CN115049924A - Building earthquake damage assessment method based on non-structural component damage identification under video monitoring - Google Patents
Building earthquake damage assessment method based on non-structural component damage identification under video monitoring Download PDFInfo
- Publication number
- CN115049924A CN115049924A CN202210631442.0A CN202210631442A CN115049924A CN 115049924 A CN115049924 A CN 115049924A CN 202210631442 A CN202210631442 A CN 202210631442A CN 115049924 A CN115049924 A CN 115049924A
- Authority
- CN
- China
- Prior art keywords
- earthquake
- damage
- structural
- structural member
- building
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/10—Terrestrial scenes
- G06V20/176—Urban or other man-made structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/80—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level
- G06V10/806—Fusion, i.e. combining data from various sources at the sensor level, preprocessing level, feature extraction level or classification level of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a building earthquake damage assessment method based on non-structural component damage identification under video monitoring, belonging to the technical field of earthquake engineering; the method comprises the following steps: acquiring an image of the whole process of damage and destruction of a non-structural component in a public building before and after an earthquake occurs, and dividing a monitoring video into video frames before, during and after the earthquake occurs; carrying out target detection, identification classification and positioning on a non-structural component in a video picture before the occurrence of the earthquake; carrying out target tracking on the successfully identified non-structural component on a video picture in the earthquake occurrence process; identifying and judging the earthquake response and damage states of various non-structural components for the video pictures after the earthquake occurs; and summarizing the damage state information of various non-structural components, and judging the earthquake damage loss degree of the non-structural system in the building. The invention utilizes the video monitoring system which is popularized and distributed in the urban public building to identify the earthquake damage of the non-structural component through the video picture, thereby effectively evaluating the earthquake damage degree in the building.
Description
Technical Field
The invention relates to a method for evaluating earthquake damage based on video monitoring, in particular to a method for evaluating building earthquake damage based on non-structural component damage identification under video monitoring, and belongs to the technical field of earthquake engineering.
Background
Damage and destruction of non-structural members under earthquakes are major factors affecting the recovery of building functions after disasters. From the experience of historical earthquake damage, the safety of the building structure is better ensured at present, but in recent earthquake events, a large number of non-structural components are damaged, so that the functions of important buildings and basic facilities are interrupted, and great obstacles are caused to emergency rescue after earthquake and urban function recovery.
After an earthquake occurs, the work of carrying out earthquake damage investigation and emergency evaluation on an engineering structure is still on-site evaluation by an expert experience method, and the method has the defects of high labor cost and low working efficiency. In addition, the structural health monitoring technology has low application popularity. Although technologies such as satellite remote sensing and unmanned aerial vehicle shooting are used in combination in recent years, damage to structural members, infilled walls and non-structural members of doors and windows, which are exposed to the appearance of buildings, can be identified and collected. However, non-structural members are more susceptible to failure than structural members. Under the condition that the appearance of the building is well represented, the damage state of the non-structural components in the building is unknown, and the loss of the building function and the repair time are difficult to estimate.
The existing earthquake damage assessment technology needs to be separately provided with hardware facilities, such as sensor systems for unmanned aerial vehicle photography and structural health monitoring, but the laying of the hardware facilities needs high extra cost; in addition, the target detection model based on CNN or ViT is adopted in the existing image processing method, but in order to improve the detection accuracy, the calculation cost of the model is often higher, which is not favorable for the wide popularization of the evaluation method. With the development of the internet of things and the internet technology, video monitoring equipment, smart phones and the like become effective ways for acquiring earthquake damage information inside buildings. In urban public buildings such as hospitals, schools and shopping malls, video monitoring systems are generally arranged for daily safety maintenance. The traffic road monitoring system and the factory production monitoring system can meet specific safety maintenance and work requirements by matching with an intelligent image recognition technology and an internet of things technology.
Therefore, it is highly desirable to construct an earthquake damage assessment method capable of capturing and recording the earthquake damage phenomenon of indoor non-structural members by using a video monitoring system in a public building, and identifying the damage of the non-structural members through video monitoring before and after the earthquake occurs, so as to effectively acquire the building function loss and accurately assess the earthquake toughness of the building.
Disclosure of Invention
The purpose of the invention is: the building earthquake damage assessment method based on the non-structural component damage identification under video monitoring is provided for solving the problems that the existing building earthquake damage assessment technical means only aims at structural system damage or building appearance damage, the indoor non-structural component earthquake damage is not considered, and the existing assessment method is low in precision and high in cost.
