Development of an Imaging Platform for Maize Kernel Morphology

Abstract

The field of computer vision can play a significant role in measuring features that describe the physical appearance of maize kernels. Extending a computer vision system to an almost automatic process can create a high througput solution that provides accurate data. Previously, such a system was designed in order to show a proof of concept. In order to fully realize that design, both positives and negatives needed to be considered to develop the fully functioning imaging platform. After developing and putting the system to use, it was discovered the original design was unable to account for all shapes, sizes, and colors of maize kernels. This lead to a redesign and implementation of the feature extraction method. The end result has led to an imaging platform that can provide accurate feature data at a high throughput rate.