Potential of Satellite Image in the End-To-End 3D City Modelling for Developing Regions: A Case Study for Cau Giay District, Hanoi, Vietnam

Abstract

Recently, the problem of missing building foundation and height data is making it difficult to develop 3D GIS in developing countries due to the high cost of accurate data collection and data sharing policy. In this study, we proposed an approach for 3D city modeling that is suitable for these areas, and then the experiment was carried out for Cau Giay district, Vietnam. First, building footprints will be obtained from the free high-resolution satellite imagery source. Then, the building footprints themself and the surrounding area images are used to extract key points by some well-known descriptors. K-means will then be applied to create bag of word (BoW) features, which are then used to build a regression model to estimate the building height over the study area. We obtained the error MAE = 3.5m and RMSE = 33.6 for our height regression model. The results achieved show the ability to apply satellite images to extract fundamental building topologies such as building boundary and building height to initially build a LoD1 3D city model.