A 3D point cloud instance segmentation method for strawberry based on SGC

Abstract

With effective protective covering and microclimate control, greenhouse crops offer significant advantages, such as high yield and quality, remaining unaffected by seasonal variations and meeting the demand for diverse agricultural products. Leaf area is a critical growth parameter influencing the indoor microclimate and the transport of nutrients within plants. This study introduces a strawberry 3D point cloud instance segmentation method based on SGC to address the challenge of stem and leaf instance segmentation in calculating plant leaf area using 3D point cloud data. High-quality point cloud data were obtained using a 3D scanner, and feature enhancement was achieved through the Leaf Vein and Boundary Preserving Sampling (LVBPS) method. The SGC network achieved an average precision of 90.41% (AP25) and 89.47% (AP50) for instance segmentation, with the precision of leaf segmentation reaching 93.63% (AP25) and 92.80% (AP50). These findings provide valuable technical support and references for greenhouse cultivation and smart agriculture applications. The source code and dataset can be accessed at https://github.com/suyangsuluo/SGC.

Plant phenotyping relevance

イチゴの3D点群から茎・葉をインスタンス分割し、葉面積算出に用いる画像解析手法を開発・評価しており、植物表現型取得が中心である。

Code and data availability

公開本文の所在を確認できませんでした。非公開または購読が必要な可能性があります。

No evidence-backed public reproduction asset is currently recorded.