Receipt Detection

Aitzetmüller Florian
5BI

Puchberger Mathias
5BI

Grömer Herbert

The app developer company “helloagain” encouraged the development of a program for recognition of receipts and the subsequent analysis of purchase data. An incoming invoice is photographed by a customer, automatically uploaded to a server and forwarded to an image processing algorithm. This algorithm crops the image isolating the invoice’s data. In addition, the invoice, if rotated, will be vertically aligned. Afterwards, the image is sent to the Google Cloud service where, using the Cloud Vision API and the OCR (optical character recognition) function, the items of the invoice and the corresponding coordinates are saved in a JSON file and stored locally. First, the BON ID, date and total amount are filtered using regular expressions and then stored in a text file. Second, to identify the products on the invoice, the JSON file is parsed and the outcome is saved in a text file in tabular format. Finally, the text files are converted into Excel format in order to enable detailed analysis with the programme “tableau”. “tableau” offers a large variety of graphic templates to study shopping behaviors.