Development of a novel image-based grain counting setup for thousand-grain weight estimation in wheat

Abstract

Thousand-grain weight (TGW) is one of the major yield-contributing traits routinely used as a selection criterion by plant breeders. It
is also an important grain quality trait that determines milling yield. Accurate phenotyping of TGW is imperative to dissect its genetics
for yield improvement. The traditional approach to TGW estimation involves manual grain counting and weighing, which is laborious,
tedious and less accurate for large sample sizes. As an alternative, we propose a customized grain counting setup for accurate estimationof TGW in wheat by assembling a photo lighting tent and a smartphone for image acquisition of grain samples. A popular open-sourcesoftware, ‘imageJ’ was used to process the images to estimate the grain count. The counted grain samples were weighed to calculatethe TGW. The TGW estimate derived from the proposed grain counting setup displayed a high degree of correlation with the manuallyestimated TGW data (r = 0.99, p <0.05). It took significantly less time to count the grain samples using the proposed setup comparedto manual counting with better accuracy and minimal labor. The error rate in grain counting using the imaging-based setup was verylow (<1%) and 30 to 40 grain samples can be imaged per hour. This setup can be extended to estimate the TGW of different crops,excluding those having spherical seeds.