BACKGROUND Considerable number of indoor navigation systems has been proposed to augment people with visual impairments (VI) about their surroundings. These systems leverage several technologies, such as computer-vision, Bluetooth low energy (BLE), and other techniques to estimate the position of a user in indoor areas. Computer-vision based systems use several techniques including matching pictures, classifying captured images, recognizing visual objects or visual markers. BLE based system utilizes BLE beacons attached in the indoor areas as the source of the radio frequency signal to localize the position of the user. METHODS In this paper, we examine the performance and usability of two computer-vision based systems and BLE-based system. The first system is computer-vision based system, called CamNav that uses a trained deep learning model to recognize locations, and the second system, called QRNav, that utilizes visual markers (QR codes) to determine locations. A field test with 10 blindfolded users has been conducted while using the three navigation systems. RESULTS The obtained results from navigation experiment and feedback from blindfolded users show that QRNav and CamNav system is more efficient than BLE based system in terms of accuracy and usability. The error occurred in BLE based application is more than 30% compared to computer vision based systems including CamNav and QRNav. CONCLUSIONS The developed navigation systems are able to provide reliable assistance for the participants during real time experiments. Some of the participants took minimal external assistance while moving through the junctions in the corridor areas. Computer vision technology demonstrated its superiority over BLE technology in assistive systems for people with visual impairments.

中文翻译：

---

针对视觉障碍者的基于计算机视觉和BLE技术的室内导航系统的比较分析。

背景技术已经提出了相当数量的室内导航系统以增强关于周围环境有视觉障碍（VI）的人。这些系统利用多种技术（例如计算机视觉，低功耗蓝牙（BLE））和其他技术来估计用户在室内区域的位置。基于计算机视觉的系统使用多种技术，包括匹配图片，对捕获的图像进行分类，识别视觉对象或视觉标记。基于BLE的系统利用附着在室内区域的BLE信标作为射频信号源来定位用户的位置。方法在本文中，我们研究了两种基于计算机视觉的系统和基于BLE的系统的性能和可用性。第一个系统是基于计算机视觉的系统，名为CamNav的CamNav使用训练有素的深度学习模型来识别位置，第二个名为QRNav的系统利用可视标记（QR代码）确定位置。使用三个导航系统时，已经对10个蒙着眼睛的用户进行了现场测试。结果导航实验的结果以及蒙住眼睛的用户的反馈表明，QRNav和CamNav系统在准确性和可用性方面比基于BLE的系统更为有效。与基于计算机视觉的系统（包括CamNav和QRNav）相比，基于BLE的应用程序中发生的错误超过30％。结论所开发的导航系统能够在实时实验中为参与者提供可靠的协助。一些参与者在穿过走廊区域的交叉路口时，得到的外部援助很少。在视觉障碍者的辅助系统中，计算机视觉技术展示了其优于BLE技术的优势。