ASABE Technical Library - Abstract

Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, view access and subscription options

Measuring Leaf Traits of Phalaenopsis Seedling Using Deep Learning

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  2026 ASABE Annual International Meeting  2600635.(doi:10.13031/aim.202600635)
Authors:   Yung-Chi Hu, Shu-Jen Lee, Jui-Hsiang Shao, Yan-Fu Kuo
Keywords:   Phalaenopsis, Computer Vision, Plant Phenotyping, You Only Look Once (YOLO)

Abstract.

Orchids are among the most substantial ornamental plants, with a global trade revenue exceeding USD 776 million in 2024. Among them, orchids in genus Phalaenopsis dominate the market. In Taiwan, Phalaenopsis generated an export revenue of USD 150 million in 2024. To accommodate the limited blooming (marketable) period, Phalaenopsis flask seedlings are first acclimated and planted, and then transplanted into pots of appropriate sizes based on plant stature for production. Following a 12-to-18-month cultivation period, these large seedlings are subsequently exported to overseas nurseries for further cultivation. Prior to export, assessment—involving the checking of leaf length and width—is typically conducted to ensure seedling quality. The assessment is conventionally conducted manually. This manual process is, however, labor-intensive and prone to inconsistencies arising from human subjectivity and fatigue. This study proposed assessing the leaf traits of Phalaenopsis seedlings automatically. Top-view and side-view images of the seedlings were collected. Leaf contours in both the top-view and side-view images were identified using the YOLOv8-seg model. Subsequently, the orientation angles of the leaves relative to the camera were measured from the top-view images and used to calibrate the corresponding leaf lengths derived from the side-view contours. These lengths were calculated using third-degree polynomial curve fitting. In parallel, leaf widths were derived from the minimum bounding rectangles of the top-view contours. Analysis indicated that the proposed phenotyping approach achieved a root mean squared error of 2.32 cm for length and 0.66 cm for width. This comprehensive framework offers a dependable, objective, and high-throughput solution for the automated evaluation of Phalaenopsis seedlings, addressing the competitive demands of Taiwan's export market with enhanced efficiency.

(Download PDF)    (Export to EndNotes)