Smartphone image capture system and image analysis pipelines enable accurate and efficient phenotyping of spaced plant mapping populations

Premise: Image-based phenotyping enhances plant trait measurement, but lack of low-cost, user-friendly image collection and analysis pipelines challenge adoption of such methods. Methods: We describe a low-cost, smartphone-based image capture system (SICS) designed to capture overhead images for phenotyping field plant populations. Two image analysis pipelines were developed using existing image analysis software (PlantCV or Biodock AI) and used to quantify plant area from the SICS images. Results: Over two field seasons, the SICS was used to collect over 12,000 images from spaced-plant genetic mapping populations of two perennial herbaceous candidate crops (Silphium L. backcross and Trifolium ambiguum M. Bieb). Both the PlantCV and Biodock AI pipelines performed with high accuracy, particularly when a single plant was growing against a dark soil background, with plant trait estimates within 30% of the values generated using ImageJ. Though accuracy declined marginally when analyzing complex image compositions (where a single plant was growing with other plants in the background), the decline in accuracy did not significantly affect trends in mean plant area estimated by either pipeline. Discussion: These advancements enable adoption of image-based plant phenotyping in field studies, advancing phenotypic data collection, analysis, and application in a wide array of plant research programs. All supplemental data files and scripts are included in the Supplementary Materials of the manuscript.