In order to achieve the purpose, the invention adopts the following technical scheme: a building earthquake damage assessment method based on non-structural member damage identification under video monitoring comprises the following steps:
s1, acquiring images of the whole damage and damage process of non-structural members in the public building before and after an earthquake occurs based on a video monitoring system arranged in the urban public building, and dividing a monitoring video into video frames before the earthquake occurs, in the earthquake occurring process and after the earthquake occurs according to the earthquake occurrence time and duration obtained after the earthquake occurs;
s2, based on the video picture before the earthquake happens after being segmented in the step S1, a multi-scale ViT network model is adopted to segment and detect the targets of the non-structural members in the video picture, the non-structural members are subjected to target identification and classification according to the connection mode and the position dependency relationship with the structural members, rectangular positioning bounding boxes of the non-structural members are obtained, then grid division is carried out on the video picture, local coordinate axes are defined, and the coordinate positions of main nodes of the outer contour of the non-structural members are determined;
s3, based on the video picture in the earthquake occurrence process after being divided in the step S1, combining the multi-scale ViT network model in the step S2 with a Deepsort target tracking algorithm, training the non-structural member which is successfully identified in the step S2 and is provided with a rectangular positioning bounding box as the input of the Deepsort algorithm, recording the coordinate change and the motion trail of the main nodes of the outer contour of each non-structural member, and tracking the target of the non-structural member;
s4, recording the coordinate position of the main node of the outer contour of the non-structural member successfully identified in the step S2 based on the video picture after the earthquake is generated after the segmentation in the step S1, and judging the motion trail of various non-structural members by using a Deepsort algorithm according to the non-structural member category completing the target identification and classification in the step S3 so as to determine the earthquake response and the damage state of various non-structural members;
s5, acquiring earthquake response and damage information of various non-structural members in video monitoring through the step S4, conjointly with an existing open non-structural member vulnerability database, according to the correlation of vulnerability models of different types of non-structural members, presuming the earthquake response and damage information of the non-structural members which are not covered by video monitoring, summarizing the damage state information of various non-structural members, and evaluating the building function loss degree by adopting repair cost and repair time based on the summarized damage state of the non-structural members;
the calculation formula of the repair cost is as follows: l is (i，j，k) ＝η 1(i，j) ×C (i，j，k) Wherein i, j, k respectively represent the component type, damage state and the number of the floor；L (i,j,k) The economic loss corresponding to the ith component in the damaged state j in the kth layer of the building structure is obtained; c (i,j,k) Is the sum of the manufacturing cost of the ith member in the damaged state j in the kth layer; eta 1(i，j) The damage coefficient of the ith type of component in a damaged state j can be evaluated according to the evaluation standard of building earthquake resistance toughness according to different types of non-structural components;
the formula for calculating the repair time is:
in the formula Q (i,k) Refers to the sum of the repair time of the ith component of the kth layer; n is a radical of an alkyl radical (i,j,k) Representing the number of the ith type components of the k layer in a damaged state j; q (i,j,k) Indicating repair man-hour, ξ, of the ith component with the kth layer in damaged state j T(i) Is a repair man-hour reduction coefficient, lambda, taking into account the repair work volume of the i-th type seismic damage component T(k) To take account of the floor influence coefficient, Q, of the floor position k at which the seismic damage component is located (i,j,k) 、ξ T(i) 、λ T(k) The value can be taken according to the evaluation standard of building earthquake resistance toughness.
In the step S1, the urban public building in which the video monitoring system is arranged includes a hospital, a school, a shopping mall, a government office building and a commercial office building; the video monitoring system can integrate, store, transmit and process analysis within the legal permission range; the non-structural components are all components and articles related to the use function of the building except the structural components in the single building.
In step S1, the earthquake occurrence time and duration are obtained as follows: after an earthquake occurs, main information of the earthquake occurrence, including magnitude, epicenter position, seismic source depth and main earthquake duration, can be issued by authoritative government affairs such as national earthquake bureau and technical departments, the earthquake table net can monitor the earthquake wave process, so that earthquake occurrence time and earthquake ending time can be obtained, and the earthquake occurrence time and the earthquake duration can be obtained through the time information.
In step S2, the specific steps of performing target segmentation and detection on the non-structural member in the video picture by using the multi-scale ViT network model are as follows:
s21, dividing an input video picture into a plurality of picture blocks at intervals of a certain region size by a multi-scale ViT network model, and extracting features with different dimensions from the divided picture blocks to obtain a multi-scale effective feature layer containing high-level semantic information, wherein the features can be shared among the multi-scale effective feature layers;
s22, inputting the extracted multi-scale effective feature layer into a feature pyramid for feature fusion to obtain { P 1 ,P 2 ,P 3 ,P 4 Four effective characteristic layers, the obtained { P } 1 ,P 2 ,P 3 ,P 4 Converting four effective characteristic layers into { T } correspondingly respectively 1 ,T 2 ,T 3 ,T 4 -multi-scale sequence representation;
s23, stacking sequence representations with different scales along the channel dimension, inputting the sequence representations into a multi-head attention mechanism to generate high-level semantic information with scale perception and perform feature extraction, inputting the high-level semantic information into a detection head to complete a detection task of the non-structural member, and finally determining a rectangular positioning bounding box of each non-structural member.
