A semi automatic off-roads and trails extraction method from Sentinel-1 data

Abstract

In this paper we propose a semi-automatic off-roads and trails extraction method using synthetic aperture radar (SAR) data from Sentinel-1 satellite. In fact, these categories of rural and desert roads, very solicited on the Tunisian borders by smugglers, are often changing and need more close monitoring. Thus, we propose here to take benefit from free SAR Sentinel-1 data with high resolution and repetitive coverage. Taking into account the particularities of these roads, having curvilinear features, and those of the data, degraded with granular noise that inherently exists in SAR, we propose a more general approach based on anisotropic diffusion and dynamic snakes adapted to the case of off-roads and trails. Application of the developed approach to the Sentinel 1 SAR data acquired over the Tunisian frontier lead to the extraction of the appropriate road edges.