Automatic extractions of road intersections from satellite imagery in urban areas

Information of the urban road areas for resource management, security monitoring, urban development and Geographic Information System (GIS) is changing with the growing world. Satellite image provides useful data that is extracted from satellite image of the urban area. Automatic extraction of the road intersections from the urban areas has been a challenging topic because the high resolution satellite images contain multiple layers that represent roads, buildings, and other high density objects. Our goals is to automatically separate the road layer from the other layers then extract the road intersections. Usually traditional image processing methods don't achieve satisfied performance in case of satellite images. In this paper we propose a cost effective method for road extraction from Satellites images. In order to find the precise road intersection of urban areas we have divided that whole process into two sequential modules: first, extraction of road line using different Morphological direction filtering to automatically eliminate the other layers from road layer and finally, extraction of road intersections to determine the road orientation and interconnectivity. For our proposed Road intersection extraction, we applied the method to a set of randomly selected satellite image from urban and semi urban areas and achieved over 95.8% precision.

I. Introduction

Geographic Information System (GIS) aggressively being popular day by day due to attractiveness of internet as well as satellite image. Google, Yahoo, Virtual Earth and other maps are examples of exhibit of those satellite images. With the availability of high resolution satellite data and its processing technologies, integration of digital image analysing systems with advance GIS systems permit compositing data sources as well as promoting a partnership between man and machine[7]. Satellite images provide opportunity in many areas like security monitoring, communication industry, urban microclimate and transportation navigation, landscape planning and visualization etc. Road extraction from remotely sensed images has been the purpose of many works in the image processing field and because of its complexity is still a challenging topic [2]. Automation has been considered the most effective way to remove the obstacles of labour intensive manual processes and reduce the cost and shorten the turnaround time of spatial database updating [5]. The traditional road extraction methods have some disadvantages such as the long computational time, the existence of some residual objects in the image which are not classified as roads and the inability to detect roads in all directions [3]. Our proposed methods try to avoid these disadvantages by performing the automatic segmentation and various morphological operations in first steps and detect various intersections aligned with non regular intervals in second steps to detect roads intersection.