Application noteComputer vision under inactinic light for hypocotyl–radicle separation with a generic gravitropism-based criterion
Introduction
Seedling heterotrophic growth is a crucial stage of the development of plants. After sowing, two successive stages have to occur starting with germination until the radicle protrudes out of the seed coat and then the heterotrophic growth in the soil until the seedling emerges out of the soil. In field conditions, germination and heterotrophic seedling growth stages Taiz and Zeiger (2010) are not easily observable and diagnosis on sources of seedling emergence failure is thus difficult, especially the separation of the respective impacts of the two stages in sowing failures. Non invasive monitoring of seedling growth that would help for a better analysis of seed quality variations is accessible in laboratory conditions with computer vision machines A common imaging system, reported in Jaffe et al., 1985, Ishikawa and Evans, 1997, Walter et al., 2002, Kimura and Yamasaki, 2003, Wang et al., 2009a, French et al., 2009, Belin et al., 2011, Subramanian et al., 2013 with various levels of automation, consists in monitoring a set of seedlings positioned on a row in a vertically settled box with agar gel. A backlight system associated with a camera then produces sequences of images of seedling during growth. From such image sequences, the temporal evolution of the length of the seedling is measurable with classical image processing such as binary image skeletonisation. The studies, made possible by such computer vision systems, target various biological processes during autotrophic growth and including plant gravitropism (Jaffe et al., 1985, Ishikawa and Evans, 1997, Subramanian et al., 2013), hypocotyl elongation in controlled light and temperature conditions (Walter et al., 2002), root length and diameter measurements (Kimura and Yamasaki, 2003, French et al., 2009), photomorphogenesis (Wang et al., 2009a). When used during the heterotrophic growth, a major limitation of these computer vision tools is the need for light in the image acquisition step. Normally, the seedling grows in the soil in the dark, and if exposed to light, its upper parts stop elongating because of the seedling light receptors, and new leaves expand. Therefore, if light is used to monitor heterotrophic growth of seedling, it has to mimic obscurity for plant, i.e. to be non “perceived” or absorbed by plant cells. We define such a light as inactinic. Some authors use infrared LED to work in absence of light visible to human eye and also because it is known in the autotrophic growth that the leaves of the plants are sensitive to the contrast between red and infrared light (Smith, 2000). This strategy however is not suited to mimic obscurity for plant during heterotrophic growth. Recently, it has been shown that seedling elongation was accessible with computer vision in complete absence of light, thanks to the use of passive thermal imaging (Belin et al., 2011, Belin et al., 2014). From thermal contrasts, Belin et al. (2011) shows that it is possible to segment the seedling in the dark and also that it is possible to discriminate in the seedling two sub parts: the radicle and the hypocotyl. Following a gravitropic response, the upper part of the seedling, the hypocotyl, grows to reach the light and activates photosynthesis, while the lower part of the seedling, the radicle, grows deeper to anchor in the soil and provide access to water and nutrients. It is important for plant phenotyping to identify hypocotyl and radicle early after their formation in the seedling, and to follow their development during seedling elongation. Such observations carry useful relevance for the better understanding of variations in plant emergence. The elongation rates of the radicle and the hypocotyl have been demonstrated as key input parameters to predict crop emergence in the soil in various models, e.g. Forcella et al., 2000, Dürr et al., 2001, Brunel-Muguet et al., 2011. The separation of radicle and hypocotyl obtained from thermal contrasts established in Belin et al. (2011) can be used as a reference for very precise measurements, but there are limitations in the use of thermal imaging to monitor seedling growth which are the high cost of thermal imaging and the rather limited spatial resolution of thermal cameras impacting the throughput of such phenotyping systems.
In this study, we address the separation of the hypocotyl and radicle of seedling in elongation with a high resolution imaging system working with inactinic lighting conditions in the visible spectrum. We consider a green LED technology for the imaging system and propose to investigate the inactinic properties of such lighting. We introduce an image processing algorithm based on the prior information of gravitropism to separate radicle and hypocotyl in these lighting conditions, and demonstrate results in accordance with the thermal contrasts and with the visual appreciation by human experts. The computer vision tool presented is shown generically efficient for various seedling species.
