Enhancing the study of seedling form and development through the application of computer vision algorithms

The field of computer vision is well suited to bring about change in the area of quantitative plant phenotyping. In order to have the largest positive influence in this area, computer vision algorithms specialized for plant morphometic phenotyping must be researched and designed. This thesis work introduces and utilizes two such examples. The first chapter presents a computer vision algorithm that is capable of measuring seedling parameters during the early stages of seedling development and establishment. A digital tool of this type could quickly screen thousands of seedlings for those which exhibit anomalous behavior or properties. The algorithm demonstrated in this chapter measures germination time, seedling area, cotyledon area, and root tip angle. The second chapter explains a computer vision algorithm which monitors the roots of young seedlings as they respond to a gravistimulation. A root's gravitropic response can be quantified by monitoring morphometric parameters such as growth rate, tip angle, and axial curvature distribution. This algorithm is capable of extracting a root midline from a series of digital images. From this midline, the above parameters can be quantified at a spatial resolution of approximately 5 mum pixel-1. This algorithm is currently being used in three labs to monitor the gravitropic response in both Lycopersicon esculentum and Arabidopsis thaliana. The third chapter of this thesis probes the root gravitropic response space as a function of non-genetic factors. Because computers are performing the measurement routines, the time constraint of human measurement is removed. Thus, trial sizes that are orders of magnitude higher than previously possible can be achieved. With the computer vision advantage comes the ability to view the gravitropic response as a suite of responses dependent on a set of non-genetic factors such as seed size, seedling age, and media composition. This chapter begins to lay the foundations of a large, multidimensional response search space within which mutants may be found to exhibit extraordinary behavior. In an era where discovering genes that have a role in plant growth and development is a possibility, computer vision is a vital tool that will expedite the transformation of this possibility into a reality.