Defect-Classification of Parts by BOW Algorithm Based on ZYNQ

In this paper, the ZYNQ platform is chosen to realize the defect-classification of cup-shaped blank based on BOW algorithm. The embedded system provides a new way to implement the parts defect classification. The training samples are extracted and described by the SURF features, then the visual dictionary is generated by K-Means++ algorithm. The Support Vector Machine (SVM) is used to discriminate each class of the images to classify the defect of cup-shaped blanks. The experimental results show that the classification system is robust to the changes in angle and clarity, and the accuracy of classification in clear imaging is 89%.