Abstract:The Douro Demarcated Region is fundamental to local cultural and economic identity. Despite its importance, the region faces the challenge of abandoned vineyard plots, caused, among other factors, by the high costs of maintaining vineyards on hilly terrain. To solve this problem, the European Union (EU) offers subsidies to encourage active cultivation, with the aim of protecting the region's cultural and environmental heritage. However, monitoring actively cultivated vineyards and those that have been abandoned presents considerable logistical challenges. With 43,843 vineyards spread over 250,000 hectares of rugged terrain, control of these plots is limited, which hampers the effectiveness of preservation and incentive initiatives. Currently, the EU only inspects 5 per cent of farmers annually, which results in insufficient coverage to ensure that subsidies are properly used and vineyards are actively maintained. To complement this limited monitoring, organisations such as the Instituto dos Vinhos do Douro e do Porto (IVDP) use aerial and satellite images, which are manually analysed to identify abandoned or active plots. To overcome these limitations, images can be analysed using deep learning methods, which have already shown great potential in agricultural applications. In this context, our research group has carried out some preliminary evaluations for the automatic detection of abandoned vineyards using deep learning models, which, despite showing promising results on the dataset used, proved to be limited when applied to images of the entire region. In this study, a new dataset was expanded to 137,000 images collected between 2018 and 2023, filling critical gaps in the previous datasets by including greater temporal and spatial diversity. Subsequently, a careful evaluation was carried out with various DL models. As a result, the ViT_b32 model demonstrated superior performance, achieving an average accuracy of 0.99 and an F1 score of 0.98, outperforming CNN-based models. In addition to the excellent results obtained, this dataset represents a significant contribution to advancing research in precision viticulture, providing a solid and relevant basis for future studies and driving the development of solutions applied to vineyard monitoring in the Douro Demarcated Region. These advances not only improve efficiency in detecting abandoned plots, but also contribute significantly to optimising the use of subsidies in the region.

Vineyard Segmentation from Satellite Imagery Using Machine Learning

Localization and Mapping on Agriculture Based on Point-Feature Extraction and Semiplanes Segmentation From 3D LiDAR Data

Deep Semantic Segmentation at the Edge for Autonomous Navigation in Vineyard Rows

Vineyard Detection and Vine Variety Discrimination from Very High Resolution Satellite Data

Vineyard Zoning and Vine Detection Using Machine Learning in Unmanned Aerial Vehicle Imagery

AI for Agriculture: the Comparison of Semantic Segmentation Methods for Crop Mapping with Sentinel-2 Imagery

Vine disease detection in UAV multispectral images with deep learning segmentation approach

Navigation Stack for Robots Working in Steep Slope Vineyard

Benchmark of visual and 3D lidar SLAM systems in simulation environment for vineyards

Local Motion Planner for Autonomous Navigation in Vineyards with a RGB-D Camera-Based Algorithm and Deep Learning Synergy

Comparative Analysis of CNNs and Vision Transformers for Automatic Classification of Abandonment in Douro's Vineyard Parcels

Vision-based Vineyard Navigation Solution with Automatic Annotation

LiDAR-based Structure Tracking for Agricultural Robots: Application to Autonomous Navigation in Vineyards

Computer-Vision Based Real Time Waypoint Generation for Autonomous Vineyard Navigation with Quadruped Robots

Position-Agnostic Autonomous Navigation in Vineyards with Deep Reinforcement Learning

Creating a Segmented Pointcloud of Grapevines by Combining Multiple Viewpoints Through Visual Odometry

Semi-supervised deep learning and low-cost cameras for the semantic segmentation of natural images in viticulture

A Deep Learning Approach for Precision Viticulture, Assessing Grape Maturity via YOLOv7

Near Real-Time Vineyard Downy Mildew Detection and Severity Estimation

Grape Cluster Detection Using UAV Photogrammetric Point Clouds as a Low-Cost Tool for Yield Forecasting in Vineyards