Abstract:In order to exploit remote sensing data operationally for precision agriculture applications, efficient and automated methods are required for the accurate detection of vegetation, crops and different crop varieties. To this end, we have designed, developed and evaluated an object-based classification framework towards the detection of vineyards, the vine canopy extraction and the vine variety discrimination from very high resolution multispectral data. A novel set of spectral, spatial and textural features, as well as rules, segmentation scales and a set of parameters are proposed based on object-based image analysis. The validation of the developed methodology was carried out on multitemporal WorldView-2 satellite data at four different viticulture regions in Greece. Concurrent in situ canopy reflectance observations were acquired from a portable spectroradiometer during the field campaigns. The performed quantitative evaluation indicated that the developed approach managed in all cases to detect vineyards with high completeness and correctness detection rates, i.e., over 89%. The vine canopy extraction methodology was validated with overall accuracy (OA) rates of above 96%. The quantitative evaluation regarding the vine variety discrimination task, including experiments with up to six different varieties, reached OA rates above 85% at the parcel level. The combined analysis of the experimental results with the spectral signatures from the in situ reflectance data indicated that certain vine varieties (e.g., Merlot) presented distinct spectral patterns across the VNIR spectrum.

An assessment of the spectral discrimination between different vine varieties was undertaken using non-destructive remote sensing observations at the véraison period. During concurrent satellite, aerial and field campaigns, in-situ reflectance data were collected from a spectroradiometer, hyperspectral data were acquired from a UAV and multispectral data from a high-resolution satellite imaging sensor. Data were collected during a three years period (i.e, 2012, 2013 and 2014) over five wine-growing regions, covering more than 1000ha, in Greece. Data for more than twenty different vine varieties were processed and analysed. In particular, reflectance hyperspectral data from a spectroradiometer (GER 1500, Spectra Vista Corporation, 350-1050nm, 512 spectral bands) were calculated from the raw radiance values and then were correlated with the corresponding reflectance observations from the UAV and satellite data. Reflectance satellite data (WorldView-2, 400nm-1040nm, 8 spectral bands, DigitalGlobe), after the radiometric and atmospheric correction of the raw datasets, were classified towards the detection and the discrimination of the different vine varieties. The concurrent observations from in-situ hyperspectral, aerial hyperspectral and satellite multispectral data over the same vines were highly correlated. High correlations were, also, established for the same vine varieties (e.g., Syrah, Sauvignon Blanc) cultivated in different regions. The analysis of in-situ reflectance indicated that certain vine varieties, like Merlot, Sauvignon Blanc, Ksinomavro and Agiorgitiko possess specific spectral properties and detectable behaviour. These observations were, in most cases, in accordance with the classification results from the high resolution satellite data. In particular, Merlot and also Sauvignon Blanc were detected and discriminated with high accuracy rates. Surprisingly different clones from the same variety could be separated (e.g., clones of Syrah), while they were confused with other varieties (e.g., with Riesling).