Multispectral and LiDAR data fusion for fuel type mapping using Support Vector Machine and decision rules

摘要: This paper presents a method for mapping fuel types using LiDAR and multispectral data. A two-phase classification is proposed to discriminate the classes of Prometheus system, which adapted ecological characteristics European Mediterranean basin. The first step mapped main groups, namely grass, shrub tree, as well non-fuel classes. phase was carried out Support Vector Machine (SVM) combining overall accuracy this 92.8% with kappa coefficient 0.9. second focused on discriminating additional categories based vertical information provided by measurements. Decision rules were applied output SVM mean height returns distribution fuels, described relative point density in different intervals. final type yielded an 88.24% 0.86. Some confusion observed between 7 (dense tree cover presenting continuity understory vegetation) 5 (trees less than 30% cover) some areas covered Holm oak, showed low pulses penetration so that vegetation not correctly sampled.