Research on Defect Detection in Printing Barcode Area Based on Machine Learning

The accuracy and aesthetics of information in the printing barcode area are of great significance in practical life. Flat printing defect detection is quite common, and there is currently almost no involvement in non flat printing defect detection. This work focuses on defect detection design for non flat printed barcode areas of glass bottles. After dividing the barcode area into two parts, barcode and character, we will perform defect detection on these two parts separately. For the barcode part, using support vector machines in machine learning to construct features and classify barcode defects. For the character part, a character quality evaluation method based on gray projection is adopted to output the character quality score. By combining the classification results of barcode defects with subjective judgment of quality scores, defect detection of barcode areas on non flat surfaces of glass bottles has been achieved.