Vineyards mapping using object based analysis

Precise vineyards digital plotting for grape-growing regions can produce thematic maps, which subsequently could be employed within Geographical Information Systems (GIS) as a part of the formation of national Vineyard Information System (VIS) in Turkey. This study proposed vineyards parcels delineation approach using sub-meter Worlview-2 (WV2) VHR satellite images. The WV2 satellite data files incorporate featured 8 (eight) multi-spectral and one panchromatic bands. They were used to differentiate and label the spatial distribution of viticulture practices in Tekirdag Province, Eastern Thrace region, which is one of the areas with highest efficiency for grape-growing practices. The applied classification process employed object-based image analysis method (OBIA, using eCognition). It is a technique that aggregates neighboring pixels into groups acknowledged as image objects. These object primitives convey similar values for several variables according to predefined spatial likelihood and homogeneity measures. Spectral, textural, customized vegetation indices, various band ratios, and other object features were integrated into the object-based image analysis with aim to produce consistent classification results. WV2 pan-sharpened images of 0.5m spatial resolution were taken as the input data. The validation of the created land cover mapping was assessed using formerly produced labeled maps from field work. The plantations with linear, straight row vineyards planting were almost completely mapped, while the allocation accuracies for the other planting types were comparatively lower. Nonetheless, these accuracy results for the dissimilar vineyards planting routines put forward that OBIA can support further in vineyards mapping for wine producers and classification of viticulture practices in general.