RustNet-3D: a high-resolution LiDAR dataset for wheat rust detection

RustNet-3D is a high-resolution LiDAR dataset specifically designed for the study of wheat rust infections. This dataset provides detailed 3D point cloud data of wheat plants acquired using the Velodyne VLP-16 LiDAR sensor, enabling researchers to analyze plant morphology and estimate disease severity based on LiDAR intensity values.

Data Collection: the dataset includes 3D scans of both inoculated and non-inoculated wheat plants, captured at various stages of rust infection. The LiDAR sensor was mounted on a mobile platform and synchronized with an odometer system for precise spatial data acquisition. The dataset also integrates ground truth measurements, including biomass, number of ears (NoE), number of grains (NoG), and grain weight (Gw), which were collected at the end of the growing season to serve as reference data for disease severity assessment.

Key features:

High-resolution LiDAR data: captured using a Velodyne VLP-16 sensor, offering a 360° horizontal field of view, a vertical resolution of 2°, and up to 600,000 points per second in dual-return mode.

Spectral information: includes LiDAR intensity values in the near-infrared (NIR) region (903 nm), which are sensitive to variations in chlorophyll content and can be used to assess rust infection severity.

Ground truth measurements: biomass, NoE, NoG, and Gw data were collected for each plant to validate disease impact. This information can be found in the DataRustNet.csv file.

Synchronization with ROS: data acquisition and processing were conducted using ROS Melodic on an Ubuntu 18.04.5 LTS system, ensuring accurate synchronization of LiDAR point clouds with odometry data.

Dataset format: the point cloud data is stored in ROS bag files, preserving raw LiDAR and odometry information for advanced analysis and reconstruction.

Potential applications:

3D plant phenotyping and structural analysis

Automated wheat rust detection and severity estimation

Development of AI models for disease classification and early detection

High-resolution 3D modeling for precision agriculture applications

RustNet-3D is an open-access dataset aimed at supporting researchers in the fields of precision agriculture, plant pathology, and remote sensing. By providing high-fidelity 3D representations of rust-infected and healthy wheat plants, it enables the development of robust computational models for disease detection and yield prediction. The code required to perform all analyses can be found at https://github.com/eapolo/agrolidarwheatrust.git.