramic image vector; The VR modeling annotation processing network includes a P-level downsampling branch and a P-level image recognition branch. The VR modeling annotation processing network generates a target single building panoramic image in the second VR modeling annotation state according to the VR modeling annotation state vector and the single building panoramic image vector, including: Determine the first-level downsampling vector generated by the first-level downsampling branch in the P-level downsampling branches according to the VR modeling annotation state vector and the single building panoramic image vector, and combine the VR modeling annotation state vector and the first-level downsampling vector to obtain a first-level combined vector; Performing image recognition on the first-level combined vector through the first-level image recognition branch in the P-level image recognition branch to obtain a first-level image descriptor; Obtaining a k-th downsampling vector generated by a k-th downsampling branch in the P-th downsampling branch, and combining the VR modeling annotation state vector, the k-th downsampling vector, and the k-1-th image descriptor to obtain a k-th combined vector; Performing image recognition on the k-th level combined vector through the k-th level image recognition branch in the P-level image recognition branches to obtain a k-th level image descriptor, wherein P and k are integers greater than 1, and k is less than or equal to P; Walk the k to obtain a P-th level image descriptor, and use the P-th level image descriptor as a panoramic image of a target single building with the second VR modeling annotation state; The step of determining the first-level downsampling vector generated by the first-level downsampling branch in the P-level downsampling branches according to the VR modeling annotation state vector and the single building panoramic image vector comprises: The VR modeling annotation state vector and the single building panoramic image vector are combined to obtain a P-th level combined vector, and the P-th level combined vector is downsampled by a P-th level downsampling branch in the P-th level downsampling branches to obtain a P-th level downsampling vector; The VR modeling annotation state vector and the k+1th level downsampling vector are combined to obtain a kth level combination vector, and the kth level combination vector is downsampled through the kth level downsampling branch in the P-level downsampling branches to obtain a kth level downsampling vector; The k is walked to obtain a first-level downsampling vector generated by a first-level downsampling branch in the P-level downsampling branches.

Images