Conférence : Semi-automatic extraction of 3D buildings from aerial imagery

In the last few years, there has been significant progress in sensor technologies for surveying applications. These include digital sensor systems for acquiring geo-referenced aerial images, such as commercial large format cameras with integrated GPS/IMU. It is therefore relatively simple to quickly acquire huge amounts of raw geo-spatial data: the problem is now in quickly interpreting them. A particularly difficult problem is the extraction of high-level information from the imagery, such as deriving buildings in the form of 3D vectors. Using current technologies (e.g. stereoplotters or CAD tools), the extraction of buildings still has to be manually performed in most cases. As a result, generation of high-level information from imagery is labour intensive. This presentation will describe a new technology for the semi-automatic extraction of 3D buildings from aerial imagery. The approach consists of first generating a digital surface model (DSM) through autocorrelation using stereo images. Then, building contours are automatically extracted using the DSM from an approximation manually inputted by the user. Once the contours have been produced, the user visually defines the approximate geometry of the building (e.g. vertices and breaklines) by tracing 2D vectors over the DSM. While the user traces these vectors, the software assists the data collection process by automatically refining the geometry and associating 3D coordinates to each vertex in real-time. Incorporated in SimActive’s Correlator3D™ photogrammetry software, the technology leads to a highly simplified process removing the need for highly-trained personnel. 3D buildings can therefore be extracted in minimum time and with limited user input. A live demonstration of the new technology will be provided and sample results shown and discussed.

Conférencier : Philippe Simard
Titre Président
Organisme SimActive Inc.
Courriel psimard@simactive.com
Profil Philippe Simard has been acting as the President of SimActive since its creation in 2003. Under his management, the company rapidly positioned itself as a leader in geospatial solutions. Before SimActive, Dr Simard was involved in numerous projects with world-class organizations such as the Department of National Defence and NASA. He holds a doctorate in electrical engineering from McGill University, where he did his research in computer vision. In 2005, he won the prestigious Young Innovator Award from the Networks of Centres of Excellence which honors top Canadian researchers whose work benefits society.