Wang Shuihua, Yang Xiaodong, Tian Yingli
Netw Model Anal Health Inform Bioinform. 2013 Jul 1;2(2):81-93. doi: 10.1007/s13721-013-0027-9.
Signage plays a very important role to find destinations in applications of navigation and wayfinding. In this paper, we propose a novel framework to detect doors and signage to help blind people accessing unfamiliar indoor environments. In order to eliminate the interference information and improve the accuracy of signage detection, we first extract the attended areas by using a saliency map. Then the signage is detected in the attended areas by using a bipartite graph matching. The proposed method can handle multiple signage detection. Furthermore, in order to provide more information for blind users to access the area associated with the detected signage, we develop a robust method to detect doors based on a geometric door frame model which is independent to door appearances. Experimental results on our collected datasets of indoor signage and doors demonstrate the effectiveness and efficiency of our proposed method.
在导航和寻路应用中，标识对于找到目的地起着非常重要的作用。在本文中，我们提出了一种新颖的框架来检测门和标识，以帮助盲人进入不熟悉的室内环境。为了消除干扰信息并提高标识检测的准确性，我们首先使用显著图提取关注区域。然后通过二分图匹配在关注区域中检测标识。所提出的方法可以处理多个标识检测。此外，为了为盲人用户提供更多信息以进入与检测到的标识相关的区域，我们基于与门外观无关的几何门框模型开发了一种鲁棒的门检测方法。在我们收集的室内标识和门的数据集上的实验结果证明了我们提出的方法的有效性和效率。
