Automatic road extraction from aerial images
 
 Group: Tun-Yu Chiang 4798971 tchiang@stanford.edu
        Yi-Hui Hsieh  4794301 yhsieh@stanford.edu
        Wilson Lau    4796082 wslau@stanford.edu
 
 Goal
 ----
 We propose a system to automatically extract a road system from a high
 resolution aerial image.
 
 
 Motivation
 ----------
 Aerial imagery is one of the standard data sources for the acquisition
 of topographic objects, like roads and buildings for geographic
 information systems (GIS). Although semi-automatic systems exist for
 building road maps from aerial images, the task is labor intensive and
 slow, requiring a human operator to pick out road seeds and to
 prune out errors. An automatic road extraction system would be able to
 do the task quicker and allow more frequent map updates.
 
 
 Approach
 --------
 The flow chart below describes the main stages of our proposed system.
 In each stage we will implement the following:
 (1)Prefiltering: image equalization and smoothing.
 (2)Road detection and tracking: identify candidate road segments and
 track the road based on the candidates.
 (3)Road network reconnection: reconstruct the road network from
 stage(2) based on knowledge about road topology.
 (4)Thinning & pruning: remove the unconnected road segments from the
 road network.
 (5)Evaluation: evaluate the result base on the correctness and the
 completeness of the extracted road network.
 
  Different algorithms will be applied in the
 road detection and tracking stage -- (i) energy minimization (snakes) [1],
 (ii) Kalman filtering [2], (iii) geometric-stochastic model and
 estimation [3]. We hope that by using these methods cooperatively we
 will be able to build a more robust system.
 
 
 
 	 ------------------
 	|    Prefiltering  |
 	 ------------------
 	         |<---------------------
 	         V			|
 	 ------------------		|>
 	|  Road detection  |		|
 	|    & tracking	   |		|
 	 ------------------		|
 	         |		 ---------------
 	         V		|  Evaluation	|
 	 ------------------	 ---------------
 	|  Road network    |		^
 	|  reconnection	   |		|
 	 ------------------		|
 	  	 |			|
 	 	 V			|
 	 -----------------		|
 	|   Thinning &	  |		|
 	|   Prunning	  |		|
 	 -----------------		|
 	 	 |---------------------->
 	 	 V
 	 -----------------
 	|     Result	  |
 	 -----------------
 
 
 
 References
 ----------
 [1] Laptev I., et al (2000). Automatic extraction of roads from aerial images
 based on scale-space and snakes. Machine Vision and Applications, 2000.
 
 [2] Vosselman G., de Kneucht J. (1995). Road tracing by profile matching and
 Kalman filtering. In: Automatic Extraction of Man-Made Objects from
 Aerial and Space Images, Birkhauser Verlag, Basel, Switzerland,
 pp. 265-274.
 
 [3] Barzohar M., Cooper D.B. (1993). Automatic Finding of Main Roads in Aerial
 Images by Using Geometric-Stochastic Models and Estimation. IEEE CVPR
 93, pp. 459-464.