Bahrom Ezzati

Date of Award

1-1-2018

Thesis Type

masters

Document Type

Thesis

Divisions

inst1

Department

Institute of Graduate Studies

Institution

University of Malaya

Abstract

Ganoderma disease in oil palm caused by Ganoderma spp. fungi have caused significant losses of Malaysia's economic income. Advances in remote sensed imagery and image processing using unmanned aerial vehicle (UAV) for Ganoderma disease detection could be developed to reduce operating cost and time as well as cover wider oil palm areas. This study examines the performance of red-green-blue (RGB) and near- infrared (NIR) digital orthophoto image acquired using modified digital cameras mounted on the UAV for aerial detection of Ganoderma disease in oil palm. The orthophoto images were filtered using eight adaptive filters with window sizes of 7×7, 9×9 and 11×11. The filtered orthophoto images then were processed using three supervised image classifiers: Maximum Likelihood (ML), Mahalanobis Distance (MD) and Neural Net (NN). The classifiers were used to classify the Ganoderma disease severities into Experiment 1: T0 (healthy), T1 (mild), T2 (moderate) and T3 (severe); and Experiment 2: healthy and unhealthy. The classification outputs were assessed using a confusion matrix. Best result was obtained from Bit Error filter with 9×9 window size using the NN algorithm with an overall accuracy of 62.41% and a Kappa coefficient of 0.3890. This study demonstrated classification from UAV-based imagery can be improved using filters for Ganoderma disease detection mapping in oil palm plantation.

Note

Dissertation (M.A.) – Institute of Graduate Studies, University of Malaya, 2018.

Download

Share

COinS