Extraction of line-networks in digital terrain models and image data
Many objects in digital terrain models and image data can be described by line-networks. Examples are given by roads and rivers in remote sensing data or vessels in medical images. The extraction of these networks is of great importance for several tasks such as the updating of road maps, flood simulations or automatic diagnostic processes.
The goal of this project is the automatic detection of line-networks in raster data. For that purpose, a model-based probabilistic approach is developed based on the method of marked point processes. Within this approach, a graph is introduced as object model. The point process is coupled with the sampling approach of Reversible Jump Markov Chain Monte Carlo sampling. In this way, several graph configurations can be constructed with the goal to find the configuration which fits the input data and prior knowledge about the configuration best.
The model is developed for the extraction of river networks in digital terrain models. In addition, applications in object detection of image data are investigated.
Fig. 1: In the sampling process the graph describing the line-network is iteratively constructed.
Fig. 2: With the developed approach river networks in digital terrain models (left) and networks in images (middle and right) are detected.
Schmidt, A.; Rottensteiner, F.; Soergel, U.; Heipke, C. (2014): Extraction of fluvial networks in lidar data using marked point processes. In: IntArchPhRS vol. XL-3, ISPRS Technical Commission III Symposium, Zurich, Switzerland, September 2014, pp. 297-304 | file |
Schmidt, A.; Rottensteiner, F.; Soergel, U.; Heipke, C. (2015): A Graph Based Model for the Detection of Tidal Channels using Marked Point Processes. In: The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XL-3/W3, pp. 115-121, 2015 | file |
Schmidt, A.; Kruse, C.; Rottensteiner, F.; Soergel, U.; Heipke, C. (2016): Network detection in raster data using marked point processes. In: Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci. XLI-B3, pp. 701-708