Analysis of Rubber Tree Recognition Based on Drone Images

Abstract

Hevea brasiliensis, the rubber tree is a tree belonging to the family Euphorbiaceae. It is the most economically important member of the genus Hevea because the milky latex extracted from the tree is the primary source of natural rubber. Rubber trees are important economic crops in tropical areas. Accurate knowledge of the number of rubber trees in a plantation area is important to predict the yield of rubber trees, manage the growing situation of the rubber trees and maximize their productivity. The aim of this study is to analyse the detection of rubber tree based on drone images. In order to solve this issue, this study proposes a novel automated method for processing and analysing the individual rubber tree using images from drone. A drone of Parrot ANAFI was used to acquire aerial images and trees in surveyed area. In order to develop the database of rubber tree based on drone images, the position of rubber tree and non-rubber tree was marked on selected three of drone images. A total number of 154 rubber tree and 149 non-rubber tree have been divided into training and testing set. Next, the features of rubber tree and non-rubber tree will be extracted by three selected algorithms of image processing. The algorithms used in this research study including Gray-Level Co-occurrence Matrix (GLCM), Wavelet Transform and Template matching. The database are tested based on the three algorithm by using Support Vector Machine (SVM) to see which algorithms gives highest accuracy and will be used for detection and enumeration of individual rubber trees using images from unmanned aerial vehicles (UAVs). The result of the analysis show Gray-Level Co-occurrence Matrix (GLCM) has the highest accuracy for window size 150 x 150 pixel which is 83.56%. This show that drone images have high potential for rubber tree recognition despite the difficulty of detecting the crown of rubber tree during marking process.