Three Dimensional Spatial Information Extraction For Buildings With Multi-source Remote Sensing Image And LiDAR Data

With the rapid development of economy and society, the requirement for the 3-D city modeling performance is becoming higher and higher for many civil and military applications, such as city planning/infrastructure construction, environment monitoring/energy consumption assessment, military reconnaissance/precise target attacking, and etc. As buildings are the most important objects in city, extracting informations for 3-D building modeling in urban scale with remote sensing data is always research focus for many relative fields. For addressing this task, satellite or aerial optical images are the most important data sources. However, restricted by key structure detection and matching, the efficiency of 3-D spatial information extraction of buildings from optical image is always low. Recent years, with the continuous mature of the airborne Light Detection and Ranging(Li DAR) technique, obvious complementarities can be seen between airborne Li DAR and remote sensing optical image in aspect to 3-D spatial information extraction of buildings. Therefore, the performance of 3-D spatial information extraction of buildings can be significantly enhanced by integrating these two kinds of data sets together. Nevertheless, almost all the research attentions for integrating these two kinds of data sets to extract 3-D spatial information of buildings are paid on methods or strategies about how to integrate them. Consequently, there are several problems needed to be addressed for exploiting advantages of these two kinds of data sets to support the corresponding integration methods. Therefore, in this dissertation, with exploiting advantages of these two kinds of data sets for integration based 3-D spatial information extraction of buildings as the approach, and with accurate and reliable 3-D spatial information extraction of buildings as the objective, there are mainly three issues concerned:(1) 3-D spatial position information extraction from multi-source remote sensing optical image pair;(2) 3-D spatial structure information restoration of airborne Li DAR data;(3) Optimal toplogical structure constrained 3-D spatial information extraction by integrating these two kinds of data sets. The main contributions are as follows:(1) For improving the capability of information description of raw data, methods for enhancing information description of these two kinds of data sets are researched. For improving the accuracy of 3-D spatial position information description of remote sensing optical image, the accuracy of camera model is improved by designing multi-information based method for selecting evenly distributed ground control points. For improving the capability of 3-D spatial structure information description of airborne Li DAR data, random noise model for the rasterized image of airborne Li DAR data is proposed by analyzing the random error propagation in the procedure of data collection and rasterized image generation.(2) For addressing limitations of the traditional 3-D spatial information extraction methods, e.g. rigorous data requirement, and limited accuracy, a novel 3-D spatial position information extraction method based on multi-source optical stereo image pair is proposed. At first, by analyzing detailed resolution at pixel level, mathematical model of 3-D spatial position information extraction from resolution adaptive multi-source optical stereo image pair is constructed based on the general least squares optimal weighting theory. During the solving procedure of the proposed mathematical model, Variance/covariance Components Estimation(VCE) technique is employed as the accuracy controlling method. Finally, accuracy of the extracted 3-D spatial position information is significantly improved.(3) For addressing the problem that the accuracy and reliability of 3-D spatial structure information of building roofs are always influenced by structure random errors, a structural sparse representation based 3-D spatial structure information restoration framework for building roofs from airborne Li DAR data is proposed. With the help of detailed analysis about the random error model for rasterized image of airborne Li DAR data, a structural sparse representation based 3-D spatial structure information restoration framework for building roofs from airborne Li DAR data is proposed by analyzing the structural properties of building roofs. This framework can realize accurate and reliable 3-D spatial information extraction for buildings fairly.(4) For addressing the problems that structural information in the optical image is always influenced by shadow and other complex detail information, and that the step feature/edge extracted by airborne Li DAR is very coarse and the corresponding performance is not stable, a hierarchical energy minimization based global optimal plane segmentation method for airborne Li DAR data is proposed to realize accrate and stable extraction of the topological structure information. Consequently, accurate and reliable 3-D spatial information of buildings can be extracted by integrating multi-source optical images pair and airborne Li DAR data.