High throughput ear phenotyping: aplication in sweet corn

Abstract

The use of digital images in high throughput phenotyping has been shown to be a promising technique to speed up and increase the genetic gains of qualitative traits linked to the quality of ears and grain productivity in breeding programs for sweet corn. Therefore, the goals of this study were to establish an efficient protocol for ear phenotyping, using digital images. The acquisition of the images was accomplished in a reduced time in a simple and low-cost platform. The images were processed in open software using image segmentation techniques. After processing the images, pixel matrices were extracted, which allowed the characterization of the ears in terms of length, width, total area, tip fill area, number of rows and number of kernels. Index of agreement were used to verify the efficiency of phenotyping through digital images when compared with manual phenotyping. The proposed methodologies were shown to be promising for the characterization of corn ears through digital images. High correlations were observed, ranging from 0.6 to 0.95, between the values estimated via images and those obtained manually. The techniques of image segmentation and the proposed methodology to determine the length and width of the ears were highly efficient. The use of images reduced the time and labor spent on characterizing the ears. The protocol established in this study can replace manual phenotyping and be adopted on a large scale, without high costs, in any breeding program. The results presented encourage further investigations related to image processing in order to speed up the phenotyping of characteristics of interest to the breeder.