Aims: This work aims to study the quality of low cost Digital Surface Models (DSMs) obtained with Unmanned Aerial Vehicle (UAV) images and to test whether these DSMs meet common requirements of the wine industry.Methods and results: Experiments were carried out on a 4-ha vineyard located 10 km north of Beziers (France). The experimental site presents slope and aspect variations representative of mechanised commercial vineyards in Languedoc Roussillon. DSMs were provided by three UAV companies selected for the diversity of their solutions in terms of image capture altitude, type of UAV and image processing software. DSMs were obtained by photogrammetry and correspond to commercial products usually delivered by UAV companies. DSMs from UAV were compared to a reference Digital Elevation Model (DEM) acquired by a laser tachymeter. Four indicators were used to test the quality of DSMs: the mean error and its dispersion in the XY plane and in elevation Z. Results show a good georeferencing of the DSMs (MeanErrorXYIntroductionThe detailed knowledge of elevation and its variations is of importance in viticulture. These data are typically contained in information layers called Digital Elevation Model (DEM). DEMs with high spatial resolution (et al., 2013) or at the within field scale (Bramley et al., 2011a). Beyond elevation, DEMs also make it possible to estimate the slope and the aspect of the fields. Therefore, DEMs provide basic information for soil erosion investigations (Martinez-Casasnovas and Sánchez-Bosch, 2000), delineation of optimal maturation zones (Olsen et al., 2011) and terroir (Carey et al., 2008), as well as identification of frost hazard risk areas (Madelin and Beltrando, 2005). Until now, the implementation of an accurate DEM (precisionSantesteban et al., 2013) or embedded on mobile machines (Bramley et al., 2011b). Photogrammetry by Unmanned Aerial Vehicle (UAV) (Stefanik et al., 2011) is a new source of information that allows the production of Digital Surface Models (DSMs) with a high spatial resolution. This technology may present an alternative to conventional methods. DSMs provided by commercial services based on UAV are already available for the wine industry. However, the use of this source of information still raises some questions. Indeed, UAVs do not necessarily measure the elevation of the ground, but the elevation of the visible objects from above (i.e. vegetation over the soil). This justifies the distinction between DSM and DEM.The interest of UAV-based DSM has been shown for applications in different areas like archeology (Sauerbier and Eisenbeiss, 2010), hydrology (Leit?o et al., 2016), engineering (Uysal et al., 2015) or agriculture to monitor crops (Bendig et al., 2013). These studies showed it was possible to estimate the elevation with errors ranging from 0.5 m to 0.05 m. However, they were all performed with specific equipment and acquisition conditions as part of a scientific work. The assessment of the quality of DSMs is made point-to-point on a small amount of ground truth sites (Uysal et al., 2015) or by type of observed object (Leit?o et al., 2016) on areas where characteristics (slope, orientation) and objects may differ drastically from that of a vineyard context.Considering the requirements of the wine industry, the objective of this work is to assess the quality of elevation measurements derived from DSMs obtained by UAVs in the specific context of a Mediterranean vineyard. Given the wide variety of parameters (type of UAV, acquisition conditions, chip size of the sensor, flight elevation and speed of the flight, image overlap and software and algorithms used) that may affect the quality of the results, this work does not aim to provide references on the best acquisition conditions. It focuses on comparing several current commercial services making the assumption that the companies in charge of these services deliver the best possible information under the conditions of the study.The originality of this work is: i) to consider several acquisition and processing chains which are currently available for DSM by a UAV: type of UAV, elevation of image acquisition, software used to compute the elevation from images, ii) to consider a study site as representative as possible of vineyard landscapes; this site takes into account specific objects which need to be described by the DEM, it also shows the specific constraints in obtaining a DEM from a DSM in these specific conditions, and iii) to propose a detailed study of the spatial distribution of the error over the whole study area in order to verify whether the error is constant over the entire study site or whether specific objects of viticulture landscape affect the quality of the DEM derived from UAV. To our knowledge, such a study has never been carried out in a commercial context.Materials and methodsStudy zoneThe study zone was a 4-ha vineyard located 10 km north-west of Beziers, in Languedoc in the south of France.
展开▼