In step S2, the non-structural members are classified into floating, fixed and suspended non-structural members according to the connection mode with the structural members, and the floating non-structural members are members and articles which are not fixedly connected with the floor or other structural members, and include tables and chairs, vertical air conditioners, shelves and bookcases; fixed non-structural members are members and articles that are fixedly connected to a floor or other structural member, including pipes, elevators, and integrated cabinets; the suspension type non-structural member is a member and an article which are flexibly connected with the floor system, and comprises a suspended ceiling, a ceiling lamp and a support and hanger frame;
classifying the non-structural members into two categories according to the position subordination relationship, wherein one category is the non-structural members which are directly placed on a floor slab or connected with the structural members, and the non-structural members comprise tables, chairs, racks, pipelines and suspended ceilings; the second category is components and items placed on the first category of non-structural components, including computers, cups, and shelving items.
In step S3, the specific steps of performing target tracking on the non-structural member by combining the multi-scale ViT network model and the Deepsort target tracking algorithm are as follows:
s31, according to the local coordinate axis defined in the step S2, determining position coordinate information and category information of the successfully identified main nodes of the outer contour of the non-structural member at the initial earthquake occurrence moment;
and S32, recording the coordinate change of the main node of the outer contour of the non-structural member at regular intervals of time frame by using a Deepsort algorithm, and realizing the tracking task of the motion trail of the detection object.
In step S4, the non-structural member damage state identified and determined mainly includes slip response and overturning damage of the floating non-structural member, connection damage of the fixed non-structural member, and sway response and connection damage of the suspended non-structural member, and the specific determination method of the damage state is as follows:
float-over non-structural member: the position of the main node of the outer contour after the earthquake is obviously changed in the horizontal direction compared with the position before the earthquake, and the slippage response is judged when the relative position of the main node of the outer contour is not changed; when the relative position of the main nodes of the outer contour changes, the main nodes are judged to be overturn damage;
fixed non-structural member: when the position of the main node of the outer contour is obviously changed compared with the position before the earthquake happens, the connection is judged to be damaged;
suspended non-structural member: when the main nodes of the outer contour generate regular reciprocating coordinate change relative to the original position in the earthquake occurrence process and after the earthquake occurrence, judging the main nodes as shaking response and recording the shaking amplitude; and judging that the connection is damaged when the main coordinate position of the main node of the outer contour is obviously changed after the earthquake occurs than before the earthquake occurs.
The invention has the beneficial effects that:
1) the invention utilizes the video monitoring system which is popularized and distributed in the modern urban public building to identify the earthquake damage of the non-structural member through the video picture so as to effectively evaluate the earthquake damage degree in the building, and the earthquake damage evaluation method forms effective complementation with the existing earthquake damage evaluation technology which mainly takes the structural member and the building appearance damage as main factors, and provides decision reference for emergency rescue and earthquake damage evaluation.
2) The invention can meet the use requirement by using the video monitoring system for daily safety maintenance in the public building, and realizes the damage identification of the non-structural member through the transmission, collection and processing of the video images so as to obtain the internal earthquake damage condition of the building. The earthquake damage assessment method is different from the existing earthquake damage assessment technology that hardware facilities such as sensor systems for unmanned aerial vehicle photography and structural health monitoring need to be arranged independently, the technology of the earthquake damage assessment method is higher in popularity, and extra cost needed by the hardware facilities is lower.
3) The invention adopts a novel multi-scale ViT network model architecture in the aspects of target detection and segmentation of non-structural members, namely, the idea of respectively extracting the features of each input image block to obtain a series of feature representations is adopted; feature representations from different scales are stacked along the channel dimension, enabling the generation of high-level semantic features with scale perception. In addition, information among multi-scale effective feature layers extracted for each input image block is shared, so that the overall layout and original position information of each image block can be obviously enhanced, and information exchange of global self-attention feature representation in a spatial dimension is enhanced. On the other hand, the multi-scale information fusion is carried out on the sequence representation extracted from each image block, so that the detection precision is ensured, and the calculation cost of the model is greatly reduced. In summary, compared with the original CNN or ViT-based target detection model, the proposed multi-scale ViT target detection model can greatly improve the detection accuracy and reduce the calculation cost of the model.
4) The method adopts the idea of combining the multi-scale ViT target detection model and the Deepsort target tracking model and optimizes the Deepsort target tracking model by using the multi-scale ViT target detection model. The proposed high-efficiency multi-scale ViT target detection model is adopted to perform rectangular positioning boundary frame positioning on the non-structural components in the surveillance video and track the sizes and the positions of various non-structural components in the surveillance video, so that the overlapping and shielding phenomena among targets can be reduced, the problem of re-recognition after the targets are lost can be solved, and the precision of a multi-target tracking task can be greatly improved.
Drawings
FIG. 1 is a flow chart of the earthquake damage assessment method of the present invention;
FIG. 2 is a schematic diagram of the monitoring video image being segmented according to the earthquake occurrence time in step S1 according to the present invention;
FIG. 3 is a schematic diagram of a multi-scale ViT network model architecture for detecting non-structural components in a surveillance video image in step S2;
FIG. 4 is a schematic diagram illustrating the steps of S2 of the method of the present invention for identifying, classifying and positioning the non-structural members in the monitored video image;
FIG. 5 is a schematic diagram of the target tracking and position change recording of the non-structural member in step S3 according to the present invention;
fig. 6 is a schematic diagram illustrating the identification and determination of the damage state of the non-structural member in step S4 according to the method of the present invention.
