ARTICLE IN PRESS

OBJECT-BASED IMAGE ANALYSIS (OBIA) ON HYPERSPECTRAL IMAGERY FROM DRONE FOR Ganoderma BASAL STEM ROT DISEASE DETECTION IN OIL PALM

MOHAMAD ANUAR IZZUDDIN1; AROF HAMZAH2*; MOHD NOOR NISFARIZA3 and ABU SEMAN IDRIS1

DOI: https://doi.org/10.21894/jopr.2024.0025
Received: 1 August 2023   Accepted: 28 December 2023   Published Online: 1 April 2024
ABSTRACT

Ganoderma basal stem rot (BSR) disease infection in oil palm causes significant yield loss to the industry. An efficient disease control application is practical during early and moderate infection of Ganoderma in oil palm. Early detection of disease caused by Ganoderma has been conducted in the lab and on the ground manually. This approach requires inspection and sampling of each oil palm in the field, which is laborious, time-consuming, and costly. Airborne detection using a drone provides a good alternative for fast Ganoderma BSR disease detection in oil palms. This study uses hyperspectral images obtained from a camera mounted on a drone and Object-based Image Analysis (OBIA) to detect Ganoderma infection in oil palms and classify the Ganoderma BSR Disease Severity Index (GDSI). In OBIA, segmentation parameters, such as the Edge were set to 30 while Merge was set to 70 to demarcate respective oil palm without overlaps. Two classifiers namely 1) Support Vector Machine (SVM) and 2) K-Nearest Neighbour (KNN) were used to classify the segmented samples for each GDSI. The results demonstrate that the SVM classifier provides a good classification of the raster image with an overall accuracy of 92.5%. The study shows that hyperspectral images captured by drones provide a viable technique for detecting Ganoderma infection and classifying its severity.

KEYWORDS:


1 Malaysian Palm Oil Board (MPOB),
6 Persiaran Institusi, Bandar Baru Bangi,
43000 Kajang, Selangor, Malaysia.

2 Department of Electrical Engineering,
Faculty of Engineering, University of Malaya,
50603 Kuala Lumpur, Malaysia.

3 Department of Geography,
Faculty of Arts and Social Sciences,
University of Malaya,
50603 Kuala Lumpur, Malaysia.

* Corresponding author e-mail: ahamzah@um.edu.my