An Approach Using Imbalanced Data Classification for Accurate Detection of Roads

Abstract: Reliably extracting information from satellite and aerial imagery is a difficult problem with many practical applications. Extraction of objects such roads, buildings etc from high resolution satellite imagery is an important task in urban planning, military applications etc. This is a difficult task because of occlusions, shadows, and other non-road objects. This paper proposes a method for road detection. It also highlights the importance of the imbalanced data classification in detecting the road in a complex scenario. First, the spectral angle of multi spectral input image is calculated and then dog filtering is applied. Then this image is subjected to morphological operations and area based filtering. At this stage, there may be some noises like buildings, rivers etc. To remove these noises, imbalanced data classification can be used.

Keywords: DOG filtering, spectral angle, Morphological operations, imbalanced data classification