Date of Publication :22nd February 2018

Abstract: This paper focuses on an automatic algorithm for fast building boundary extraction from satellite imagery and it presents an experimental comparison of the bilateral filter (BF) and adaptive bilateral filter (ABF). The study and experimental result prove that outcomes of ABF are much better than outcomes of BF. ABF produces a more promising result than BF. Old and conventional building boundary extraction models are very complex and time-consuming. The proposed procedure of building boundary extraction consists of three main stages: (1) Edge-preserving and smoothing by using adaptive bilateral filter, (2)detection of a line segment using ED Line algorithm, (3) identification of polygonal building boundary by using perceptual grouping technique. Our proposed algorithm is tested on HR (high resolution) Quick Bird satellite images and obtained results are promising and nearly real-time. Hence the experimental results are useful enough with an overall accuracy of 88.24%, which is accurate enough for further image understanding building boundary detection and identification of the target in real time environment and can contribute to solving the problem of identification of unauthorized and illegal building construction on its early stage

Reference :