Skip to content
This repository has been archived by the owner on Feb 11, 2024. It is now read-only.

Latest commit

 

History

History
69 lines (42 loc) · 2.82 KB

sohani2018optical.md

File metadata and controls

69 lines (42 loc) · 2.82 KB

Optical Character Recognition Engine to extract Food-items and Prices from Grocery Receipt Images via Templating and Dictionary-Traversal Technique

Ali Sohani, Rafi Ullah, Faraz Ali, Athaul Rai, Richard Messier

Browse

@inproceedings{sohani2018optical,
author = {Ullah, Rafi and Sohani, Ali and Ali, Faraz and Rai, Athaul},
year = {2018},
month = {03},
pages = {},
title = {Optical Character Recognition Engine to extract Food-items and Prices from Grocery Receipt Images via Templating and Dictionary-Traversal Technique}
}

Pipeline

Receipt detection Receipt localization Receipt normalization Text line segmentation Optical character recognition Semantic analysis
❌ ✔️ ✔️ ❌ ❗ ✔️

Receipt localization

  • Image Background Removal

Receipt normalization

  • Otsu’s Image binarization

  • Image and Text De-skewing

  • Image Resizing

Optical character recognition

  • Tesseract OCR Library

Semantic analysis

  • Fields extracted:

    • item names,
    • item quantities,
    • item prices,

  • Section-level extraction

  • Our system then bifurcates the image into sections (like Price, Item-Names, Quantity are dealt separately from one another) according to given template layouts

  • We have used the stored templates of stores in database. And while testing the image, we retrieved that specific store template. Store template have the (x, y) coordinate points, width and height information.

  • image-20200215141726973

  • We had the location of items, prices and quantity in the template. We used this knowledge to retrieve image’s specific portion.

  • Using above technique we retrieved the image portions/sections using template information independent of requested image size, whether greater or lesser than template image.

  • Context sensitive spelling correction is a technique of correcting OCR results by matching them with dictionaries of stores.

  • store specific and grocery related dictionary

  • Words such as tax, total, subtotal, discount etc are included in that corpus. These words are excluded at the very first stage from OCR result.

Notes

  • Before localization, Image stitching is performed if more than one photo of receipt is available.
  • image-20200215141940638