The Use of Voting Strategy for Building Extraction from High Resolution Satellite Images

초록

This paper proposes the use of voting strategy for extracting buildings from high resolution satellite images. Previously, the grouping strategy has been proposed and widely used for extraction of man-made features from images. In order to apply grouping we need to extract one complete line per each building side. However, this requirement may not be met for satellite images such as IKONOS images due to their spatial resolution. Often a long side of a building produces several lines with their orientation not necessarily identical. Lines from short side are often missing. In this situation we proposed to use a voting strategy. We vote line elements within a small region of interest for finding line position and orientation. The orientation and position are refined by a least squares matching process. We assessed the performance of our algorithm against an IKONOS image. Our algorithm extracted building lines from over 83% of buildings tested with an average angular accuracy of 2 degrees and average positional accuracy of 1.2 pixels. This promising result supports the use of voting strategy as proposed in this paper.