ABSTRACT
Phenotypes are a plant’s traits, such as its morphological and developmental observable characteristics, physiological traits, behaviors, and biochemical properties. Advances in imaging technology, analysis techniques, and in machine learning and statistical approaches can significantly advance the inquiry into phenotyping research. Image-based phenotyping is non-destructive, and therefore, allows for tracking of the dynamics of phenotypes during a plant’s life cycle. This chapter demonstrates and illustrates machine learning and statistical approaches, focusing on two-dimensional Red Green Blue and hyperspectral images, and three-dimensional computed tomography images. It illustrates different machine learning and statistical approaches used in plant phenotyping in the following order: unsupervised dimensional reduction approaches, supervised machine learning approaches for continuous traits, supervised machine learning approaches for binary traits, image segmentation approaches, and approaches combining multiple tracks of information or using temporary traits. To visualize the data in low-dimensional space and to predict traits from images, dimensional reduction approaches have been used.
