Conditional Random Field models for road network extraction

The aim of this work is to extract the road network from aerial images. What makes the problem challenging is the complex structure of the prior: roads form a connect network of smooth, thin segments which meet at junctions and crossings. This type of a-priori knowledge is more difficult to turn into a tractable model than standard smoothness or co-occurrence assumption. We develop a novel CRF formulation for road labeling, in which the prior is representated by higher-order cliques that connect sets of superpixels along straight line segments.

Representative Publications

Mind the gap: modeling local and global context in (road) networks
J. Montoya, J.D. Wegner, L. Ladicky, K. Schindler.
German Conference on Pattern Recognition (GCPR), Münster, Germany, 2014

A higher-order CRF model for road network extraction
J.D. Wegner, J. Montoya, K. Schindler
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2013

Road networks as collections of minimum cost paths
Wegner, J.D., Montoya, J., Schindler, K.
ISPRS Journal of Photogrammetry and Remote Sensing, vol. 108, 2015, pp. 128-137

On the evaluation of higher-order cliques for road network extraction
Montoya, J., Wegner, J.D., Ladický, L., Schindler, K.
Joint Urban Remote sensing Event (JURSE), Lausanne, Switzerland, 2015

Semantic segmentation of aerial images in urban areas with class-specific higher-order cliques
Montoya, J., Wegner, J.D., Ladický, L., Schindler, K.
ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. II(3/W4), 2015, pp. 127-133. PIA15+HRIGI15 best paper award

Contact person: Jan Dirk Wegner and Javier Montoya