I. Introduction
Accurate detection of road maps is necessary for efficient navigation systems, autonomous driving [1] [2], urban planning, disaster management, etc. In the modern world, road detection [3], [4] is necessary to generate an efficient navigation system. In Kerala alone, 30 new highways of about 549 km are targeted for completion by 2024. These numbers are the highways alone, discarding the local routes. So this shows how fast the routes are newly added or changed within a short time span. Manually updating an existing digital data related with geographical features is truly inefficient. To overcome these limitations, there are several types of road detection techniques that differ between their performance. Present satellite images can have high spatial resolutions (up to 50 cm per pixel). From satellite images that have a spatial resolution of 10cm per pixel, roads can be easily detected by a normal human eye. So with this accuracy, roads can be annotated manually that could cover larger distances within a small spatial resolution.