Classification for non infected and infected ganoderma boninense of oil palm trees using ALOS PALSAR-2 backscattering coefficient

Abstract

Malaysia’s monthly export of oil palm product in 2015 was 25,370,294 tonnes valued at about RM60 million. Consequently, Malaysia is now one of the leading manufacturers and exporters of palm oil and its derivatives in the world. However, oil palm plantations in Malaysia are now facing the threat of a Basal Stem Rot (BSR) disease that is caused by fungus called the Ganoderma boninense. This disease reduces oil palm production as an infected mature oil palm dies after 2-3 years of being infected. A decision tree classification approach is proposed in this study for discriminating between non infected and infected of G. boninense in oil palm tree using backscatter values of ALOS PALSAR 2 in FELCRA Seberang Perak 10, Kampung Gajah, Perak. The methodology involves (1) collection of ALOS PALSAR 2 image which include dual polarization HH (Horizontal - transmit and Horizontal - receive) and HV (Horizontal - transmit and Vertical - receive); (2) infection status of the oil palm trees in the study area that comprise 92 trees; and (3) image pre-processing that includes radiometric calibration, speckle filtering and linear conversion to dB. The final stage is the backscatter classification of G. boninense health status using the Decision Tree classifier. The overall accuracy for HH and HV backscatter classification were 45.65% and 56.52% respectively. Further investigations may need to be carried out to improve existing accuracy.