See Research

Semantic segmentation of pole-like road furniture in mobile laser scanning data

Student:Fashuai Li
Timeline:September 2014 - 25 September 2018

In the past years with the technical development of earth observation, 3D virtual models and 3D maps have become popular. Street furniture plays an important role in 3D city models, maps, urban environment analysis and planning. Thus street furniture automatic detection and classification are very necessary and in urgent need. The development of mobile laser scanner provides much denser and more accurate data. The quality of mobile laser scanning data is better than before. There is some knowledge that can be used. Many algorithms proposed in other research fields such as computer vision and machine learning give more technical supports than before.

Most current methods which utilize generic features to classify street furniture are not able to describe street furniture accurately. Shape decomposition is a promising tool that can segment street furniture into meaningful components, which are able to describe these objects accurately. Every street furniture can be decomposed or labelled with basic elements or trained features, which can avoid establishing numerous models for every kind of street furniture.

My research will focus on learning, detecting, decomposing and classifying street furniture. It will start by decomposing street furniture into meaningful components with multi-scale features and part template fitting. The methods and strategies for the use of decomposed components to improve the performance of street furniture detection and classification will be proposed.

Meet the team

F. Li (Fashuai)
Graduate Student M.G. Vosselman (George)
Promotor S.J. Oude Elberink (Sander)
Research theme
Acquisition and quality of geo-spatial information

Developments in sensor and web technology have led to a vast increase in earth observation data. Advanced methodology is needed for interpretation and integration of such big geo-data to support decision making.

Show more