The quality of corn seed production and new variety breeding are affected by the problem of abnormal corn ears. Taking the whole corn ear as research object, the sorting method of three abnormal grains (namely moldy corn ears, worm-eaten corn ears and mechanically damaged corn ears) was researched based on two-dimensional fast imaging technology. Firstly, the portable image acquisition device was constructed based on the monocular vision and the corn ear image was acquired. According to these characteristics of corn ear images, six color features in RGB model and HIS model and five texture features in gray scale images were extracted and normalized to build the classification model of these abnormal corn ears. The classifiers were trained with the support vector machine (SVM) and BP neural network for comparison analysis by using the known feature vectors. The result showed that the SVM classifier had higher accuracy than BP neural network classifier. The accuracies of moldy corn ears sorting, worm-eaten corn ears sorting and mechanically damaged corn ears sorting were 96.0%, 93.3% and 90.0%, respectively. The study made an important foundation for realizing the automatic machine screening of abnormal corn ears and had high application value in improving the corn seed quality.