Smartphone‐based proximal sensing as an accessible platform for improved physiological phenotyping

Abstract We tested whether smartphone‐based sensing could deliver accurate physiological traits in an accessible and low‐cost platform. A smartphone, carried by someone walking through a field, was used to generate a high‐resolution time series of images for 10 genotypes of cowpea ( Vigna unguiculata L. Walp.). We used this data to track genotypic differences in vegetative growth at the leaf‐scale as well as reproductive dynamics at the scale of individual flowers and pods. In total, we extracted 12 sensed traits as well as 23 manually measured ground truth traits. Ten of the manually measured ground truth traits could serve as direct proxies for our sensed traits, allowing us to validate the sensing approach. Not only were our high‐throughput sensed traits comparable to the labor‐intensive measurements ( R 2 ≥ 0.78), they provided more detailed information than could be feasibly captured with traditional phenotyping. Our sensed traits uncovered a tight coupling between vegetative and reproductive phenology and tentatively identified maximum pod length and the lag between the start of flowering and peak flowering as traits with implications for pod harvest index and yield. We believe our general methodology of producing a high‐resolution time series of vegetative and reproductive dynamics has broad applicability and can potentially unlock more detailed physiological traits for breeders at minimal cost.