Adaptive Per-Tree Canopy Volume Estimation Using Mobile LiDAR in Structured and Unstructured Orchards

2025-06-09

AI Summary

To address the challenge of real-time, non-invasive estimation of individual tree canopy volume in both structured and unstructured orchards, this paper proposes an end-to-end mobile LiDAR–based estimation framework. The method integrates LiDAR-inertial odometry (LIO) for high-precision pose estimation, introduces an adaptive point cloud segmentation and geometric reconstruction pipeline, and—novelly—combines DBSCAN with spectral clustering to significantly enhance robustness in segmenting individual trees under dense, overlapping canopies. Evaluated in a structured open-center orchard and an unstructured, densely planted almond orchard, the approach achieves tree segmentation success rates of 93% and 80%, respectively. Canopy volume estimates correlate strongly with UAV-based remote sensing measurements (R² > 0.95). This framework enables autonomous robotic orchard inspection and scalable, precision monitoring of commercial orchards.

Technology Category

• Intelligent Robots: State Estimation
• Computer Vision: Segmentation
• Machine Learning: Ensemble Methods

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Economics, Online Markets and Human Computation: Architectures and workflows that use LLMs for crowd work
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs

Abstract

We present a real-time system for per-tree canopy volume estimation using mobile LiDAR data collected during routine robotic navigation. Unlike prior approaches that rely on static scans or assume uniform orchard structures, our method adapts to varying field geometries via an integrated pipeline of LiDAR-inertial odometry, adaptive segmentation, and geometric reconstruction. We evaluate the system across two commercial orchards, one pistachio orchard with regular spacing and one almond orchard with dense, overlapping crowns. A hybrid clustering strategy combining DBSCAN and spectral clustering enables robust per-tree segmentation, achieving 93% success in pistachio and 80% in almond, with strong agreement to drone derived canopy volume estimates. This work advances scalable, non-intrusive tree monitoring for structurally diverse orchard environments.

Problem

• Real-time per-tree canopy volume estimation using mobile LiDAR
• Adapting to varying orchard structures with LiDAR-inertial odometry
• Robust segmentation for diverse orchard environments

Innovation

• Mobile LiDAR for real-time canopy volume estimation
• Adaptive segmentation via LiDAR-inertial odometry
• Hybrid clustering with DBSCAN and spectral methods