Detailed Description
The invention is further explained below with reference to the figures and the embodiments.
Example (b): as shown in fig. 1-6, the method for evaluating damage of a non-structural member based on video surveillance for evaluating earthquake damage of a building, as shown in the flow chart of fig. 1, comprises the following steps:
s1, acquiring images of the whole process of damage and damage of non-structural components in public buildings before and after an earthquake occurs based on a video monitoring system arranged in the urban public buildings, wherein the image is the acquired video monitoring image in a certain retail commercial building under a strong earthquake event shown in FIG. 2;
according to the earthquake occurrence time and duration obtained after the earthquake occurs, the monitoring video is divided into video frames before the earthquake occurs (figure 2a), in the earthquake occurrence process (figure 2b) and after the earthquake occurs (figure 2c), the earthquake duration is about 42 seconds, and a single frame at three moments after the video image is divided is shown in figure 2.
S2, based on the video picture before the earthquake happens after being segmented in the step S1, adopting a multi-scale ViT network model to segment and detect the targets of the non-structural components in the video picture, identifying and classifying the targets of the non-structural components according to the connection mode and the position subordination relation with the structural components to obtain a rectangular positioning bounding box of each non-structural component, then carrying out grid division on the video picture and defining local coordinate axes, and determining the coordinate positions of main nodes of the outer contour of each non-structural component;
as shown in fig. 3, the specific steps of performing target segmentation and detection on a non-structural member in a video picture by using a multi-scale ViT network model are as follows:
s21, dividing an input video picture into a plurality of picture blocks at intervals of a certain region size by a multi-scale ViT network model, and extracting features with different dimensions from the divided picture blocks to obtain a multi-scale effective feature layer containing high-level semantic information, wherein the features can be shared among the multi-scale effective feature layers;
s22, inputting the extracted multi-scale effective feature layer into a feature pyramid for feature fusion to obtain { P 1 ,P 2 ,P 3 ,P 4 Four effective characteristic layers, the obtained { P } 1 ,P 2 ,P 3 ,P 4 Converting four effective characteristic layers into { T } correspondingly respectively 1 ,T 2 ,T 3 ,T 4 -multiscale sequence representation, wherein N represents the number of sequences;
s23, stacking sequence representations with different scales along the channel dimension, inputting the stacked sequence representations into a multi-head attention mechanism to generate high-level semantic information with scale perception, extracting features, inputting the high-level semantic information into a detection head part to complete a detection task of the non-structural member, and finally determining a rectangular positioning boundary box of each non-structural member.
The non-structural members are classified into floating type, fixed type and suspension type non-structural members according to the connection mode with the structural members, the floating type non-structural members are members and articles which are not fixedly connected with a floor slab or other structural members, and the floating type non-structural members comprise tables and chairs, vertical air conditioners, goods shelves and bookcases; fixed non-structural members are members and articles that are fixedly connected to a floor or other structural member, including pipes, elevators, and integrated cabinets; the suspension type non-structural member is a member and an article which are flexibly connected with the floor system, and comprises a suspended ceiling, a ceiling lamp and a support and hanger frame;
classifying the non-structural members into two categories according to the position subordination relationship, wherein one category is the non-structural members which are directly placed on a floor slab or connected with the structural members, and the non-structural members comprise tables, chairs, racks, pipelines and suspended ceilings; the second category is components and items placed on the first category of non-structural components, including computers, cups, and shelving items.
As shown in fig. 4, for the video frame before the earthquake, the decorative articles on the chair, the stereo, the pot and the cabinet are successfully identified, and the rectangular positioning boundary frame of each non-structural member is represented by black frame lines.
S3, based on the video picture in the earthquake occurrence process segmented in the step S1, combining the multi-scale ViT network model in the step S2 with a Deepsort target tracking algorithm, training the non-structural component with the rectangular positioning bounding box successfully identified in the step S2 as the input of the Deepsort algorithm, recording the coordinate change and the motion trail of the main nodes of the outer contour of each non-structural component, and tracking the target of the non-structural component;
the specific steps of tracking the target of the non-structural member by combining the multi-scale ViT network model and the Deepsort target tracking algorithm are as follows:
s31, according to the local coordinate axis defined in the step S2, determining position coordinate information and category information of the successfully identified main nodes of the outer contour of the non-structural member at the initial earthquake occurrence moment;
and S32, recording the coordinate change of the main nodes of the outer contour of the non-structural member at regular time frames (such as 0.1S and 0.5S) by using a Deepsort algorithm, and realizing the tracking task of the motion trail of the detection object.
As shown in fig. 5, a single frame of image in the earthquake occurrence process is displayed, and compared with the coordinate position of each non-structural member before the earthquake occurs, it can be identified that the potted plant is toppled, the two sound boxes slide and shake, the chair slides, and part of ornament articles on the rack cabinet fall off.
