Co-Registration of UAV and Handheld LiDAR Data for Fine Phenotyping of Rubber Plantations with Complex Canopies.

Abstract

Rubber tree phenotyping is transitioning from labor-intensive manual techniques toward high-throughput intelligent sensing platforms. However, the advancement of high-throughput phenotyping remains hindered by complex canopy architectures and pronounced seasonal morphological variations. To address these challenges, this paper introduces a unified phenotyping framework that leverages a novel Wood Salient Keypoint (WSK)-based registration algorithm to achieve seamless data fusion from unmanned aerial vehicle laser scanning (ULS) and handheld laser scanning (HLS) systems. The proposed approach begins by extracting stable wooden structures through a region-of-interest (ROI) segmentation process. Repeatable WSKs are then generated using a newly proposed wood structure significance (WSS) score, which quantifies and identifies salient regions across multi-view data. For transformation estimation, descriptor matching, WSS constraints, and geometric consistency optimization are integrated into a fast global registration (FGR) pipeline. Extensive evaluation across 25 plots covering 5 sites at the National rubber plantation base in Danzhou, Hainan, China, demonstrates that the method achieves a mean co-registration accuracy of 9 cm. Further analysis under varying seasonal canopy complexities confirms its robustness and critical role in enabling high-precision rubber tree phenotyping.

Plant phenotyping relevance

ゴム樹の表現型取得を目的に、UAVおよびハンドヘルドLiDARのデータ融合・位置合わせ手法を開発し、複数圃場で精度と季節変動下の頑健性を評価しているため、方法が中心的である。

Code and data availability

The authors publicly released the ULS-HLS rubber plantation point cloud dataset (25 plots, 5 sites, leaf-on/leaf-off, >400M points) used for this paper's phenotyping/registration analysis, via an explicit Data Availability Statement with a Hugging Face URL matching an allowed URL.