An airborne LiDAR-based methodology for vineyard parcel detection and delineation

Abstract

An airborne lidar-based technique to delineate vineyard parcels from surrounding land uses is proposed and assessed in the Texas Hill Country American Viticultural Area near Austin, Texas, USA. Although most vineyard site analyses are based on multispectral aerial and satellite images, this study takes advantage of the height-based uniqueness of vineyard land uses inherent in the vine-trellising structure to differentiate vineyard areas from non-vineyard areas. A normalized digital surface model was created from lidar data and smoothed with a focal statistics method to identify vine rows and delineate vineyard land-use parcels. A simple unsupervised classification of the three study sites was performed to identify low vegetation areas. The vineyard areas were extracted from the low vegetation class and compared with manually digitized versions. The results suggest that lidar-based data sets can efficiently differentiate vineyard from non-vineyard land use. Our study yielded a mean classification accuracy of 97.55% and successfully extracted vineyard parcel area (mean accuracy 88.79%).