About this Research Topic
The integration of machine learning, computer vision, and big-data analytics, together with life sciences, has opened up new opportunities for plant science research. Multi-factorial models of certain phenotypes can be dynamically generated from big biological datasets to characterize phenotypic features, as well as to predict complex trends while plants are interacting with their environments. Such methodological and technical advances have enabled plant scientists to unravel the genetics of otherwise-complicated plant phenotypes at the level of the cell, organ, tissue, plant, and population.
Together with the rise of whole-genome sequencing of many plant species, large-scale and high-throughput plant phenotyping and associated phenotypic analysis has become a bottleneck that needs to be urgently relieved. Plant genetics and crop breeding studies can therefore be accelerated by several recent rapid technical advances, from sensors to data extraction, combined with systems integration and decreasing costs. For example, novel image-based feature and dynamic (organ, root, etc.) growth traits could be determined and had shown the potential power to detect new candidate genes combined with genetic analysis tools. Another good case is that the unmanned aerial vehicle (UAV) and remote sensing had been flexibly applied to phenotype large-scale crop populations in the field. Now, plant morphological and physiological traits can be assessed non-destructively and repeatedly, across large-scale populations, and throughout growth and development, for which the current technologies are still at early stage and under active development.
X-ray computed tomography (CT) and corresponding CT image pipeline had been developed to quantify root system architectures (RSA). In addition, more optical imaging techniques developed and applied in medicine, such as thermoacoustics tomography (TAT), associated particle imaging (API) using neutrons, tomographic electrical rhizosphere imaging (TERI), low-cost X-ray CT, etc. could be transferred to quantify RSA, dynamic organ growth, and final yield. UAV equipped with various optical sensors will generate huge amounts of data, which need special data processing and standardized analytical software for a broader range of users. Machine learning and deep learning have shown power in identification and classification of pest and other targets, and more applications should also be paid attention, such as quantification and prediction of biotic/abiotic stress, etc.
Combining with the regular surveys of trends in crop phenotyping by International Plant Phenotyping Network (IPPN), in our opinion, the future challenges mainly include dynamically quantifying organ-level growth to provide new insights in crop development, innovation in root phenotyping, low-cost and flexible field phenotyping, deploying 3D imaging techniques during the growing season and in post-harvest phenotyping (x-ray CT, 3D laser scanning, etc.), and novel phenotypic analysis techniques (computer vision, machine learning, etc.). In this Research Topic, we encourage the submissions describing state-of-the-art phenotyping techniques and applications in plant-related studies, with a particular focus on these areas:
• Dynamic organ phenotyping in crop science
• Root system architectures phenotyping: the innovation below ground
• Field phenotyping: unmanned aerial vehicle (UAV), Unmanned Ground Vehicle (UGV), distributed phenotyping, pocket phenotyping etc.
• Post-harvest phenotyping: 3D grain phenotyping, seed germination, etc.
• Machine learning and deep learning in phenotypic analysis.
Keywords: GxE dynamics, Phenotyping, Phenotypic analysis, Machine learning, Dynamics