Digital Repeat Photography Application for Flowering Stage Classification of Selected Woody Plants.

Abstract

Digital repeat photography is currently applied mainly in geophysical studies of ecosystems. However, its role as a tool that can be utilized in conventional phenology, tracking a plant's seasonal developmental cycle, is growing. This study's main goal was to develop an easy-to-reproduce, single-camera-based novel approach to determine the flowering phases of 12 woody plants of various deciduous species. Field observations served as binary class calibration datasets (flowering and non-flowering stages). All the image RGB parameters, designated for each plant separately, were used as plant features for the models' parametrization. The training data were subjected to various transformations to achieve the best classifications using the weighted k -nearest neighbors algorithm. The developed models enabled the flowering classifications at the 0, 1, 2, 3, and 5 onset day shift (absolute values) for 2, 3, 3, 2, and 2 plants, respectively. For 9 plants, the presented method enabled the flowering duration estimation, which is a valuable yet rarely used parameter in conventional phenological studies. We found the presented method suitable for various plants, despite their petal color and flower size, until there is a considerable change in the crown color during the flowering stage.

Plant phenotyping relevance

単一カメラの反復デジタル画像と機械学習により、木本植物の開花段階・開花期間を推定する手法を開発しており、植物フェノタイピング手法が研究の中心である。

Code and data availability

The paper's phenotyping inputs (phenocamera image time series, ROI-derived RGB color indices, and field BBCH flowering observations) and analysis are not publicly deposited. The Data Availability Statement says raw data are available only on request, and no author code or model repository is given. The MDPI supplement仅

No evidence-backed public reproduction asset is currently recorded.