Analysis of Oil Palm Tree Recognition using Drone-Based Remote Sensing Images

Abstract

The oil palm tree, or scientifically called as Elaeis guineensis is native to West Africa, where it grows in the wild, transformed into a crop that later was introduced to Malaysian industry. The cultivation of oil palm improved rapidly under the agricultural sector causes degradation, particularly when the oil palm plantation goes uncontrolled. Tree plantation identification is very important for plantation management, environmental management, biodiversity monitoring and many other applications. Accurate inventories and monitoring oil palm estates can be a challenge and critical towards the plantation management and plant area expansion. Managing oil palm estate manually can almost be impossible, so do the tree counting. Manual field-based tree counting is time-consuming and high cost. Conventional method for tree counting can be carried out by manually marked on images or carry out field surveying using GPS to collect the positions of oil palm trees and display their position on image. Developing easier, simpler and cheaper method for tree counting is needed. The aim of this study is to analyse oil palm trees using drone-based remote sensing images. The algorithms used in this research study including Gray-Level Co-occurrence Matrix (GLCM), wavelet transform and template matching. The database of oil palm tree been developed with a total of 131 oil palm trees and 161 of non-oil palm trees have been collected. The window size of oil palm tree been analysed where 250 x 250 pixels which GLCM showed the best overall accuracy of 73.10% for both oil palm and non-oil palm. In this specific window, the oil palm crown can be covered and the result given is more accurate compared to other window sizes. The resulting analysis shows that wavelet transform algorithm gives the highest overall accuracy value which is 82.07%. The other eight statistic parameters can also used to modify the GLCM in order to observe the accuracy and identify which give the best classification accuracy. The availability and ubiquity of drone technologies with high resolution images and regular basis monitoring, new techniques in image and pattern recognition using drone-based remote sensing images let the idea of high accuracy oil palm tree detection become a reality.