Use of Unmanned Aerial Systems for multispectral survey and tree classification: a test in a park area of northern Italy

In the frame of project FoGLIE (Fruition of Goods and Landscape in Interactive Environment), UAS were used to survey a park area in its less accessible zones, for scenic and stereoscopic videos, 3D modeling and vegetation monitoring. For this last application, specifically, through the acquisition of very high resolution images taken with two UAS-borne compact cameras (RGB and NIR), a DSM of a small vegetated area and the corresponding orthoimages were produced and co-registered. Planimetric and height accuracies in block adjustments and orthophotos are in the range of 0.10 m horizontally and 0.15 m in height. Then, after the derivation of synthetic channels, both unsupervised classification and supervised one were performed in order to test the algorithms' ability to distinguish between different bushes and trees species: some of them were correctly classified by the latter method but misclassifications still remain. The overall accuracy for the unsupervised classification is about 50% while the supervised one yields an overall accuracy around 80%.