Vector One | Lidar Building Detection – Combined Point Clouds and Rasters

Search

Aerial Sensing

Lidar Building Detection – Combined Point Clouds and Rasters

By Vector One ⋅ July 13, 2009 ⋅ Post a comment

Last week in Salzburg at the GI Forum, Werner Mucke and colleagues at the Vienna University of Technology gave a presentation on the detection of buildings using lidar and raster data acquired from imagery. The development of 3D city models requires that work be done rapidly and the trend is toward automated techniques. To achieve this, this research group developed what they term an “Echo Ratio” approach.

To achieve the highest accuracies, airborne imaging is used together with lidar. This also recognises that buildings are sometimes covered with vegetation or objects which interferes with imagery, but not lidar.

The Echo Ratio is explained as sample areas which are inverted tube shaped. Each is sampled for lidar points in 3D and 2D, the ratio then computed, and a roughness determination made for a particular surface type. The roughness is then used for a region classification. The idea here to find candidate regions for growing the building to its edges.

It was interesting to learn that the official survey cadastre often proved inaccurate when determining foundations to compare this work.

The technique is based on 3 echo/ m2 and included 90 million data points. Completeness reached was about 97% and correctness was 73% – signalling a relatively high success rate using the technique.

The software used was developed at the university and is called ‘OPACS’ .

For more information

[Comment update July 16 - I would like to make a comment on your blog entry. We do not use airborne image data in our approach. We solely use the lidar point cloud and derive an nDSM and the echo ratio raster. These are our only input data for the outline detection. The object based classification process is carried out using a roughness parameter also derived directly from the lidar data.]