PyExpress – A pipeline for fast and reliable UAV image processing in vineyards

Abstract

Increasing drought poses a challenge to viticulture, with complex impacts on grape yield and quality. The use of Unmanned Aerial Vehicles (UAV) in Precision Viticulture offers a valuable tool to detect drought stress, capturing its spatio-temporal variability and thus, supports management strategies. UAV-based infrared or multispectral imagery enables high-resolution, non-invasive assessment of vegetation characteristics and drought stress. However, despite the increasing use of UAVs in vineyard management, a robust and comprehensive methodology for continuous and rapid image processing – providing orthophotos for timely decision support and research analysis – is still lacking.

To address this gap, the PyExpress software package was developed to enable rapid and automated photogrammetric image analysis with particular suitability for applications in vineyards. It builds on the functionality of the Agisoft Metashape API and adapts it to create fully customized and automated workflows for, e.g., UAV-based image processing. Sample workflows have been developed to automate the generation of diverse data products, including e.g. point clouds, camera and reference parameters, and digital elevation models (DEMs). These outputs serve as input data for a variety of applications and further analysis to provide decision support for vineyard management or research, e.g., using Geographic Information Systems (GIS) or 3D point cloud processing software such as CloudCompare. To manage large image and project data sets, PyExpress incorporates the interaction with MinIO, a high-performance object storage system. The automated workflows were extensively tested on infrared UAV datasets for temporal and spatial dynamics assessment of drought stress in a steep slope vineyard (Seußlitz, Germany).

PyExpress reliably provides high-quality analysis results across different measurement, flight, and drought stress conditions. Object storage systems such as MinIO demonstrated excellent suitability for UAV data processing workflows. To ensure precise separation of grapevine canopies from surrounding vineyard environment, PyExpress integrates a classification functionality for distinguishing canopy and soil, with adjustable parameters tailored to specific datasets. This filtering of the point cloud was a critical step in subsequent data processing, mainly concerning the differentiation between, e.g., the temperature of vine canopy and the surrounding vineyard.