S4, recording the coordinate position of the main node of the outer contour of the non-structural member successfully identified in the step S2 based on the video picture after the earthquake is generated after the segmentation in the step S1, and judging the motion trail of various non-structural members by using a Deepsort algorithm according to the non-structural member category completing the target identification and classification in the step S3 so as to determine the earthquake response and the damage state of various non-structural members;
the non-structural component damage state of identification judgement mainly includes the slippage response and the overturning damage of the floating non-structural component, the connection damage of the fixed non-structural component, the shaking response and the connection damage of the suspension non-structural component, and the specific judgement method of the damage state is:
float-over non-structural member: the position of the main node of the outer contour after the earthquake is obviously changed in the horizontal direction compared with the position before the earthquake, and the slippage response is judged when the relative position of the main node of the outer contour is not changed; when the relative position of the main nodes of the outer contour changes, the main nodes are judged to be overturn damage;
fixed non-structural member: when the position of the main node of the outer contour is obviously changed compared with the position before the earthquake occurs, judging that the connection is damaged;
suspended non-structural member: when the main nodes of the outer contour generate regular reciprocating coordinate change relative to the original position in the earthquake occurrence process and after the earthquake occurrence, judging as shaking response and recording the shaking amplitude; and judging that the connection is damaged when the main coordinate position of the main node of the outer contour is obviously changed after the earthquake occurs than before the earthquake occurs.
In the single frame image of the video image after the earthquake shown in fig. 6, it can be determined that the potted plant is overturned, one stereo is overturned, the other stereo is slid, the chair is slid, and the goods of furniture on the rack are dropped.
S5, acquiring earthquake response and damage information of various non-structural members in video monitoring through the step S4, conjointly with an existing open non-structural member vulnerability database, according to the correlation of vulnerability models of different types of non-structural members, presuming the earthquake response and damage information of the non-structural members which are not covered by video monitoring, summarizing the damage state information of various non-structural members, and evaluating the building function loss degree by adopting repair cost and repair time based on the summarized damage state of the non-structural members;
the calculation formula of the repair cost is as follows: l is (i，j，k) ＝η 1(i，j) ×C (i，j，k) Wherein i, j and k respectively represent the component type, the damage state and the number of the floor where the component is located; l is (i,j,k) The economic loss corresponding to the ith component in the damaged state j in the kth layer of the building structure is obtained; c (i,j,k) Is the sum of the manufacturing cost of the ith member in the damaged state j in the kth layer; eta 1(i，j) The damage coefficient of the ith type of component in a damaged state j can be evaluated according to the evaluation standard of building earthquake resistance toughness according to different types of non-structural components;
the formula for calculating the repair time is:
in the formula Q (i,k) Refers to the sum of the repair time of the ith component of the kth layer; n is (i,j,k) Representing the number of the ith type components of the k layer in a damaged state j; q (i,j,k) Indicating repair man-hour, ξ, of the i-th class member with the k-th layer in a damaged state j T(i) Is a repair man-hour reduction coefficient, lambda, taking into account the repair work volume of the i-th type seismic damage component T(k) To take account of the floor influence coefficient, Q, of the floor position k at which the seismic damage component is located (i,j,k) 、ξ T(i) 、λ T(k) The value can be taken according to the evaluation standard of building earthquake resistance toughness.
According to the identification judgment of the step S4, the overturning damage number of the floating type non-structural component identified by the camera is about 20, and the total value is 4000 yuan. According to the literature and the evaluation standard of earthquake-resistant toughness of buildings (GB/T38591-2020), the loss ratio eta of the overturn damage is assumed 1(i，j) Is 0.5, and can be calculated to be 2000 yuan by adopting a repair cost formula. Presumed floatingRepair man-hour Q for placing type non-structural component in overturn damage state (i,j,k) 1 hour, ξ T(i) Values of 0.6, λ T(k) The value is 1.0, and the repair time is calculated to be 12 hours according to a repair time formula.
The invention utilizes the video monitoring system which is popularized and distributed in the modern urban public building to identify the earthquake damage of the non-structural member through the video picture so as to effectively evaluate the earthquake damage degree in the building, and the earthquake damage evaluation method forms effective complementation with the existing earthquake damage evaluation technology which mainly takes the structural member and the building appearance damage as main factors, and provides decision reference for emergency rescue and earthquake damage evaluation.
The above description is only for the purpose of illustrating the technical solutions of the present invention and not for the purpose of limiting the same, and other modifications or equivalent substitutions made by those skilled in the art to the technical solutions of the present invention should be covered within the scope of the claims of the present invention without departing from the spirit and scope of the technical solutions of the present invention.
