Building Modeling Based On Target Separation

The three-dimensional modeling of buildings is of great significance in many areas such as urban planning,virtual tourism,and cultural relics protection.Compared to the time-consuming and labor-intensive man-made modeling,automated three-dimensional modeling using image processing algorithms is more suitable for large-scale applications or mass modeling scenes,but often due to the complexity of the environment in which the buildings are located,image-based 3D modeling of buildings has not been able to meet the efficiency and precision requirements for commercial use,so this is also a difficult issue in the current study.Usually,in addition to the target building,the building image,as well as the existence of disturbance information such as trees,grasslands,etc.,these additional existences greatly affect the correctness of the matching,which in turn affects the quality of the final generation model.The processing of these existences will become a key issue for improving the quality and efficiency of reconstructed building models.Based on the study of the reconstruction technology of buildings,this paper applies the common target separation idea to the three-dimensional reconstruction of the building.Before the three-dimensional reconstruction,the contour information of the buildings in each image is extracted first,and these contour information are used as baseline,the feature points in the corresponding image are divided into two types of building feature points and non-building feature points,and the two types of feature points are respectively matched to deal with each other.This improves the accuracy of matching on one hand,and on the other hand,it can speed up the matching procedure.The main work of the dissertation can be divided into the following three aspects:Firstly,a building image contour extraction algorithm combining visual salient model and level set algorithm is proposed,based on the analysis of common image contour extraction algorithms.And on the basis of the extracted visual saliency maps of building images,the SLIC superpixel blocks are used as a processing unit to use the texture information of each block to further screen the building area.Then,the contours are optimized by the level set algorithm,and the effectiveness of the method was verified by experiments.Secondly,two key issues of feature matching and view selection in 3D modeling of building targets using incremental SFM algorithm are studied.According to the characteristics of building images,the SURF algorithm is used to extract the feature information in the image.Based on the extracted building outline information,a matching strategy based on the separation of buildings and non-building targets is further proposed.The matching algorithm can improve the accuracy of matching;for the two main problems of the initial SLR pair selection and the best view selection in the incremental SFM algorithm,the solution to the target separation is also provided in combination with the building target itself.Thirdly,based on OpenCV,Ceres Solver,OpenMVG and other open source 3D reconstruction platforms,the 3D modeling system for buildings based on target separation is implemented.The experimental results show that the three-dimensional reconstruction of buildings using the idea of target separation is feasible.And compared with the conventional SFM method,it can improve the efficiency and accuracy of building 3D modeling to a certain extent.