Plant canopy classification methodology based on hyperspectral data cube on a polluted mining site

Canopy analysis was carried out in order to classify the differences between vegetation types at the Szárazvölgy flotation sludge reservoir. Supervised classification methods were used to distinguish 8 vegetation types based on the spectral properties of the area: forest (Quercus sp.), young deciduous forest, reed (Phragmites sp.) and aquatic plants, false indigo (Amorpha fruticosa), Australian pine (Pinus nigra), shrub – mainly sloe (Prunus spinosa) and dog rose (Rosa silvestre), blackberry (Rubus caesius), low biomass. The results of the classifications were compared to a ground truth image in order to know the best process for classification. The ground truth image is based on ortophoto, topographic map, and GPS based field data collection. According to results of the comparison, the parallelepiped classification method is proved to be appropriate method based on the overall accuracy of canopy classification, which was 54 % due to heterogeneity of the vegetation. The results showed that hyperspectral remote sensing is an effective tool for the characterization of canopy and monitoring of canopy changes at the examined polluted sites so that the obtained information can be a valid base for modelling soil degradation and erosion.