Computer vision under inactinic light for hypocotyl-radicle separation with a generic gravitropism-based criterion

This article proposes a computer-vision based protocol, useful to contribute to high-throughput automated phenotyping of seedlings during elongation, the stage following germination. Radicle and hypocotyl are two essential organs which start to develop at this stage, with the hypocotyl growing towards the soil surface and the radicle exploring deeper layers for nutrient absorption. Early identification and measurement of these two organs are important to the characterization of the plant emergence and to the prognosis of the adult plant. In normal conditions, this growth process of radicle and hypocotyl takes place in the soil, in the dark. Identification and measurement of these two organs are therefore challenging, because they need to be achieved with no light that could alter normal growth conditions. We propose here an original protocol exploiting an inactinic green light, produced by a controlled LED source, coupled to a standard low-cost gray-level camera. On the resulting digital images, we devise a simple criterion based on gravitropism and amenable to direct computer implementation. The automated criterion, through comparison with the performance of human experts, is demonstrated to be efficient for the detection and separation of radicle and hypocotyl, and generic for various species of seedlings. Our protocol especially brings improvement in terms of cost reduction over the current method found in the recent literature which resorts to higher-cost passive thermal imaging to perform the same task in the dark, and that we also consider here for comparison. Our protocol connected to automation of image acquisition, can serve to improve high-throughput phenotyping equipments for analysis of seed quality and genetic variability.