Share Email Print

Proceedings Paper

Semiautomatic road extraction as a model driven optimization procedure
Author(s): Armin Gruen; Haihong Li
Format Member Price Non-Member Price
PDF $17.00 $21.00

Paper Abstract

In this paper we propose a model driven optimization framework for semi-automatic road extraction from digital images. Semi-automatic means that a road is extracted automatically after some seed points have been given coarsely by the human operator, through activation of a mouse using a convenient interactive image-graphics user interface. In the model driven optimization framework, a road is represented by a generic road model that specifies both photometric and geometric constraints and defines an objective function which embodies a notion of the 'best road segment'. Then the problem of road extraction is treated as one of evaluating the objective function and generating the optimal fit of the model to the image data. Two different techniques, based on dynamic programming and least squares principles respectively, are discussed in the paper. With dynamic programming, the optimization problem is set up as a discrete multistage decision process and is solved by a 'time-delayed' algorithm. It ensures global optimality, is numerically stable and allows for hard constraints to be enforced on the solution. In the least squares approach, we combined three types of observation equations, representing the photometric part of the road model, the geometric part (modeled by B-spline) and the boundary constraints defined by operator-given seed points. The solution is obtained by solving a pair of independent normal equations to estimate the parameters of the spline. Therefore this new snake concept is called 'LSB-snakes' (least squares B-spline snakes). The issues related to the mathematical modeling and the practical implementation of both methods are discussed and experimental results of the different approaches are shown.

Paper Details

Date Published: 1 December 1995
PDF: 12 pages
Proc. SPIE 2646, Digital Photogrammetry and Remote Sensing '95, (1 December 1995); doi: 10.1117/12.227864
Show Author Affiliations
Armin Gruen, Swiss Federal Institute of Technology (Switzerland)
Haihong Li, Swiss Federal Institute of Technology (Switzerland)

Published in SPIE Proceedings Vol. 2646:
Digital Photogrammetry and Remote Sensing '95
Eugeny A. Fedosov, Editor(s)

© SPIE. Terms of Use
Back to Top