Development of an automated plant phenotyping system for evaluation of salt tolerance in soybean

Abstract Plant high-throughput phenotyping technology is taking more and more important roles in soybean breeding and genetic research thanks to the advance in sensing and data analytic technologies. However, commercial high-throughput phenotyping systems of general purpose are expensive and complicated for many research groups, and their data analytic methods are designed for specific research projects. The goal of this study was to develop and validate a customized image-based phenotyping system that was used to automatedly collect, process and analyze imagery data of soybean cultivars to evaluate their response to salt stress in controlled environments. The imaging system consisted of a consumer-grade digital camera and an automated platform was used to take sequential images of soybean plants of five cultivars under salt stress during the experimental period. An image processing and analytic pipeline was developed to automatically extract image features and evaluate their tolerance to salt stress. Results indicated that two image features, i.e. canopy area and ExV (the difference of excess green and excess red) were highly correlated with salinity tolerance trait of soybean. The image saturation and blue channel values were able to extract salt stress characteristics and identify different types of salt stress characteristics. In addition, the ratio of damaged leaf area to canopy area was extracted as a novel image feature to quantify the salinity tolerance grade. The overall results indicated that the automatic plant phenotyping system based on low-cost image sensors and automation platform was able to quantify plant stress due to salt stress and would be useful in soybean breeding programs.