Algorithms for Plant Phenotyping Imaging: Turning Today’s Limitations into Tomorrow’s Strengths
For more information about the Special Issue, please see:
https://www.mdpi.com/journal/remotesensing/special_issues/phenotyping_imaging
Summary :
The last decade has seen an exponential increase in methodological articles in the area of high throughput phenotyping imaging. Most of them aim to extract quantitative and qualitative information about plants development in response to their environments. This diversity in algorithms is reflecting:
- the broad range of species and their growing conditions (control environment to field, single plant to plot canopy or tree),
- the growth dynamic across the lifecycle (young seedling to mature plants – reproductive stages),
- the organs considered,
- the wide range of traits,
- the type of technology: sensors/camera (RGB, multi- hyperspectral, fluorescence, thermal infrared, lidar technologies and so on) and the vectors (from hand-pole to aircraft or nanosatellite).
The computer science community either aims to develop algorithms enabling a direct quantification of the desired trait or to build proxies replacing the more traditional data collection methods.While some traits measurements (ex: height) rely on well-established methodologies and algorithms, others (ex: proxy for visual scoring of stage of development in field condition) remain complicated to estimate. Besides, the difficulty often resides in having a robust algorithm to extract the information in a dynamic environment (ambient illumination, crop growth…) rather than in the trait itself.
The strength and potentials of the proposed methods are generally well-highlighted in the articles whereas, their weaknesses and limitations are too infrequently discussed in detail. Today’s limitations should be shaping the next generation of studies. As most of the algorithms rely on multiple steps, a deeper understanding of their underlying processes would provide meaningful information on both strengths and weaknesses, thus defining the next challenges for the community.
In this special issue, authors are invited to submit research paper providing new methods or using current ones, emphasizing their limitations in the context of their studies. Review or opinion papers are also welcome.
All species, phenotyping platforms and sensor technologies are welcome. The aims here is to highlight the challenges in algorithm development for tomorrow.
Dr. Nicolas Virlet
Guest Editor
