Wearable barcode scanning: Advancements in visual code localization, motion blur compensation, and gesture control: Advancements in visual code localization, motion blur compensation, and gesture control

Visual codes like barcodes and quick response (QR) codes are the most prevalent linking elements between physical objects and digital information. They are found in numerous consumer applications such as shopping, electronic payments, ticketing, or marketing campaigns, but they are also found in logistics and enterprise asset tracking and provide employees access to detailed service records. Reading visual codes is also often the first step for pairing and interaction with physical appliances in various research projects in the area of pervasive computing. While these codes are found almost everywhere, the reading of codes usually requires expensive scanner devices which hinders the original goal of easy access to information about every physical object. Technological advancements in wearable computing and mobile computer vision may radically expand the adoption of visual codes because smartphones, smartwatches, smartglasses, and other wearables enable instant barcode scanning on the go. The wide accessibility, the computing performance, the intuitive user interface, and the relatively low price make personal wearable computers strong competitors for traditional scanners while they also enable new use cases for visual codes. As these devices are primarily designed for other purposes, there are also a few shortcomings. In this dissertation, we describe methods to overcome these shortcomings and to add advanced features that can make wearable barcode scanning an attractive alternative to traditional barcode scanning even outside the consumer domain. We present fast and robust solutions to the following problems on computationally restricted unmodified wearable devices: (i) Fast and robust localization of visual codes: Current smartphone-based barcode scanning solutions require the user to hold and align the tagged object close to the camera. This is especially problematic with smartglasses and leads to lower user acceptance. On the other hand, today’s wearable cameras have a high enough resolution to scan visual codes that are further away, once they are segmented in a preprocessing step. We propose a fast algorithm for joint 1D and 2D visual code localization in large digital images. The proposed method outperforms other solutions in terms of accuracy while it is invariant to scale, orientation, code symbology, and is more robust to blur than previous approaches.