Research on Feature Extraction for Chip Resistors defect based on PCA

摘要

Principal component analysis (PCA) is the common method of compressing data for extracting sample statistical feature under the condition of meeting the optimal standard deviation. In this paper, it is to improve the recognition speed that the PCA was used to extract the image features of the chip resistors surface defects when the image is as much as possible to compress image data under the condition of retain the image defect information as much as possible. The result shows that PCA can greatly compress the images data during recognizing the chip resistor defect and improve the recognition accuracy, recognition rate is improved by increasing the training samples under the condition of not affect the recognition time, and the number of principal components has a suitable value. The defect recognition rate is the best when the main component number is the 78.57% of the eigenvectors of the training set covariance matrix.