Research

The development of image and 3D processing technology promotes the digitalization of residential environment. Refined 3D reconstruction and large-scale realistic presentation of urban vegetation have application value in urban digitalization, ecological benefit evaluation, space layout planning, and management decision. This project develops new methods and techniques in image, vision and 3D processing, and solves the technical problems of acquisition, reconstruction, processing and presentation of vegetation information in urban digitalization. According to multiple-source data of urban vegetation, research contents are urban vegetation information processing, refined 3D reconstruction of an individual plant and a group of plants based on 3D analysis, interactive reconstruction of plants in complex environment based on similarity analysis, plant modeling based on input restrictions and botanic principals, coding and simplification of plant models, and large-scale fast and realistic presentation. The innovations and the key scientific problems in this project are as follows: segmentation for vegetation and plants in images, refined segmentation of branches, leaves, and stems, hyperspectral image based environmental factor calculation, 3D point cloud acquisition from multi-view images in a large-scale scene, high-precision reconstruction from incomplete plant point cloud, the combination of plant reconstruction and plant growth, inverse process modeling of real trees, urban vegetation scene shape analysis, and fast presentation. The key difficulty is the analysis for coherence of multiple-source data.