Street view data segmentation and modelling
Zhexi Wang · 2013
Large scale urban scene scanned data processing attracts much attention these years. There are many research topics along the whole process, including data capturing, alignment, registration, segmentation, and modelling. My research work is related to data capturing, segmentation and modelling. First we present a real-time stereo reconstruction system which can be used for data capturing. It is used on airplane mounted capturing system and do real-time structure-from-motion by using monocular or stereo cameras, which provides camera pose estimation and similarity reconstruction or metric reconstruction respectively. Then we address the problem of separating objects from 3D scanned point clouds of urban scene. The proposed approach hierarchically performs the grouping from the 3D points to curve segments, object elements, and finally objects. We introduce the relation attributes that describe relations for pairs of object primitives, learn a preference function over such attributes via ranking-SVM which is used to compute the degree that two object primitives belong to an object, and finally merge object elements that are very likely to be contained in the same object. Unlike previous 3D points segmentation algorithms that require object priors to annotate 3D points, our approach only exploits the relation prior that is not limited to any specific object and can separate general urban objects. At last, we present an automatic approach to reconstruct 3D road network models as a part of city models from terrestrial LiDAR and photo data. Terrestrial data provide much higher resolution then aerial data, however, the common terrestrial LiDAR data suffers from occlusion, inconsistency between multiple scans, and the lack of topology information. We introduce the prior knowledge of roads in the form of 2D topology maps, which are widely available on Internet, to assist the reconstruction of roads. A cross-domain alignment method is designed to align the captured data and 2D topology maps. Topology-aware partitioning helps splitting captured data into manageable partitions so that a model for each partition can be generated individually. Finally, all partitions are merged to a global consistent model. Our pipelines are tested in large scale datasets such as San Francisco, New York City, and Paris.