Receipt Recognition Project

-- Core Algorithm Development Overview

Step 1: Preprocessing (CV Group)

Edge Detection: Cropping the receipt part correctly from the images and removing the redundant background noises

• Pre-processing image to enhance the document.
• Finding receipt extraction points in the image.
• Extracting the receipt from the points.
• Post-processing to correct distortions and imperfections.

Ideal example: The algorithm of auto-cropping places the focus on receipts exactly and ignores the background noises (i.e. finger, cup, spoon) perfectly

Step 2: Text Detection (CV Group)

• Outline the exact contour of each piece of information on receipts

Ideal examples:

Step 3: Text Recognition (CV Group + NLP Group)

• Recognising and extracting the exact information from receipts into texts
• The functionality of auto-correcting words may be required

Ideal examples:

Step 4: Information extraction (NLP Group)

• Extract the important information (i.e. Date, Total expense, Name and price of each purchased item) from the results of text recognition
• Ongoing...