Computer Vision as a Tool to Study Plant Development
Morphological phenotypes due to mutations frequently provide key information about the biological function of the affected genes. This has long been true of the plant Arabidopsis thaliana, though phenotypes are known for only a minority of this model organism's approximately 25,000 genes. One common explanation for lack of phenotype in a given mutant is that a genetic redundancy masks the effect of the missing gene. Another possibility is that a phenotype escaped detection or manifests itself only in a certain unexamined condition. Addressing this potentially nettlesome alternative requires the development of more sophisticated tools for studying morphological development. Computer vision is a technical field that holds much promise in this regard. This chapter explains in general terms how computer algorithms can extract quantitative information from images of plant structures undergoing development. Automation is a central feature of a successful computer vision application as it enables more conditions and more dependencies to be characterized. This in turn expands the concept of phenotype into a point set in multidimensional condition space. New ways of measuring and thinking about phenotypes, and therefore the functions of genes, are expected to result from expanding the role of computer vision in plant biology.
Key wordsComputer vision morphometrics plant development image processing
The author thanks Nathan Miller, Tessa Durham, and Liya Wang for many discussions that generated the ideas expressed in this chapter. This project was supported by NSF grant DBI-0621702.