Agricultural-field extraction on aerial images by region competition algorithm

The problem of segmenting agricultural fields in aerial images is still a manual work in most geographic information system requiring repetitive, tedious and time-consuming human work. Here, we address the problem of semiautomatic segmenting agricultural fields by region competition technique that integrates region growing and deformable models. The deformable model dynamically adapts its contour, analyzing homogeneous parcels in an energy-minimizing framework. To assure the optimal image segmentation and practical applicability of the approach, we study different aspects: parameterization, convergence criteria and user interaction. The successful results obtained have allowed introducing the region competition technique in a teledetection environment.