Detection of cars in complex urban areas

Abstract

Detection of cars in airborne images of typical urban areas has various applications in several domains, such as surveillance, military and remote sensing. It is a tremendously-challenging problem, mainly because of the significant inter-class similarity among various objects in urban environments. In this paper, a novel framework is introduced that adopts a sliding-window approach and it depicts, in a novel way, the local distribution of gradients, colours and texture. A linear support vector machine classifier is used to differentiate between descriptors that belong to cars and descriptors that belong to other objects in a hyperspace of 3838 dimensions. Descriptors are computed over a newlyproposed adaptive distribution of cells that enables the use of various rotation-variant image descriptors. The proposed framework has been evaluated on the Vaihingen dataset and results corroborate its superiority as it achieves a higher precision for a given recall than the state of the art.