Claims (7)
1. A building earthquake damage assessment method based on non-structural member damage identification under video monitoring is characterized in that: the method comprises the following steps:
s1, acquiring images of the whole damage and damage process of non-structural members in the public building before and after an earthquake occurs based on a video monitoring system arranged in the urban public building, and dividing a monitoring video into video frames before the earthquake occurs, in the earthquake occurring process and after the earthquake occurs according to the earthquake occurrence time and duration obtained after the earthquake occurs;
s2, based on the video picture before the earthquake happens after being segmented in the step S1, a multi-scale ViT network model is adopted to segment and detect the targets of the non-structural members in the video picture, the non-structural members are subjected to target identification and classification according to the connection mode and the position dependency relationship with the structural members, rectangular positioning bounding boxes of the non-structural members are obtained, then grid division is carried out on the video picture, local coordinate axes are defined, and the coordinate positions of main nodes of the outer contour of the non-structural members are determined;
s3, based on the video picture in the earthquake occurrence process after being divided in the step S1, combining the multi-scale ViT network model in the step S2 with a Deepsort target tracking algorithm, training the non-structural member which is successfully identified in the step S2 and is provided with a rectangular positioning bounding box as the input of the Deepsort algorithm, recording the coordinate change and the motion trail of the main nodes of the outer contour of each non-structural member, and tracking the target of the non-structural member;
s4, recording the coordinate position of the main node of the outer contour of the non-structural member successfully identified in the step S2 based on the video picture after the earthquake is generated after the segmentation in the step S1, and judging the motion trail of various non-structural members by using a Deepsort algorithm according to the non-structural member category completing the target identification and classification in the step S3 so as to determine the earthquake response and the damage state of various non-structural members;
s5, acquiring earthquake response and damage information of various non-structural members in video monitoring through the step S4, conjointly with an existing open non-structural member vulnerability database, according to the correlation of vulnerability models of different types of non-structural members, presuming the earthquake response and damage information of the non-structural members which are not covered by video monitoring, summarizing the damage state information of various non-structural members, and evaluating the building function loss degree by adopting repair cost and repair time based on the summarized damage state of the non-structural members;
the calculation formula of the repair cost is as follows: l is (i，j，k) ＝η 1(i，j) ×C (i，j，k) Wherein i, j and k respectively represent the component type, the damage state and the number of the floor where the component is located; l is (i,j,k) The economic loss corresponding to the ith component in the damaged state j in the kth layer of the building structure is obtained; c (i,j,k) Is the sum of the manufacturing cost of the ith type component in the damaged state j in the kth layer; eta 1(i，j) The damage coefficient of the ith type of component in a damaged state j can be evaluated according to the evaluation standard of building earthquake resistance toughness according to different types of non-structural components;
the formula for calculating the repair time is:
in the formula Q (i,k) Refers to the sum of the repair time of the ith component of the kth layer; n is (i,j,k) Representing the number of the ith type components of the k layer in a damaged state j; q (i,j,k) Indicating repair man-hour, ξ, of the ith component with the kth layer in damaged state j T(i) Is a repair man-hour reduction coefficient, lambda, taking into account the repair work volume of the i-th type seismic damage component T(k) To take account of the floor influence coefficient, Q, of the floor position k at which the seismic damage component is located (i,j,k) 、ξ T(i) 、λ T(k) The value can be taken according to the evaluation standard of building earthquake resistance toughness.
2. The method for assessing the earthquake damage of a building based on the identification of the damage of a non-structural member under video surveillance as claimed in claim 1, wherein: in step S1, the city public building in which the video monitoring system is disposed includes a hospital, a school, a shopping mall, a government office building and a commercial office building; the video monitoring system can integrate, store, transmit and process analysis within the legal permission range; the non-structural components are all components and articles related to the use function of the building except the structural components in the single building.
3. The method for assessing the earthquake damage of a building based on the identification of the damage of a non-structural member under video surveillance as claimed in claim 1, wherein: in step S1, the earthquake occurrence time and the earthquake duration time are obtained by: after an earthquake occurs, main information of the earthquake occurrence, including magnitude, epicenter position, seismic source depth and main earthquake duration, can be issued by authoritative government affairs such as national earthquake bureau and technical departments, the earthquake table net can monitor the earthquake wave process, so that earthquake occurrence time and earthquake ending time can be obtained, and the earthquake occurrence time and the earthquake duration can be obtained through the time information.
4. The building earthquake damage assessment method based on non-structural member damage identification under video surveillance as claimed in claim 1, wherein: in step S2, the specific steps of performing target segmentation and detection on the non-structural member in the video picture by using the multi-scale ViT network model are as follows:
s21, dividing an input video picture into a plurality of picture blocks at intervals of a certain region size by a multi-scale ViT network model, and extracting features with different dimensions from the divided picture blocks to obtain a multi-scale effective feature layer containing high-level semantic information, wherein the features can be shared among the multi-scale effective feature layers;
s22, inputting the extracted multi-scale effective characteristic layer into a characteristic pyramid for characteristic fusion to obtain { P 1 ,P 2 ,P 3 ,P 4 Four effective characteristic layers, the obtained { P } 1 ,P 2 ,P 3 ,P 4 Converting four effective characteristic layers into (T) respectively 1 ,T 2 ,T 3 ,T 4 -multi-scale sequence representation;
s23, stacking sequence representations with different scales along the channel dimension, inputting the sequence representations into a multi-head attention mechanism to generate high-level semantic information with scale perception, extracting features, inputting the extracted features into a detection head to complete a detection task of the non-structural member, and finally determining a rectangular positioning bounding box of each non-structural member.
