Automatic Detection and Assessment of Barcodes on Document Images

In this work, the detection and evaluation of barcodes used in the digitization and tracking processes of internal correspondence of Yurtiçi Kargo Servisi AŞ. were performed. To ensure that all correspondence conducted on paper within the company is digitized, stored, and tracked, barcode labels are affixed to the documents. Barcode numbers are used for searching, responding, and reporting documents in the digital environment. Previously, documents and the barcodes on them were scanned with a special scanner hardware, and the documents were recorded in the system along with the barcode number. However, this situation introduced dependency on specific model hardware and increased costs. In this study, barcodes on documents were detected using scanned document images. For this purpose, a UNet deep neural network model was trained. Barcode images detected by the model were rotated to make them readable by barcode reader libraries. The barcode value was recorded in the system along with the digital version of the document. The system developed as a result of our study has yielded successful results and has been implemented in the internal correspondence of Yurtiçi Kargo Servisi AŞ.