Oil palm age classification in Ladang Tereh Selatan,Johor,Malaysia using remote sensing technique

Abstract

Determining and classifying the age of oil palm is important in predicting oil palm yield, planning replanting activities and oil palm age is also an important criteria in estimating the carbon sequestration and storage potential of oil palm trees.. Nevertheless, determining its age with conventional method is costly, time consuming and tedious process. Alternatively remote sensing methods are used with only a moderate success. Previous studies using remote sensing have shown limitations to classify more than five age classes of oil palm trees. This studyused SPOT-5 multispectral image to classify 12 different age classes of oil palm trees at Ladang Tereh Selatan, Kluang, Malaysia. . Three different classifiers namely Support Vector Machine (SVM), Artificial Neural Network (ANN), and Maximum Likelihood Classifier (MLC) were employed and it was found that all these techniques that rely on spectral information from the image could only classify the ages with low overall accuracy of 32.46%, 29.92% and 37.41% respectively. In order to improve the classification, Grey-Level Co-occurrence Matrix (GLCM) texture measurement was added into the MLC classifier. Various combinations of textures and window sizes were tested in order to find the optimum texture combination. The overall accuracy of the classification was improved to 89.6% with the incorporation of eight texture combinations with 39 × 39 window. This study also found that, window size is more important than the type of texture in determining the stand age of the palm trees, where all the window sizes were statistically significant at 95% confidence level. The method used in this study should be extended to other plantations to test the applicability of the technique in classifying more age classes.