Exploring the Potential of High Resolution Remote Sensing Data for Mapping Vegetation and the Age Groups of Oil Palm Plantation

8th International Symposium of the Digital Earth

地球科学;计算机科学

Kamiran, N.^1 ; Sarker, M.L.R.^1,2

Department of Geoinformation, Universiti Teknologi Malaysia, Malaysia^1

Department of Geography and Environmental Studies, University of Rajshahi, Bangladesh^2

关键词: Classification accuracy;    Classification algorithm;    Grey-level co-occurrence matrixes;    High resolution remote sensing;    Image processing technique;    Oil palm plantations;    Sustainable production;    Textural information;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/18/1/012181/pdf
DOI  :  10.1088/1755-1315/18/1/012181

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

The land use/land cover transformation in Malaysia is enormous due to palm oil plantation which has provided huge economical benefits but also created a huge concern for carbon emission and biodiversity. Accurate information about oil palm plantation and the age of plantation is important for a sustainable production, estimation of carbon storage capacity, biodiversity and the climate model. However, the problem is that this information cannot be extracted easily due to the spectral signature for forest and age group of palm oil plantations is similar. Therefore, a noble approach "multi-scale and multi-texture algorithms" was used for mapping vegetation and different age groups of palm oil plantation using a high resolution panchromatic image (WorldView-1) considering the fact that pan imagery has a potential for more detailed and accurate mapping with an effective image processing technique. Seven texture algorithms of second-order Grey Level Co-occurrence Matrix (GLCM) with different scales (from 3×3 to 39×39) were used for texture generation. All texture parameters were classified step by step using a robust classifier "Artificial Neural Network (ANN)". Results indicate that single spectral band was unable to provide good result (overall accuracy = 34.92%), while higher overall classification accuracies (73.48%, 84.76% and 93.18%) were obtained when textural information from multi-scale and multi-texture approach were used in the classification algorithm.