Comparison of object-based image classification of worldView-2 and small format aerial photography images for vegetation mapping

High-spatial resolution remote sensing images play an important role in discriminating and mapping vegetation and non-vegetation objects in coastal areas. The advancement of remote sensing sensors and commercially available unmanned aerial vehicle (UAV) offered opportunities for detailed mapping and monitoring vegetation objects. This study aims to compare the utility of a pan-sharpened WorldView-2 (WV-2, 0.5 m pixel size) and Small Format Aerial Photograph (SFAP, 0.32 m pixel size) images in discriminating and mapping vegetation and non-vegetation objects in Perancak Estuary, Bali, Indonesia. An object-based image classification was selected to perform the classification process. This method is well-suited for high-spatial resolution image data where there is high value of heterogeneity of pixel values within a single object on the image. A multi-resolution segmentation was applied to segment both images into object candidates. To select the most relevant segmentation parameters, a systematic simulation between scale parameters, color, and shape of the segments was performed. In the classification stage, a rule-based classification was applied to classify the object candidates into meaningful land-cover with two classes, including vegetation and non-vegetation. The method emphasizes the use of vegetation index transformation in the visible bands to distinguish vegetation and non-vegetation classes. Vegetation index transformation includes Green Red Vegetation Index (GRVI), Kawashima Index (IKAW), Red Green Ratio Index (RGRI), Visible Atmospheric Resistance Index (VARI), and Green Leaf Index (GLI). The accuracy of object-based image classification mapping used visual interpretation. The results of this study show that optimum rule-based segmentation and classification parameters depend on the image used for classification, and Green Leaf Index (GLI) produced highest vegetation class accuracy for both WV-2 and SFAP images. © 2021 Elsevier B.V., All rights reserved.