An object-based image analysis in QGIS for image classification and assessment of coastal spatial planning

Abstract

In practice, urban and regional planners often use a pixel-based method for image classification. Unfortunately, it produces lower accuracy than an Object-Based Image Analysis (OBIA) method, especially for the high-resolution images. To assess spatial planning, scholars rarely used the OBIA method in open-source software. This paper aims to develop a method for classifying land cover and assessing coastal spatial planning. We used Sentinel-2A in 2015 and 2020 as the basic data. For image classification, we used the OBIA method in Quantum GIS (QGIS) 3.10.6 and Orfeo ToolBox 7.1.0. Furthermore, we used Artificial Neural Network (ANN) and Cellular Automata (CA) algorithms in QGIS 2.18.20 for projecting future land cover change, and then used the projected land cover map to assess the spatial planning in 2031. The results show that the OBIA method is useful for image classification, achieving 94.50 and 90.98 percent of the overall accuracy for the imageries in 2015 and 2020, respectively. Our coastal spatial planning assessment shows that the plan has not considered adequately the rapid land cover change of the region, especially the increase in waterbodies. We advocate that the local government should consider this issue when evaluating the spatial planning. The methodology using an open-source software such as QGIS in a developing country context also provides a promising exemplar that other local governments can use for assessing their spatial planning.