An automatic extraction and classification method of electronic component tags combining OCR and object detection

Aiming at the problems of low accuracy and slow speed of label recognition in the electronic manufacturing industry, this paper proposes an automatic extraction and classification method for electronic component labels combining OCR and object detection. This method uses the YOLOv5 model to quickly obtain label positioning information and then uses PaddleOCR to recognize the label character content. On this basis, Classify and structure each field by using regular expressions and keyword matching. After experimental verification, this method has a high accuracy rate and good robustness in various practical scenarios.