A decision support system for hybrid corn classification
Abstract High yielding corn is primarily derived from a cross-pollination among superior appearing male and female plants. Cross-pollination is closely linked at the tasseling/flowering stage, marked by the emergence of tassel for 5-10 days. With the advancement of machine learning, there are opportunities to apply deep learning models to control the purity of plants. The research aims to develop a decision support system based on deep learning to enable earlier identification and removal of contamination/off-type plants during seed production. The datasets containing 1,587 tassel images taken by high resolution camera. The results of the training and the validation sequence indicated a highly correlated accuracy score. A quite contrasting tassel morphology makes it easier for the model to distinguish on and off-type plants. The loss value during the training and the validation stages was 0.05 and 0.1 respectively. A stand-alone graphical user interface (GUI) was deployed to support the early detection of tassels in the field. This tool can be used to support national corn seed production programs.