Research On Rapid Reconstruction Of Structural 3D Model Based On Image Processing

An earthquake is a natural disaster with great destructive power.After the earthquake,the evaluation of the safety and habitability of buildings in the earthquake zone under the predetermined function is one of the key points of emergency work at the earthquake site.The traditional assessment method is to conduct field visits by experts or professionals with training experience.However,there are various problems with traditional evaluation methods: difficulty in obtaining data;time-consuming and laborious;strong dependence on experts;and potential safety hazards in the process.In response to the problems in traditional assessment and the development of image processing technology,the three-dimensional(3D)model reconstruction technology based on digital image processing has gradually attracted attention in the evaluation of earthquake damaged buildings.However,the research of image three-dimensional reconstruction method in the field of civil engineering is at an early stage.Although the feasibility of this method in civil engineering has been verified,the long reconstruction time of large-scale structural models is still one of the main factors restricting the application of this technology to the rapid assessment of seismic damage structures.In order to solve these problems,this paper proposes to first use image compression and image segmentation techniques to pre-process the two-dimensional digital image,to achieve the purpose of reducing the reconstruction time of the structural threedimensional model.The main research work and innovation results are as follows:(1)In order to reduce the reconstruction time of the structural 3D model,this paper uses principal component analysis algorithm(PCA)to compress and pre-process the structure target image.The basic idea of this method is as follows:(1)According to the actual engineering requirements,the image is collected around the target,and the position of the camera is kept on the same level;(2)Use the principal component analysis algorithm to pre-compress the collected target image;(3)Extract the feature points of the image,and calibrate the camera position by matching the images,and then establish the correspondence between the twodimensional image and the three-dimensional space position to generate a sparse threedimensional point cloud;(4)Use the multi-view stereo vision algorithm to densify the point cloud to generate a dense three-dimensional point cloud;(5)Use Poisson surface reconstruction algorithm to reconstruct dense point cloud into three-dimensional surface;(6)Map texture to the surface of the model to generate a 3D model corresponding to the actual structure.The feasibility of this method is verified through experiments.The feasibility and accuracy of this method are verified by compressing and 3D modeling the actual structure images.It can be concluded through experiments: On the premise of not affecting the accuracy of the model,introducing the image compression algorithm into the technology of 3D model reconstruction can greatly reduce the reconstruction time of the model,about one third of the original time,and can meet the application requirements in actual engineering.(2)When image acquisition is performed on the actual structure,because of the complexity of the situation around the actual structure,the acquired image will contain a lot of useless information in addition to the target.When the image is used to reconstruct the target object,the useless information is also reconstructed,which greatly wastes modeling time and makes the built model very complicated.The image segmentation technology can extract the target objects in the image to reduce the useless information,but there is no general segmentation technology that is suitable for the segmentation of all target objects.Based on the existence of the above problems,this paper proposes an improved image segmentation algorithm,through simple manual interaction to complete the required target segmentation.The target data is extracted through the interaction of the front,back,left,and right images of the target,and the remaining hundreds of images are extracted through the obtained data.Model the new image obtained after extraction.The method was verified by modeling experiments on Rainbow Bridge of Dalian University of Technology.It can be seen from the experimental results: without affecting the accuracy of the model,the image segmentation technology can greatly reduce the modeling time and meet the application in actual engineering.(3)In order to further reduce the modeling time,this paper combines image segmentation technology and image compression technology.First,the image segmentation technology is used to segment the image,and then the segmented image is compressed.The feasibility of this method was verified by modeling experiments on Rainbow Bridge of Dalian University of Technology.According to the experimental results,this method can greatly reduce the modeling time on the basis of meeting the actual engineering requirements.