5. The method for assessing the earthquake damage of a building based on the identification of the damage of a non-structural member under video surveillance as claimed in claim 1, wherein: in step S2, the non-structural members are classified into floating, fixed and suspended non-structural members according to the connection mode with the structural members, and the floating non-structural members are members and articles which are not fixedly connected with the floor or other structural members, and include tables and chairs, vertical air conditioners, shelves and bookcases; fixed non-structural components are components and articles which are fixedly connected with a floor slab or other structural components, and comprise pipelines, elevators and integral cabinets; the suspension type non-structural member is a member and an article which are flexibly connected with the floor system, and comprises a suspended ceiling, a ceiling lamp and a support and hanger frame;
classifying the non-structural members into two categories according to the position subordination relationship, wherein one category is the non-structural members which are directly placed on a floor slab or connected with the structural members, and the non-structural members comprise tables, chairs, racks, pipelines and suspended ceilings; the second category is components and items placed on the first category of non-structural components, including computers, cups, and shelving items.
6. The building earthquake damage assessment method based on non-structural member damage identification under video surveillance as claimed in claim 1, wherein: in step S3, the specific steps of performing target tracking on the non-structural member by combining the multi-scale ViT network model and the Deepsort target tracking algorithm are as follows:
s31, according to the local coordinate axis defined in the step S2, determining position coordinate information and category information of the successfully identified main nodes of the outer contour of the non-structural member at the initial earthquake occurrence moment;
and S32, recording the coordinate change of the main node of the outer contour of the non-structural member at regular intervals of time frame by using a Deepsort algorithm, and realizing the tracking task of the motion trail of the detection object.
7. The method for assessing the earthquake damage of a building based on the identification of the damage of a non-structural member under video surveillance as claimed in claim 1, wherein: in step S4, the non-structural member damage state identified and determined mainly includes slip response and overturning damage of the floating non-structural member, connection damage of the fixed non-structural member, and sway response and connection damage of the suspended non-structural member, and the specific determination method of the damage state is as follows:
float-over non-structural member: the position of the main node of the outer contour after the earthquake is obviously changed in the horizontal direction compared with the position before the earthquake, and the slippage response is judged when the relative position of the main node of the outer contour is not changed; when the relative position of the main nodes of the outer contour changes, the main nodes are judged to be overturn damage;
fixed non-structural member: when the position of the main node of the outer contour is obviously changed compared with the position before the earthquake occurs, judging that the connection is damaged;
suspended non-structural member: when the main nodes of the outer contour generate regular reciprocating coordinate change relative to the original position in the earthquake occurrence process and after the earthquake occurrence, judging as shaking response and recording the shaking amplitude; and judging that the connection is damaged when the main coordinate position of the main node of the outer contour is obviously changed after the earthquake occurs than before the earthquake occurs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210631442.0A CN115049924B (en) | 2022-06-06 | 2022-06-06 | Building earthquake damage assessment method based on non-structural member damage identification under video monitoring |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210631442.0A CN115049924B (en) | 2022-06-06 | 2022-06-06 | Building earthquake damage assessment method based on non-structural member damage identification under video monitoring |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115049924A true CN115049924A (en) | 2022-09-13 |
| CN115049924B CN115049924B (en) | 2023-04-14 |
Family
ID=83158840
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210631442.0A Active CN115049924B (en) | 2022-06-06 | 2022-06-06 | Building earthquake damage assessment method based on non-structural member damage identification under video monitoring |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115049924B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116310791A (en) * | 2023-01-19 | 2023-06-23 | 中国地震台网中心 | Rapid judgment method and electronic equipment for extremely disaster area based on building earthquake damage detection |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20180089598A (en) * | 2017-01-31 | 2018-08-09 | 연세대학교 산학협력단 | Configuration Method of Seismic Performance Evaluation for Building Structures Considering Repair Cost of Structural and Nonstructural Components |
| CN109918710A (en) * | 2019-01-21 | 2019-06-21 | 北京科技大学 | It is a kind of consider non-structural element indoor shake after three-dimensional virtual scene construction method |
| CN110516556A (en) * | 2019-07-31 | 2019-11-29 | 平安科技（深圳）有限公司 | Multi-target tracking detection method, device and storage medium based on Darkflow-DeepSort |
| CN110580443A (en) * | 2019-06-19 | 2019-12-17 | 深圳大学 | Low-altitude near-real-time building earthquake damage assessment method |
