Détection simultanée de l'ombre et de la végétation sur des images aériennes couleur en haute résolution

We propose a new method to simultaneously detect shadows and vegetation in remote sensing images. Several shadow and vegetation indices are merged with the Dempster-Shafer (DS) theory so as to increase the reliability and accuracy of the segmentation. The per-pixel DS fusion is incorporated in a Markovian context to obtain an optimal and regularized segmentation result. Evaluation on aerial images clearly shows that this new method is robust and accurate.