SharpmetriX Dataset: IoT-Based Harvest Monitoring for Yield and Labor Productivity Mapping in Viticulture

"SharpmetriX: A Cost-Effective IoT Solution for Mapping Harvest Vineyards Yields and Labor Productivity Indicators", introducing the SharpmetriX system for in-harvest monitoring in viticulture.

The dataset was collected during field experiments conducted in September 2024 at the Centro de Estudos Vitivinícolas do Dão (CEVDAO), Nelas, Portugal, using custom-developed IoT add-ons attached to harvesting tools (secateurs and collection containers), combined with a smartphone-based data acquisition system.

The dataset comprises three complementary experimental scenarios, reflecting both controlled validation conditions and real-world harvesting operations.

The dataset is organized into three main components, each corresponding to a distinct experimental scenario.

1. Parallel Harvest Dataset

File: Device_cuts_CEVDAO_PH.csv

This dataset corresponds to a real harvesting operation where four workers harvested vine rows simultaneously, each equipped with instrumented secateurs.

Contents

• Timestamped cut events
• Geographic coordinates (latitude, longitude)
• Device identifier (corresponding to each harvester/tool)
• Event-related information associated with each cut

Characteristics

• Data reflect real-world variability and operator behavior
• Suitable for:
• Spatial attribution (e.g., row assignment algorithms)
• Multi-operator analysis
• Evaluation of unsupervised methods

2. Controlled Harvest Datasets

Files:

• Vine_3.csv
• Vine_15.csv

These datasets correspond to controlled harvesting experiments, where a single operator harvested individual vines under monitored conditions, allowing comparison with ground truth.

Contents

• Timestamped data stream including:
• Cut events
• Accumulated harvested weight (grams)
• Acceleration measurements (milli-g)
• Geographic coordinates

Data Encoding

• The column “grams & cuts” contains mixed information:
• The value "CUT" indicates a cutting event
• Numeric values correspond to accumulated harvested weight

3. RTK Benchmark Dataset

File: RTK_data.csv

This dataset was collected to evaluate positioning accuracy, comparing smartphone-based GNSS with RTK-corrected positioning.

Contents

• Latitude and longitude:
• With RTK correction
• Without RTK correction
• Cut state:
• cut = 1 → secateurs closed (cut event)
• cut = 0 → secateurs open

Enables analysis of:

•  
• GNSS positioning error
• Effect of operator movement and posture
• Spatial accuracy improvements with RTK