High throughput automated seedling phenotyping system

This thesis describes the design, development and analysis of a system to aid in automated phenotyping of seedlings in controlled environments. The task of phenotyping involves observing physical or behavioral changes (phenotype) in a plant as a consequence of an alteration to a particular gene (genetic mutation) in its genome. One method currently gaining a lot of interest for phenotype observation and analysis is the use of camera based acquisition and analysis. But with plant genomes containing upwards of 10000 genes, comprehensive studies will require automated, high throughput methodologies.;To address this automation and throughput requirement without sacrificing spatial or temporal resolution, a robotic gantry system utilizing visual servoing and a sample-holding fixture were developed. A 1 meter square sample fixture holds 36 Petri plates in which seedlings grow along the surface of agar gel perpendicular to the optical axis of two 2 CCD cameras with macro lenses, mounted on the end effector. One camera collects images at low magnification to aid in positioning and the other collects experimental data at higher magnification. Direct drive linear servo motors move the payload at a speed of 1 m/s with an accuracy of 1 micron, and a repeatability of 0.1 micron. A probabilistic image analysis algorithm are used to locate the root of seedlings in the Petri plate and a normalized gray scale variance measure was used to focus the camera on the root. The typical size of a root is about 45 to 200 micron wide and about 4 mm in length. The custom hardware and software together provide a robust seedling imaging platform capable of tracking growth of 72 roots (2 seedlings per Petri plate) in parallel at approximately 3 min intervals. The time series of images are then be subjected to phenotypic analysis for studies of gene function. The entire workflow is designed to have minimal human intervention.