Robotic system with tactile-enabled leaf tracking for high-resolution hyperspectral imaging device for autonomous corn leaf phenotyping in controlled environments

Abstract

Hyperspectral imaging of individual corn leaves provides valuable data for analyzing nutrient content and diagnosing diseases. However, existing leaf-level imaging techniques face challenges such as low spatial resolution and labor-intensive processes. To address these limitations, this study developed a robotic system integrated with a high-resolution line-scanning hyperspectral imaging device to autonomously scan a corn leaf. The hyperspectral imaging device used a vision-based tactile sensor for active leaf tracking throughout the scanning process, ensuring high image quality. Additionally, the device incorporated an in-hand leaf manipulation mechanism that ensured the leaf was properly positioned on the tactile sensing area at the start of every scanning. The scanning process was executed by a robotic arm equipped with an RGB-D camera and integrated with the Segment Anything Model (SAM), enabling autonomous leaf detection, localization, grasping, and scanning. The system was tested on V10-stage corn plants and the success rate was 91.4 % with an average 4.8 s for leaf detection and localization and an average leaf scanning time of 38.3 s.

Plant phenotyping relevance

トウモロコシ葉の高解像度ハイパースペクトル画像取得を自動化するロボット・触覚追跡システムを開発し、検出・走査成功率や処理時間で評価しており、表現型取得手法が中心である。

Code and data availability

公開論文であることは確認できましたが、現在の公式API・許可済み取得経路では本文を自動取得できませんでした。

No evidence-backed public reproduction asset is currently recorded.