Plant phenotyping is now being widely used to study and increase the yield of row-crop plants. Phenotyping is defined as a set of observable characteristics of an individual that results from its interaction of its genome with the environment. Therefore, the collection of physical and observable traits is the primary task of any phenotyping study. While current phenotyping methods are painstakingly slow and tedious, advances in digital imagery and computer technology have unlocked new avenues for this arduous task. High-resolution im-ages can now easily be obtained with practically any camera whereas improvements in com-puter technology mean that images taken can be processed at a shorter time.

Phenotyping generally can be classified into two categories, below ground phenotyp-ing and above ground phenotyping. Below ground phenotyping typically pertains to roots or parasites that are in the soil. The study results from below-ground phenotyping are of the root system architecture of a plant or the cause and effect of below ground parasites. Above ground phenotyping encompasses more variety of traits which includes flowers, fruits, leaves and more. This thesis discusses a computational platform for rapid phenotyping of two prob-lems: root phenotyping and maize flower phenotyping. Both of these phenotyping studies involved collaborative works with a plant science group.

The first phenotyping platform was intended for a study of seedling root traits, which offer an opportunity to study Root System Architecture of a plant without having to wait for the plant to be fully grown. A framework was developed that would take root images and output traits of the plants using image segmentation and graph-based algorithms. The frame-work can also be extended easily to any another kind of roots. The input to the framework would just be a picture of a root with great contrast to the background, and the program

would output the traits out in a simple and easily understandable manner. The ease of use not only means that phenotyping can be done in a very time, cost and labor efficient manner, but also just about anyone could use the program.

The next phenotyping platform was intended to extract phenotyping traits of maize tassels. On field, time series images from two different plantings were provided by the Plant Science Institute for the development of the framework. The planting consisted of nearly four thousand different genotypes. The developed framework could identify the object of interest (the tassels) and analyzed it using image analysis techniques and deployed on the ISU super-computer, CyEnce. Utilizing feature detection and extraction along with segmentation meth-ods, the tassel location could be identified and separated from the background. Then, graph-based techniques and morphological operations were used to extract the various traits of the tassels. By plotting the extracted traits, the growth, and development of the maize tassel over time could be seen and further studied. This framework is also easily extendable to other types of above ground phenotyping. However, due to the nature of having feature detection, significantly more dataset is needed for training the detection algorithm.

This thesis will illustrate how the combination of high-performance computers, image analysis, and machine learning are ushering a revolution in the field of agriculture. The fact that computer processing speed are almost doubling every 18 months provides access to new methods that were not possible before. Just as the landscape of technology is constantly be-ing innovated, phenotyping studies will ensure that the field of agronomy not be left behind.