On Lemon Defect Recognition with Visual Feature Extraction and Transfers Learning

Applying machine learning to lemon defect recognition can improve the efficiency of quality detection. This paper proposes a deep learning-based classification method with visual feature extraction and transfer recognize lemons (i.e., green mold defects). First, data enhancement brightness compensation techniques are used for prepossessing. The is quantify defects determine variables as bandit basis classification. Then we construct convolutional neural network an embedded Visual Geometry Group 16 based (VGG16-based) using learning. proposed model compared many benchmark models such K-nearest Neighbor (KNN) Support Vector Machine (SVM). Results show that achieves highest accuracy (95.44%) in testing set. research provides new solution recognition.