OHSLIC Algorithm

Online Hyperspectral SLIC (OHSLIC)

OHSLIC is a computationally efficient superpixel segmentation algorithm specifically designed for processing hyperspectral imagery from UAV platforms in real-time.

The Problem

UAV-based hyperspectral imaging generates massive data volumes (hundreds of spectral bands per pixel), making real-time processing infeasible with traditional segmentation methods. Agricultural phenotyping requires timely analysis, not post-processing days later.

Algorithm Innovation

OHSLIC adapts the classic SLIC (Simple Linear Iterative Clustering) algorithm for:

• Online processing: Processes data as it streams from the sensor
• Hyperspectral efficiency: Handles hundreds of spectral bands without computational explosion
• Memory constraints: Operates within limited onboard computing resources

The algorithm achieves superpixel segmentation of individual plant phenotypes directly during flight operations.

Performance

• 200x faster than batch processing methods on hyperspectral data
• Enables real-time phenotype extraction during UAV missions
• Successfully deployed in large-scale vineyard monitoring campaigns
• Preserves spectral fidelity while achieving spatial coherence

Technical Approach

• Spectral distance metrics optimized for high-dimensional data
• Incremental clustering that updates with each scan line
• Adaptive bandwidth selection for varying ground sampling distances

Related Publication: Online Hyperspectral Phenotype Segmentation for UAVs (WACV 2025)

Impact

This algorithm enables precision agriculture at scale by making UAV hyperspectral analysis practical for operational deployment rather than just research demonstrations.