Upland Cotton (Gossypium hirsutum L.) Fuzzy-Seed Counting by Image Analysis

Abstract

In recent years cottonseed size has been reduced as a result of the substantial fiber yield increases cotton breeders have made. Small cottonseed size has been associated with reduced germination, low seedling vigor and stand establishment, and has created production problems for downstream whole seed users. The potential loss in revenue to the cotton industry due to small seed size is substantial and has prompted a renewed effort by breeders to generate high-yielding, high-quality varieties with increased seed size. To aid these efforts and enable a better understanding of the effects of seed characteristics on fiber, a fuzzy-seed imaging method was developed. The method utilizes inexpensive, off-the-shelf equipment and an open source image processing pipeline to derive the number of seeds, seed index, and seed area, height, width, and perimeter. The time to image the seed and process the image takes less than three minutes per sample on average. The seed counts and seed index were strongly correlated with manual measurements at r = 0.967 and 0.693, respectively. Associations among seed characteristics and fiber indicate seed area, when used to calculate lint density, could be a useful selection criterion for breeders to increase both yield and seed size.