We developed convolutional neural network (CNN) models using drone pictures to estimate vineyard leaf area index (LAI) and percent canopy cover. These parameters are traditionally measured using hand-held devices (e.g., AccuPAR for LAI and Ceptometer for percent canopy cover) and calculated manually, which is labor intensive and hard to apply to large-scale areas. We collected airborne images or videos by flying a low-altitude drone with a built-in digital camera over a large-scale vineyard. The airborne images convey all necessary information for developing CNN models. To date, we have collected data from the same vineyard over a couple of years. The ground truth values were manually measured using AccuPAR and Ceptometer at the same time of airborne imaging. Specifically, we trained five CNN models to estimate percent canopy cover and leaf area index (LAI). The estimated results over a large vineyard will help guide planting intercrops or cover crops to prevent soil erosion, calculating the correct amounts of expected residue in fall, and foliar sprays of pesticides and fungicides, and characterization of vegetation-atmosphere interactions.
|