Phenotype-based Robotic Screening Platform for Leafy Plant Breeding

Abstract

Abstract: Automatically measuring the dynamics of plant phenotype is fundamental to the enhancement of our ability to dissect the agriculturally important traits and the understanding of plant development processes. This paper describes a high-throughput, automatic phenotyping platform to trace the phenotype of leafy plant and complete the screening function. First, a binocular stereo vision system is introduced to acquire images and transfer them to a host computer which processes, analyses and obtains some certain morphological parameters, such as leaf area and height. Second, according to the parameters obtained at different time points, a quantitative phenotype database and a prediction model of growth are established. Third, based on the models, a robotic arm executes the transplanting instructions to screen the plant with undesirable characteristics. The experiment results of leaf area show the measurement accuracy is higher than 90%, and this method can be applied in accurate measurement of plant phenotypic parameters. This work demonstrates how a high-throughput phenotyping equipment can construct an evaluation index system of plant growth during its whole cultivation period with high spatial and temporal resolution by machine vision, and offers an automated approach to the screening in plant breeding.