| CN113095127A (en) * | 2021-03-01 | 2021-07-09 | 兰州大学 | Building post-earthquake positioning and damage state evaluation method based on satellite images |
| CN114170411A (en) * | 2021-12-06 | 2022-03-11 | 国能大渡河大岗山发电有限公司 | Picture emotion recognition method integrating multi-scale information |
| CN114239108A (en) * | 2021-12-17 | 2022-03-25 | 四川大学 | Urban building group loss distribution calculation method after earthquake based on monitoring Internet of things |
| CN114417472A (en) * | 2022-01-21 | 2022-04-29 | 四川大学 | Non-structural system seismic loss assessment method considering multi-dimensional seismic input |
-
2022
- 2022-06-06 CN CN202210631442.0A patent/CN115049924B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR20180089598A (en) * | 2017-01-31 | 2018-08-09 | 연세대학교 산학협력단 | Configuration Method of Seismic Performance Evaluation for Building Structures Considering Repair Cost of Structural and Nonstructural Components |
| CN109918710A (en) * | 2019-01-21 | 2019-06-21 | 北京科技大学 | It is a kind of consider non-structural element indoor shake after three-dimensional virtual scene construction method |
| CN110580443A (en) * | 2019-06-19 | 2019-12-17 | 深圳大学 | Low-altitude near-real-time building earthquake damage assessment method |
| CN110516556A (en) * | 2019-07-31 | 2019-11-29 | 平安科技（深圳）有限公司 | Multi-target tracking detection method, device and storage medium based on Darkflow-DeepSort |
| CN113095127A (en) * | 2021-03-01 | 2021-07-09 | 兰州大学 | Building post-earthquake positioning and damage state evaluation method based on satellite images |
| CN114170411A (en) * | 2021-12-06 | 2022-03-11 | 国能大渡河大岗山发电有限公司 | Picture emotion recognition method integrating multi-scale information |
| CN114239108A (en) * | 2021-12-17 | 2022-03-25 | 四川大学 | Urban building group loss distribution calculation method after earthquake based on monitoring Internet of things |
| CN114417472A (en) * | 2022-01-21 | 2022-04-29 | 四川大学 | Non-structural system seismic loss assessment method considering multi-dimensional seismic input |
Non-Patent Citations (5)
| Title |
|---|
| ALEXEY DOSOVITSKIY ET AL.: "AN IMAGE IS WORTH 16X16 WORDS:TRANSFORMERS FOR IMAGE RECOGNITION AT SCALE", 《ICLR 2021》 * |
| NICOLAI WOJKE ET AL.: "SIMPLE ONLINE AND REALTIME TRACKING WITH A DEEP ASSOCIATION METRIC", 《ARXIV》 * |
| XIAONING ZHU ET AL.: "ViTT: Vision Transformer Tracker", 《SENSORS》 * |
| 任军宇 等: "ＧＢ／Ｔ３８５９１—２０２０《建筑抗震韧性评价标准》解读", 《建筑结构学报》 * |
| 黄凯文 等: "基于改进 ＹＯＬＯ和 ＤｅｅｐＳＯＲＴ的实时多目标跟踪算法", 《电子测量技术》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116310791A (en) * | 2023-01-19 | 2023-06-23 | 中国地震台网中心 | Rapid judgment method and electronic equipment for extremely disaster area based on building earthquake damage detection |
| CN116310791B (en) * | 2023-01-19 | 2023-09-05 | 中国地震台网中心 | Rapid judgment method and electronic equipment for extremely disaster area based on building earthquake damage detection |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115049924B (en) | 2023-04-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112396027B (en) | Vehicle re-identification method based on graph convolution neural network | |
| CN104123544B (en) | Anomaly detection method and system based on video analysis | |
| KR20200006167A (en) | Vessel automatic tracking method and system based on deep learning network and average movement | |
| WO2020170014A1 (en) | Object counting and instance segmentation using neural network architectures with image-level supervision | |
| CN110147743A (en) | Real-time online pedestrian analysis and number system and method under a kind of complex scene | |
| CN108073888A (en) | A kind of teaching auxiliary and the teaching auxiliary system using this method | |
| CN107016357A (en) | A kind of video pedestrian detection method based on time-domain convolutional neural networks | |
| CN101635835A (en) | Intelligent video monitoring method and system thereof | |
| CN115049924B (en) | Building earthquake damage assessment method based on non-structural member damage identification under video monitoring | |
| CN114239108B (en) | Urban building group loss distribution calculation method after earthquake based on monitoring Internet of things | |
| CN112115906A (en) | Open dish identification method based on deep learning target detection and metric learning | |
| CN105760846A (en) | Object detection and location method and system based on depth data | |
| CN109272554A (en) | A kind of method and system of the coordinate system positioning for identifying target and semantic map structuring | |
| CN100565557C (en) | System for tracking infrared human body target based on corpuscle dynamic sampling model | |
| CN107657215B (en) | Indoor behavior track motion semantic analysis method based on passive infrared sensor | |
| CN109918710A (en) | It is a kind of consider non-structural element indoor shake after three-dimensional virtual scene construction method | |
| CN110276379B (en) | Disaster information rapid extraction method based on video image analysis | |
| CN106683077A (en) | Escalator floor board large-object retention detection method | |
| CN114707604A (en) | Twin network tracking system and method based on space-time attention mechanism | |
| CN103425958B (en) | A kind of method of motionless analyte detection in video | |
| CN108876824A (en) | Method for tracking target, device, system and ball machine | |
| CN106056030A (en) | Method and Apparatus for counting the number of person | |
| Yao et al. | Using machine learning approach to construct the people flow tracking system for smart cities | |
| CN116187634A (en) | Intelligent queuing system and prediction method for same | |
| CN115393753A (en) | Dangerous chemical management and control device and method based on edge cluster and machine vision |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |