Developing techniques for counting strawberry flowers in movable-bench systems in a greenhouse

H. Naito, K. Yoshinaga, T. Fukatsu, S. Hayashi, S. Tsubota, S. Yamamoto
A circulating movable-bench system was installed in a greenhouse at Yamamoto, Miyagi Prefecture, Japan, to demonstrate new strawberry production techniques. To optimize the profitability of the movable-bench system, a fixed-position system for counting flowers using machine vision has been studied. The counting system collects time-series RGB images of the strawberry plants, detects flower regions by analyzing color features, and records the change in the number of flowers during the growing period. Using hue-saturation-intensity images, we propose three flower detection methods, namely 1) the pistil method, 2) the pistil-and-petal method, and 3) the area-divided pistil-and-petal method. Based on the appearance rate of actual flowers in the images, we evaluated the flower counting accuracies of the proposed methods. The percentages of appeared flower number in images varied between 35 and 77% during the growing season. The proposed pistil-and-petal and area-divided pistil-and-petal methods maintained their precision percentages and exhibited better performance in the movable-bench system. The coefficient of determination between the number of actual flowers and the number of flowers detected was 0.73 in the pistil-and-petal method. Using a movable-bench system for fixed-position observation provides novel value to farmers by showing them the change in the number of flowers. However, further work is needed to clarify the relationship between the number of counted flowers and the yield.
Naito, H., Yoshinaga, K., Fukatsu, T., Hayashi, S., Tsubota, S. and Yamamoto, S. (2018). Developing techniques for counting strawberry flowers in movable-bench systems in a greenhouse. Acta Hortic. 1227, 401-408
DOI: 10.17660/ActaHortic.2018.1227.50
https://doi.org/10.17660/ActaHortic.2018.1227.50
fixed-position observation, petal, pistil, horticulture, HSI color space, machine vision, yield estimation
English

Acta Horticulturae