Use of Satellite Imagery for Updating Terrain Information
in Topographic Databases
K. D. Parakum Shantha
Institute of Surveying & Mapping
Diyatalawa
Sri Lanka
Tel: 94-57-3016..Fax: 94-57-2004
Abstract
Current terrain-information capturing procedures are mostly based on conventional photogrammetric methods. These procedures are slow and data acquisition is expensive. Developing countries cannot afford the time and cost required to capture terrain-information for updating topographic databases. However, with the increasing development of the remote sensing technology, the satellite data, has wide application potential in terrain-information production.
The main objective of this experiment was to investigation the use of SPOT imagery for updating terrain information in topographic databases. The study covers: Identifying the inherent characteristics of optical satellite images, Investigating a method for extraction of linear features (eg.,roads) in SPOT multispectral images using model based .image analysis technique. The results shows that the feasibility of updating road network at 1 : 50000 scale.
Introduction
The demand for timely and accurate geo-information has been increased for the management of the earth related disciplines. In response to this demand, geo -information are being producing in map form as well as in digital form. In most of the developing countries are now in a position to update the ' existing geo-information for the use of their. development
activities. Hence, the updating of this information should be less cost effective and less time consuming. The use of optical satellite imagery has this facility since its multi temporal and high resolution characteristics. With the use of ' SPOT multi spectral data, this study emphasis to update road network in topographic database. In this context, the study enable an opportunity to develop image understanding techniques to support the automated feature extraction of optical imagery that goes beyond the traditional extraction
Automated Feature Extraction
Automated feature extraction is performed by segmentation techniques. Segmentation is a process of generating image segments in which image elements having the same properties are merge together. According to the basic principle, the segmentation method can be divided into different categories, such as edge based segmentation, region based segmentation and segmentation in measurement space (spectral clustering).
In the edge based segmentation the goal is to find discontinuities in the image while region based segmentation is interested in the inner pixel of homogenous area, not the boundaries. Spectral clustering is based on the spectral values of the image pixel and it implies a grouping of pixel in a multispectral space. These three segmentation method can be considered as conventional procedures for image segmentation.
The new approach of image segmentation is segmentation with model based image analysis. Model based image analysis is based on the analysis of the image model that considers the recognition of objects by assuming a model for how real world features are represented in the image.
