A Mail Sorting System Based on Handwritten Thai Characters

Recent work on a mail sorting system for the Thai Post Office has shown that the recognition error rate was high. This research developed a more efficient mail sorting system by taking the destination province that appeared on envelopes into the processing system. The analysis tools used were an envelope-data-acquiring hardware system and image processing software for sorting mail. The software processed handwritten letters on envelopes and forwarded the data to the processing unit to identify the result. The system was evaluated based on handwritten samples obtained from 30 participants who wrote provinces and zip codes on 231 standard envelopes. There were 2,510 samples used to train the system which were different from the set of handwritten envelopes used to verify the system in order to obtain the closest result to actual usage. The experimental results showed that the recognition performance of the proposed system was significantly increased and can be applied in practical use.