Identification of shadows in the Atlantic Forest for preservation monitoring

Abstract

The presence of shadows in aerial images of forest areas can be useful for several tasks, such as detection of emergent layer specimens, determination of bamboo proliferation or evaluation of other forest structural parameters. Invariant features in images have special value since the process of capture of images involves considerable levels of variation. The aim of the present study was to evaluate an algorithm based on an illumination invariant shadow ratio to segment RGB aerial images of the Atlantic Forest in shadowed and non-shadowed areas. Some parameters of the algorithm were tuned with satisfactory results regarding the success rate comparing with manual selection, and thus allowed obtaining a procedure to achieve an automatic process. On the other hand, the processing time was reasonably short, making this methodology a suitable option for rainforest conservation monitoring.