A marked point process of rectangles and segments for automatic analysis of digital elevation models

Area of Science:

• Computer Vision
• Geographic Information Science
• Computational Geometry

Background:

• Automatic feature extraction is crucial for image analysis.
• Existing methods often struggle with incorporating prior knowledge about feature distribution.
• Digital Elevation Models (DEMs) require robust methods for analyzing urban altimetry.

Purpose of the Study:

• To present a framework for automatic feature extraction from images using stochastic geometry.
• To incorporate a priori knowledge on the spatial distribution of features.
• To apply the framework to the analysis of Digital Elevation Models (DEMs).

Main Methods:

• Modeling image features as realizations of spatial point processes (segments and rectangles).
• Defining an energy function to favor object connections and alignments.
• Estimating features by minimizing the energy using simulated annealing.

Main Results:

• Demonstrated successful application to Digital Elevation Models (DEMs) of dense urban areas.
• Showcased the framework's ability to detect linear discontinuities and segment homogeneous areas.
• Presented results on low-quality DEM data from the IGN (French National Geographic Institute).

Conclusions:

• The proposed stochastic geometry framework effectively extracts features from images, particularly DEMs.
• The model's flexibility allows integration of prior spatial information for improved accuracy.
• This approach offers a robust solution for analyzing complex altimetric data.