The ability to non-destructively image and automatically phenotype complex root systems, like those of rice (Oryza sativa), is fundamental to identifying genes underlying root system architecture (RSA). Although root systems are central to plant fitness, identifying genes responsible for RSA remains an underexplored opportunity for crop improvement. Here we describe a non-destructive imaging and analysis system for automated phenotyping and trait ranking of RSA. Using this system we image rice roots from 12 genotypes. We automatically estimate RSA traits previously identified as important to plant function. In addition, we expand the suite of features examined for RSA to include traits that more comprehensively describe monocot RSA but that are difficult to measure with traditional methods. Using 16 automatically acquired phenotypic traits for 2297 images from 118 individuals, we observe (i) wide variation in phenotypes among the genotypes surveyed; (ii) greater inter-genotype variance of RSA features than variance within a genotype. RSA trait values are integrated into a computational pipeline which utilizes supervised learning methods to determine which traits best separate two genotypes, and then ranks the traits according to their contribution to each pair-wise comparison. This trait ranking step identifies candidate traits for subsequent QTL analysis and demonstrates that depth and average radius are key contributors to differences in rice RSA within our set of genotypes. Our results suggest a strong genetic component underlying rice RSA. This work enables the automatic phenotyping of RSA of individuals within mapping populations, providing an integrative framework for QTL analysis of RSA.