High-resolution spatial images are the main data source in photo-interpretation projects that produce official land cover and vegetation maps. High resolution aerial imagery – highly useful in visual analysis- is excessive for pixel-based analysis, when our goal is the differentiation of objects defined by limits. This is due to the fact that the huge spectral variability of cover or because of the few spectral difference between contiguous objects. In addition, beyond the pixel-based classification, there are qualities that can define space objects better: scale, texture, shape, and context. In this paper, we propose an object-based classification to detect physiognomic-ecological units using aerial orthophotographic imagery from the National Plan for Aerial Orthophotography (PNOA). These images have four bands (visible and near infra-red) with a 50 cm spatial resolution. We performed a multiscale image segmentation by analysing spectral and spatial components. We defined a set of classification rules to calibrate the components of the objects. We used field data and validated the results with regions of interest. The maps derived allow the differentiation of physiognomic-ecological units with a great level of detail, suitable for general studies of vegetation, landscape ecology and erosion models.

Póster presentado en el XVI Congreso de la Asociación Española de Teledetección, celebrado en Sevilla del 21 al 23 de octubre de 2015.

Peer reviewed