Classification of landcover from combined LiDAR and orthophotos using support vector machine

• Main Authors: Pula, Rolando A., Concepcion, Ronnie, Ilagan, Lorena, Tobias, Rogelio Ruzcko
• Format: text
• Published: Animo Repository 2019
• Subjects: Land cover—Classification; Support vector machines; Orthophotography; Digital elevation models
• Online Access: https://animorepository.dlsu.edu.ph/faculty_research/2704
• Institution: De La Salle University

© 2019 IEEE. The study is based on the Landcover classification from combined light detection and ranging (LiDAR) data and orthophotos. Five land classes were extracted namely: barren, build up, low vegetation, mango, and non-agricultural trees. Support vector machine (SVM) was the algorithm used for the classification. Different LiDAR derivatives and orthophoto were used as an input which are intensity, digital terrain model (DTM), digital surface model (DSM), normalized digital surface model (NDSM), and RGB combination of orthophotos. The applied algorithm has 100% accuracy based on the confusion matrix which means that SVM is a good algorithm in classification of landcover from combined LiDAR and orthophotos given that the right LiDAR derivatives were used.