ABSTRACT
This chapter provides an in-depth discussion of the data-driven techniques that are useful to analyze hyperspectral image sequences for plant phenotyping. It introduces a graphical user interface (GUI)-based HypeRPheno toolbox to assist information extraction from hyperspectral imaging (HSI), using machine learning techniques. The chapter describes a dataset that consists of hyperspectral image sequences of tobacco plants. It also describes the dataset that is used to illustrate three core data-driven techniques widely applicable for plant phenotyping, namely dimensionality reduction, clustering, and classification. In summary, the state-of-the-art research strongly supports the observation that HSI has excellent potential in plant phenotyping in detecting abiotic or biotic stresses in the plants. Dimensionality reduction is the process of decreasing the number of variables used in downstream analysis. It is critical in hyperspectral image analysis because the information is captured in narrow wavelengths in hundreds of bands.
