本論文提出一套針對帳單類型郵件所設計的小型化自動分揀系統，旨在透過整合機構設計與影像處理演算法，解決地方型郵局因空間與預算受限，難以導入大型自動設備所造成的郵務處理瓶頸。本系統主要由三大模組所組成：郵件發放裝置、郵遞區號辨識演算法，以及輸送帶分揀平台。郵件發放裝置先將郵件由一整疊的狀態被逐封送出，再透過郵遞區號辨識演算法確定該封信件的郵遞區號，最後由輸送帶分揀平台將信件分別送至對應的郵件分類箱中。
本論文的硬體部分使用SolidWorks進行設計與繪製，並使用雷射切割機與3D列印機產出零組件，以達到低成本、可維護性高之模組化的目的。本系統採用筆記型電腦作為上位控制單元，負責執行影像處理與決策運算。為了達成分散控制與提升模組化彈性，本系統使用兩組Arduino Nano作為下位控制單元，分別負責控制信封發送裝置與運輸帶分揀平台。兩組Arduino Nano皆使用USB介面與筆記型電腦連接，透過UART串列通訊與筆電進行訊號的交換，上位機依照辨識結果對下位機傳送對應指令，同時接收回饋訊號以同步狀態並確保流程連貫性，此種多下位機架構可有效提升系統穩定度，且有利於後續系統維護或擴充。
影像處理部分，本系統配備雙攝影機拍攝信封的雙面影像，結合模板匹配、霍夫圓轉換與HSV色彩空間轉換進行郵遞資訊框定位，並使用PaddleOCR進行字元辨識。
在實驗部分，以250封帳單信封做為測試樣本，進行整體流程驗證。郵件發放成功率達95.6%，郵遞區號辨識準確率達96.2%，最終由輸送帶導引至正確分類箱之成功率為99.6%。整體結果顯示本系統具備良好的辨識穩定度，且機構體積精簡，適合部署於空間與資源有限的地區性郵政據點。
本論文證明透過結合低成本機構設計與多階段影像處理演算法，可實現一套具擴展性的郵件自動分揀系統，為中小型郵局的郵務自動化處理提供具體可行的解決方案。

This thesis presents a compact automatic sorting system tailored for bill-type mail. Aimed at addressing the limitations of local post offices—such as restricted space and limited budgets that hinder the adoption of large-scale automation—this system integrates mechanical design with image processing algorithms to streamline mail handling. The system comprises three core modules: a mail dispensing unit, a ZIP code recognition algorithm, and a conveyor-based sorting platform. The mail dispensing unit extracts individual envelopes from a stacked pile, the recognition module identifies the ZIP code, and the conveyor system directs the envelope to the corresponding sorting bin.
The hardware was designed using SolidWorks and fabricated through laser cutting and 3D printing, ensuring a modular, cost-effective, and maintainable structure. A laptop serves as the central control unit, responsible for image processing and decision-making. To achieve distributed control and modular flexibility, the system employs two Arduino Nano boards as local controllers—one managing the mail dispensing module and the other handling the sorting mechanism. These controllers communicate with the laptop via USB using UART serial communication. The central unit sends commands based on recognition outcomes and receives feedback signals to synchronize operations and maintain seamless workflow. This multi-controller architecture improves overall system stability and supports future maintenance or expansion.
The image processing pipeline involves dual cameras capturing both sides of each envelope. Using template matching, Hough circle transform, and HSV color space conversion, the system locates the postal information window, followed by character recognition using PaddleOCR.
To evaluate system performance, experiments were conducted using 250 bill-type envelopes. The mail dispensing success rate reached 95.6%, ZIP code recognition accuracy was 96.2%, and the final sorting success rate was 99.6%. These results demonstrate the system’s high recognition reliability and compact structure, making it well-suited for deployment in local postal facilities with limited space and resources.
This research validates that a scalable and practical automated mail sorting system can be realized by combining low-cost mechanical design with multi-stage image processing. It offers a viable solution for automation in small- to medium-sized post offices.
Keywords: mail sorting, image processing, computer vision, embedded systems, automation systems

[1] 中華郵政股份有限公司, 中華民國112年郵政統計要覽. 台北市: 中華郵政股份有限公司, 2024.
[2] 政府電子採購網. "決標公告." https://web.pcc.gov.tw/tps/atm/AtmAwardWithoutSso/QueryAtmAwardDetail?pkAtmMain=NzA1MTAxODU= (accessed Sep, 2024).
[3] S. Basu, N. Das, R. Sarkar, M. Kundu, M. Nasipuri, and D. Basu, Recognition of Numeric Postal Codes from Multi-Script Postal Address Blocks. 2009, pp. 381–386.
[4] R. O. Duda and P. E. Hart, "Use of the Hough Transformation to Detect Lines and Curves in Pictures," Commun. ACM, vol. 15, no. 1, pp. 11–15, 1972. [Online]. Available: https://doi.org/10.1145/361237.361242.
