Real time visualisation of 3D city models in street view based on visual salience

• Street level visualization is an important application of the 3D city models. Challenges in the street level visualization are the cluttering of the detailed buildings and the performance. In this paper, a novel method for street level visualization based on visual salience evaluation is proposed. The basic idea of the method is to preserve these salient buildings in a view and remove the non-salient ones. The method is composed by pre-process and real-timevisualization. The pre-process starts by converting 3D building models in higher Levels of Detail (LoDs) into LoD1 with simplified ground plan. Then a number of index view points are created along the streets; these indexes refer both to the positions and the direction of the sights. A visual salience value is computed for each visible simplified building in respective index. The salience of the visible building is calculated based on the visual difference of the original and generalized models. We propose and evaluate three methods for visual salience: local difference, global difference and minimum projection area. The real-time visualization process starts by mapping the observer to its closest indexes. Then the street view is generated based on the building information stored in theindexes. A user study shows that the local visual salience gives better result than the global and area, and the proposed method can reduce the number of loaded building by 90% while still preserve the visual similarity with the original models.

Ämnesord

NATURVETENSKAP  -- Data- och informationsvetenskap -- Annan data- och informationsvetenskap (hsv//swe)

NATURAL SCIENCES  -- Computer and Information Sciences -- Other Computer and Information Science (hsv//eng)

Nyckelord

3D city models

street level visualization

selection

visual salience

Publikations- och innehållstyp

ref (ämneskategori)

art (ämneskategori)