ABSTRACT
The state-of-the-art computer vision-based methods aim to analyze the image sequences of a group of plants captured by different types of cameras, i.e., visible light, infrared, near-infrared, fluorescent, and hyperspectral, at regular time intervals from multiple viewing angles under varying environmental conditions, to achieve high-throughput plant phenotyping analysis. The phenotypes obtained from an experiment in a high-throughput phenotyping system are stored in a database and then analyzed using statistical methods, data mining algorithms, or visual exploration. Temporal phenotypes provide crucial information about a plant’s development over time. Temporal phenotypes, computed by image sequence analysis, are subdivided into two groups, namely trajectory-based and event-based phenotypes. This chapter presents an image-processing pipeline to compute holistic and component structural phenotypes, using 2-D visible light image sequences. It uses a skeleton-graph transformation approach to identify the components of a maize plant.