[5] A. Roy, N. Mazumder, N. Das, R. Sarkar, S. Basu, and M. Nasipuri, "A New Quad Tree Based Feature Set for Recognition of Handwritten Bangla Numerals," in 2012 IEEE International Conference on Engineering Education: Innovative Practices and Future Trends (AICERA), 19–21 July 2012 2012, pp. 1–6.
[6] D. E. Rumelhart, G. E. Hinton, and R. J. Williams, "Learning Representations by Back-Propagating Errors," Nature, vol. 323, no. 6088, pp. 533–536, 1986/10/01 1986. [Online]. Available: https://doi.org/10.1038/323533a0.
[7] S. Basu, N. Das, R. Sarkar, M. Kundu, M. Nasipuri, and D. Kumar Basu, "A Novel Framework for Automatic Sorting of Postal Documents with Multi-Script Address Blocks," Pattern Recognition, vol. 43, no. 10, pp. 3507–3521, 2010/10/01/ 2010. [Online]. Available:https://www.sciencedirect.com/science/article/pii/S0031320310002426
https://www.sciencedirect.com/science/article/pii/S0031320310002426?via%3Dihub.
[8] C. Cortes and V. Vapnik, "Support-Vector Networks," Machine Learning, vol. 20, no. 3, pp. 273–297, 1995/09/01 1995. [Online]. Available: https://doi.org/10.1007/BF00994018.
[9] Y. LeCun, C. Cortes, and C. J. C. Burges. MNIST Handwritten Digit Database. [Online]. Available: http://yann.lecun.com/exdb/mnist/
[10] G. Cohen, S. Afshar, J. Tapson, and A. Van Schaik, "EMNIST: Extending MNIST to Handwritten Letters," in 2017 international joint conference on neural networks (IJCNN), 2017: IEEE, pp. 2921–2926.
[11] M. Rana, M. H. Kabir, and A. Sobur, "Comparison of the Error Rates of MNIST Datasets Using Different Type of Machine Learning Model," 06/08 2023.
[12] J. Canny, "A Computational Approach to Edge Detection," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PAMI-8, no. 6, pp. 679–698, 1986.
[13] I. Sobel and G. Feldman, "A 3×3 Isotropic Gradient Operator for Image Processing," Pattern Classification and Scene Analysis, pp. 271–272, 01/01 1973.
[14] A. P. Witkin, "Scale-Space Filtering," in Readings in Computer Vision, M. A. Fischler and O. Firschein Eds. San Francisco (CA): Morgan Kaufmann, 1987, pp. 329–332.
[15] D. Lowe, "Object Recognition from Local Scale-Invariant Features," Proceedings of the IEEE International Conference on Computer Vision, vol. 2, 01/13 2001.
[16] D. Lowe, "Distinctive Image Features from Scale-Invariant Keypoints," International Journal of Computer Vision, vol. 60, p. 91, 11/01 2004.
[17] C. Chengtao and L. Mengqun, "Tire Pattern Similarity Detection Based on Template Matching and LBP," in 2019 3rd International Conference on Electronic Information Technology and Computer Engineering (EITCE), 18–20 Oct. 2019 2019, pp. 419–423.
[18] S. L. Tanimoto, "Template Matching in Pyramids," Computer Graphics and Image Processing, vol. 16, no. 4, pp. 356–369, 1981/08/01/ 1981. [Online]. Available: https://www.sciencedirect.com/science/article/pii/0146664X81900460.
[19] H. K. Yuen, J. Princen, J. Illingworth, and J. Kittler, "Comparative Study of Hough Transform Methods for Circle Finding," Image and Vision Computing, vol. 8, no. 1, pp. 71–77, 1990. [Online]. Available: https://www.sciencedirect.com/science/article/pii/026288569090059E.
[20] R. Smith, "An Overview of the Tesseract OCR Engine," in Ninth International Conference on Document Analysis and Recognition (ICDAR 2007), 23–26 Sept. 2007 2007, vol. 2, pp. 629–633.
[21] Y. Du et al., "PP-OCR: A Practical Ultra Lightweight OCR System," arXiv preprint arXiv:2009.09941, 2020.
[22] S. Patil, B. Basutkar, P. Kotwal, and K. Sonawane, "Object Sorting by Robotic Arm using Image Processing," International Journal of Creative and Arts Studies, vol. 9, 04/14 2021.
[23] M. C. Schwaiger, T. Thummel, and H. Ulbrich, "A 2D-Motion Platform: The Cybercarpet," in Second Joint EuroHaptics Conference and Symposium on Haptic Interfaces for Virtual Environment and Teleoperator Systems (WHC'07), 22–24 March 2007 2007, pp. 415–420.
[24] M. Q. Zaman and H. M. Wu, "Intelligent Motion Control Design for an Omnidirectional Conveyor System," IEEE Access, vol. 11, pp. 47351–47361, 2023.
[25] Compona. "2253RTC型錄." https://www.compona.com.tw/proimages/pdf/2022-2253RTC%E5%9E%8B%E9%8C%84.pdf (accessed Jun, 2025).