Retrieving of urban morphology from satellite images and machine learning
Project Description

Urban information is very important in a lot of scientific research areas, for example, meteorological modeling, air quality modeling, energy consumption studies, and city planning. One of the key urban information is the dimensions of buildings. However, that type of information is not easy to acquire and is normally expensive. Various studies indicate that there is a high correlation between building dimensions (building height and area coverage) and the satellite images' signal in different bands. While many open-source satellite images and digital elevation model data are available, their resolutions are normally too low to deduce the detailed building structure. However, with the advance in online map engines and street view images(e.g., Baidu, Google, and Open street map), high-resolution RGB images provide the possibility of more accurate building identification through image segmentation and object recognition methods. In this project, a combination of different open-source satellite images and Google or other online map APIs would be used as the input in machine learning algorithms to retrieve the building dimensions in the Greater Bay Area.

Supervisor
FUNG Jimmy Chi Hung
Quota
3
Course type
UROP1000
UROP1100
UROP2100
Applicant's Roles

The student needs to learn various machine learning techniques

Applicant's Learning Objectives

The outcome of this project is expected to be very useful to provide better open-source urban information for modeling studies in the research community worldwide. It is also scalable to other regions in the world, thus highly facilitating the global scale climate modeling studies.

Complexity of the project
Moderate