Abstrato

Road Network Extraction from High Resolution Satellite Images

Srilakshmi D, Nazneen, Mustafa B

Images taken from satellites provide us with a host of useful information. One of the most crucial data it provides is about the modern road networks. It can provide information about roads in real-time of places which may not be accessible. This road data can then be used to extract the road network in those places. One of the most common application is to use the extracted road features for guiding Unmanned Vehicles (UMV). Another important application is to generate maps which may be used in areas like GPS navigation. Although an expert cartographer can extract those features manually, it is often time consuming and prone to errors. Here we are introducing a system to manually extract the road network features. The system will take a high resolution satellite image as its input. It will apply a bilinear filter to it for reducing the number of colour pixels. Then a canny edge detection algorithm will be applied which will extract the image features and generate a spatial voting matrix out of it. Using the spatial voting matrix, we can detect the road network pixel with a fair degree of accuracy. To the feature extracted from this step, we will apply a tracking algorithm to further increase the accuracy of the road network extraction. We will look for areas and operations to parallelize in order to improve feature extraction time