Abstract
دستگاه های طبقه بندی نامه، سامانه هایی هستند که برای دسته بندی نامه ها، بر اساس آدرس مقصد، در ادارات پست مورد استفاده قرار می گیرند. در این پژوهش تلاش کردیم سامانه ای طراحی و ساخته شود که نامه های ورودی به آن با توجه به نام شهر در آدرس دستنویسِ گیرنده، طبقه بندی شوند. این سامانه نمونه ای اولیه از یک دستگاه کاربردی دسته بندی پاکتهای نامه است که برای اداره پست مرکزی استان خراسان رضوی طراحی شده است. داده های تصویری این پژوهش کاملا اصیل بوده و طی چند مرحله تصویر برداری از پاکتهای نامه موجود در مرکز پست، تهیه شده اند. دسته بندی داده ها با استفاده از ماشینهای بردار تکیه گاه (SVM) انجام شد که نتایج خوبی را به ارمغان آورد. داده های ورودی به SVM تصاویر دستنویس فارسی مربوط به نامِ شهرِ مقصد بودند. برای اجرای طرح پیش فرضهایی در رابطه با رعایت اصول اولیه در آدرس نویسی، الصاق تمبر، و عدم درج آگهی بر روی پاکت نامه لحاظ شده است. در این مقاله، جداسازی آدرس از تصویر پاکت نامه و تشخیص کلمات دستنویس توضیح داده می شود. این پژوهش مبتنی بر امکانات موجود در داخل کشور، سیستم کارآمدی را برای دسته بندی خودکار نامه ها ارائه می دهد.
Abstract in English:
Mail sorting machines are systems that are used in post offices to sort mail based on the destination address. In this research, we tried to design and build a system to classify incoming letters according to the name of the city in the recipient's handwritten address. This system is a prototype of a practical device for sorting mail envelopes designed for the central post office of Razavi Khorasan province. The image data of this research are completely original and have been prepared during several stages of photographing the envelopes in the post office. Data classification was done using Support Vector Machines (SVM), which brought good results. The input data to SVM were Persian handwritten images related to the name of the destination city. For the implementation of the plan, presuppositions related to compliance with the basic principles in writing addresses, attaching stamps, and not placing advertisements on the envelopes have been included. This article explains how to extract addresses from envelope images and recognize handwritten words. This research, based on the facilities available inside the country, provides an efficient system for automatic sorting of letters.
Abstract in English:
Mail sorting machines are systems that are used in post offices to sort mail based on the destination address. In this research, we tried to design and build a system to classify incoming letters according to the name of the city in the recipient's handwritten address. This system is a prototype of a practical device for sorting mail envelopes designed for the central post office of Razavi Khorasan province. The image data of this research are completely original and have been prepared during several stages of photographing the envelopes in the post office. Data classification was done using Support Vector Machines (SVM), which brought good results. The input data to SVM were Persian handwritten images related to the name of the destination city. For the implementation of the plan, presuppositions related to compliance with the basic principles in writing addresses, attaching stamps, and not placing advertisements on the envelopes have been included. This article explains how to extract addresses from envelope images and recognize handwritten words. This research, based on the facilities available inside the country, provides an efficient system for automatic sorting of letters.
| Translated title of the contribution | Classification of envelopes using machine vision in reading and processing handwritten addresses |
|---|---|
| Original language | Persian (Iran, Islamic Republic of) |
| Title of host publication | کنفرانس ماشین بینایی و پردازش تصویر ایران |
| Place of Publication | Iran |
| Publisher | University of Mashhad |
| Number of pages | 7 |
| Publication status | Published - 4 Nov 2008 |
| Event | 5th Iranian Conference on Machine Vision and Image Processing - University of Tabriz, Tabriz, Iran, Islamic Republic of Duration: 4 Nov 2008 → 6 Nov 2008 Conference number: 5th https://profdoc.um.ac.ir/list-confname-%DA%A9%D9%86%D9%81%D8%B1%D8%A7%D9%86%D8%B3%20%D9%85%D8%A7%D8%B4%DB%8C%D9%86%20%D8%A8%DB%8C%D9%86%D8%A7%DB%8C%DB%8C%20%D9%88%20%D9%BE%D8%B1%D8%AF%D8%A7%D8%B2%D8%B4%20%D8%AA%D8%B5%D9%88%DB%8C%D8%B1%20%D8%A7%DB%8C%D8%B1%D8%A7%D9%86.html |
Conference
| Conference | 5th Iranian Conference on Machine Vision and Image Processing |
|---|---|
| Country/Territory | Iran, Islamic Republic of |
| City | Tabriz |
| Period | 4/11/08 → 6/11/08 |
| Internet address |
Keywords
- handwritten
- address recognition
- Optical character recognition (OCR)
- Support vector machines (SVM)
- Letter classification machine (LSM)