| Smart cities are a form of urban operation and management that utilizes information and communication technologies to enhance the sustainability of urban ecological,economic,social,and environmental development,thereby improving the quality of life for city residents.The City Information Modelling(CIM)provides a comprehensive data platform for smart cities,integrating and analyzing data from various urban domains to assist decision-makers in better city management.CIM requires the integration and organization of multi-source heterogeneous data,with a key step being the integration and organization of multi-source 3D model data and Geographic Information System(GIS).However,the spatial reference or geographic coordinate information of the 3D model data may be missing or inaccurate due to its different sources,making it challenging to achieve geo-referencing of multi-source 3D model data in CIM.Therefore,the development of new geo-referencing methods for3 D models is necessary to address the geo-referencing issues of multi-source 3D model data with missing spatial reference or geographic coordinate information in CIM.Remote sensing imagery is one of the main sources of GIS map data,including satellite images and aerial photographs.Although remote sensing imagery and 3D model data are non-homogeneous,identifying and matching corresponding feature points in both the remote sensing imagery and the 3D models can be used for georeferencing the 3D model data,particularly in the case of multi-source 3D model data in CIM.However,calculating feature points that match the 3D model in large-scale remote sensing imagery consumes significant time and computer memory,which hinders the application of feature points for geo-referencing 3D models in CIM.This paper addresses the issues and focuses on the following research contents:(1)Proposed a distributed geo-referencing framework for 3D models.Firstly,based on the scale of the remote sensing imagery and the top view of the 3D models,the largescale remote sensing imagery is divided into smaller image slices using a sliding window approach.Next,a distributed geo-referencing computing cluster is established,where each geo-referencing computing unit consists of three computation stages: image matching between 3D model and remote sensing image slices,calculation of pixel coordinate transformation matrices between the top view of the 3D model and the remote sensing imagery,and geo-referencing of the 3D models.(2)Proposed an online geo-referencing service for 3D models.Firstly,a global feature point database of remote sensing image slices is established,and feature point matching between the top view of the 3D model and each image slice is performed in the distributed feature point matching computing cluster.Then,the pixel coordinates of the feature points on the image slices are transformed back to the original remote sensing imagery.Clustering algorithms are applied to cluster the transformed feature points as data points,and geo-referencing of the 3D models is finally performed within each generated cluster.This paper proposes a distributed geo-referencing framework and an online georeferencing service for 3D models,addressing the challenges of fast and accurate georeferencing of multi-source 3D model data in GIS for CIM.This method lays the foundation for the integration and organization of multi-source 3D model data and GIS data in CIM,promoting the development of CIM in smart cities. |
