Reading QR codes on challenging surfaces

Abstract

[en] QR Codes are a common way of representing information in a machine-readable format, thanks to its fast reading speed and reliability. The increasing use of this technology has led to one remarkable limitation: the standard algorithms for reading QR Codes assume that the surface where it lies is flat. Therefore, QR Codes that lie non-flat surfaces suffer from slow reading speed and high rate of unsuccessful decoding with the standard algorithms. In this work, we will compare four different correction methods for QR Codes found in challenging surfaces. First, we will implement the method described in the standard of the QR Codes. Secondly, we will study the correction method used in the majority of the commercial readers. Thirdly, we will explore the state of the art for the correction of cylindrical deformations, because it is a common occurrence with some existing solutions. Finally, we will propose using a well-known surface interpolation spline named Thin Plate Spline to target arbitrary deformations. To make this comparison, we will implement a modular library for decoding QR Codes. Using this library we will be able to compare and test the results achieved by each correction method. For making the tests, we will create three different datasets, each with labelled images of QR Codes with different deformations.