Abstract: Using machine vision to replace artificial vision and realizing navel orange quality grade detection, the mathematical morphology was employed to separate navel orange from background. The bulk features, surface defect features, color features and texture features were extracted, which were the input feature vectors of the support vector machine (SVM) , and SVM is used in training and classification of those features. The trained classifier was used to detect navel oranges. Experimental results show that the classifier has the feature of higher rate of correct identification and real-time, and it can be used in real-time detection of navel oranges.