Section snippets
Imaging systems
We use two different imaging systems. The main imaging system is an original system we propose with inactinic green light and gravitropism-based image processing criterion to distinguish two organs: radicle and hypocotyl. The second imaging system based on thermal imaging is only used to compare results obtained with main imaging system according to the distinction between organs. The two different imaging systems are acquiring images at a sampling time period of 2 h during time periods of 50 to
Inactinic properties of green LED lighting
We first demonstrate the inactinic properties of green LED light. As visible in Table 1, the study, realized on 30 seedlings of M. truncatula, shows no significant effect of green LED on seedlings development by comparison with development in darkness. This is specially true after 4 days which approximately corresponds to the duration of the monitoring presented in this article.
Validation of automated measurements by confrontation with experts notation
Since no impact of the green LED light is observed on seedlings development, we now assess the image processing
Conclusion
For the automated phenotyping of seedling by computer vision the separation between hypocotyl and radicle is an important step not routinely handled by standard image analysis systems. To contribute on this important phenotyping task, we have proposed here a green LED imaging system associated with a simple criterion based on the recording of the germination point. We have tested the validity and robustness of this criterion by performing hypocotyl–radicle separation on computer images from
Acknowledgment
This work received supports of “Région des Pays de la Loire” in the framework of Phenotic project and from the French Government supervised by the “Agence Nationale de la Recherche” in the framework of the program “Investissements d’Avenir” under reference ANR-11-BTBR-0007 (AKER program). Landry BENOIT gratefully acknowledges financial support from Angers Loire Métropole and GEVES-SNES for the preparation of his PhD. We are grateful to Lydie LEDROIT for her precious technical assistance.
References (26)
- et al.
Thermography as non invasive functional imaging for monitoring seedling growth
Comput. Electron. Agric.
(2011) - et al.
An image acquisition system for automated monitoring of the germination rate of sunflower seeds
Comput. Electron. Agric.
(2004) - et al.
Modelling early growth under different sowing conditions: a tool to predict variations in intercrop early stages
Eur. J. Agron.
(2014) - et al.
Modeling seedling emergence
Field Crop Res.
(2000) Image registration by local approximation methods
Image Vis. Comput.
(1988)- et al.
Thermal imaging for evaluation of seedling growth
- et al.
Analysis of the variability of sugar beet (beta vulgaris l.) growth during the early stages. I-influence of various conditions of crop establishment
Agronomy
(1992) - et al.
Photoreceptors in plant photomorphogenesis to date. Five phytochromes, two cryptochromes, one phototropin, and one superchrome
Plant Physiol.
(2001) - et al.
Simulating the impact of genetic diversity of Medicago truncatula on germination and emergence using a crop emergence model for ideotype breeding
Ann. Botany
(2011) - et al.
Active contours without edges
IEEE Trans. Image Process.
(2001)
Simple: a model for simulation of plant emergence predicting the effects of soil tillage and sowing operations
J. Sci. Soil Soc. Am.
High-throughput quantification of root growth using a novel image-analysis tool
Plant Physiol.
Novel software for analysis of root gravitropism: comparative response patterns of Arabidopsis wild-type and axr1 seedlings
Plant Cell Environ.
Cited by (11)
High throughput phenotyping dataset related to seed and seedling traits of sugar beet genotypes
2020, Data in BriefCitation Excerpt :A growth chamber was specifically equipped for measurements of seedling heterotrophic growth. A method commonly used for mimicking darkness for plants is to observe them under green light, as the plant photoreceptors are assumed to be not activated under wave-lengths in the 515–550 nm range and are sensitive only either to red/far red or blue wavelengths expositions respectively [14,15]. Green LEDs, with narrow light spectrum in the range 515–550 nm were used to light the seeds or seedlings during image acquisition.
Virtual modeling based on deep phenotyping provides complementary data to field experiments to predict plant emergence in oilseed rape genotypes
2016, European Journal of AgronomyCitation Excerpt :The present study used only five genotypes because seedling growth phenotyping was done by hand, while all germination measurements were automated. Different tools are currently being developed to increase the throughput of seedling phenotyping (Demilly et al., 2014; Benoit et al., 2015), which will facilitate the phenotyping of larger populations. This means that such a phenotyping-modelling approach is realistic.
Root Elongation Method for the Quality Assessment of Metal-Polluted Soils: Whole Soil or Soil-Water Extract?
2020, Journal of Soil Science and Plant NutritionDeep learning-based detection of seedling development
2020, Plant MethodsLow-cost vision machine for high-throughput automated monitoring of heterotrophic seedling growth on wet paper support
2019, British Machine Vision Conference 2018, BMVC 2018
- 1
The first three authors contributed equally to the work.