Skip to Main content Skip to Navigation

Lecture automatique de tickets de caisse

Abstract : The large increase in multimedia data volume and especially the dematerialization of paper documents requires the implementation of solutions to automatically analyze these documents in order to facilitate their storage and their use. Moreover, there is currently a strong interest of companies or institutes to access consumer information of populations or population groups in order to have a better understanding of consumer behavior. The sales receipt is a solution to obtain this information without strongly soliciting the consumer. The objective of this thesis is to propose a solution to automatically analyze the contents of a sales receipt from a photo taken by a smartphone. We begin by explaining the industrial objectives and, through the development of demonstrator, we highlight the scientific obstacles of the realization of such a system, from the acquisition of the picture to the extraction of the textual data contained in the ticket. At the end of this study, we propose an original processing chain to best meet all expectations and constraints. Then, we realize a state of the art detailing methods of detection of objects based in particular on deep neural networks (logo detection, text detection...). We also present text recognition methods and existing associated tools (OCR). Finally, we end up evoking some approaches concerning semantic analysis. The first part of the realization of the chain is the pre-treatment. This phase has several goals : checking the presence of a sale receipt within the image, ticket in order to crop it and straighten it, and then to determining the brand of the receipt. In order to minimize false alarms, each of these objectives is obtained after merging the results of two methods based on different sources (image and text). The second part is to analyze the content of the receipt, starting with the semantic segmentation of the receipt areas (header, logo, product list, bottom of receipt, etc.), then performing optical recognition and finally applying a semantic analysis to extract the different relevant information.
Complete list of metadata

Cited literature [174 references]  Display  Hide  Download
Contributor : Patrick Lambert <>
Submitted on : Monday, March 30, 2020 - 5:54:04 PM
Last modification on : Friday, November 6, 2020 - 3:27:21 AM


Thèse Rizlène Raoui.pdf
Files produced by the author(s)


  • HAL Id : tel-02525154, version 1



Rizlène Raoui-Outach. Lecture automatique de tickets de caisse. Traitement des images [eess.IV]. Université Grenoble Alpes (France), 2019. Français. ⟨tel-02525154⟩



Record views


Files downloads