Image Mining in Remote Sensing for Coastal Wetlands Mapping: from Pixel Based to Object Based Approach

2nd International Conference of Indonesian Society for Remote Sensing 2016

地球科学;计算机科学

Farda, N.M.^1 ; Danoedoro, P.^1 ; Hartono^1 ; Harjoko, A.^2

Faculty of Geography, Universitas Gadjah Mada, Sekip Utara Bulaksumur, Yogyakarta

55281, Indonesia^1

Faculty of Mathematics and Natural Sciences, Universitas Gadjah Mada, Sekip Utara Bulaksumur, Yogyakarta

55281, Indonesia^2

关键词: Back propagation neural networks;    Bare soil indices;    Ecosystem services;    Mean-shift segmentation;    Multi-temporal image;    Remote sensing images;    Vegetation index;    Visible and near infrared;   

Others  :  https://iopscience.iop.org/article/10.1088/1755-1315/47/1/012002/pdf
DOI  :  10.1088/1755-1315/47/1/012002

学科分类：计算机科学（综合）

来源: IOP

【 摘 要 】

The availably of remote sensing image data is numerous now, and with a large amount of data it makes "knowledge gap" in extraction of selected information, especially coastal wetlands. Coastal wetlands provide ecosystem services essential to people and the environment. The aim of this research is to extract coastal wetlands information from satellite data using pixel based and object based image mining approach. Landsat MSS, Landsat 5 TM, Landsat 7 ETM+, and Landsat 8 OLI images located in Segara Anakan lagoon are selected to represent data at various multi temporal images. The input for image mining are visible and near infrared bands, PCA band, invers PCA bands, mean shift segmentation bands, bare soil index, vegetation index, wetness index, elevation from SRTM and ASTER GDEM, and GLCM (Harralick) or variability texture. There is three methods were applied to extract coastal wetlands using image mining: pixel based - Decision Tree C4.5, pixel based - Back Propagation Neural Network, and object based - Mean Shift segmentation and Decision Tree C4.5. The results show that remote sensing image mining can be used to map coastal wetlands ecosystem. Decision Tree C4.5 can be mapped with highest accuracy (0.75 overall kappa). The availability of remote sensing image mining for mapping coastal wetlands is very important to provide better understanding about their spatiotemporal coastal wetlands dynamics distribution.