利用房间指纹基于走廊自动构建室内地图

Hallway based automatic indoor floorplan construction using room fingerprints

       人们70%以上的时间都处在室内空间。了解室内环境信息对广泛新兴的移动社交应用至关重要。在这些应用中往往需要室内楼层平面图。然而，很多环境中的室内地图不可用或者通过人工制作地图往往耗时、耗力且存在很多误差。本文中作者设计了一种自动构建室内地图的系统。该系统通过利用Wi-Fi指纹和用户的运动特征信息，通过三个关键步骤自动构建室内地图。

       第一步、构建房间邻接图，以确定房间之间的邻接关系；

       第二步、学习走廊布局，以估计沿着每个走廊的房间大小和房间顺序；

       第三步、进行伸缩处理，以调整房间大小并优化整个楼层平面图的精度。

       该系统可以用在新兴的移动应用中用于生成、使用、共享室内地图。

   People spend approximately 70% of their time indoors. Understanding the indoor environments is therefore important for a wide range of emerging mobile personal and social applications. Knowledge of indoor floorplans is often required by these applications. However, indoor floorplans are either unavailable or obtaining them requires slow, tedious, and error-prone manual labour. 

     This paper describes an automatic indoor floorplan construction system. Leveraging Wi-Fi fingerprints and user motion information, this system automatically constructs floorplan via three key steps: 

    (1) room adjacency graph construction to determine which rooms are adjacent;

    (2) hallway layout learning to estimate room sizes and order rooms along each hallway;

   (3) force directed dilation to adjust room sizes and optimize the overall floorplan accuracy. Deployment study in three buildings with 189 rooms demonstrates high floorplan accuracy. The system has been implemented as a mobile middleware, which allows emerging mobile applications to generate, leverage, and share indoor floorplans. 

    The system functions for a variety of indoor environments, and is robust to Wi-Fi noise and variation in occupant motion patterns. Deployment at three building sites with 189 rooms yields an average room position accuracy of 91%, room area estimation error of 33% and room geometric aspect ratio error of 24%, using on average only 20 data points per location for the system to converge.

     Keywords  Indoor floorplan; context sensing; indoor localization  

Automatic indoor floorplan construction system overview.

Jiang, Y., et al. (2013). Hallway based automatic indoor floorplan construction using room fingerprints. Proceedings of the 2013 ACM international joint conference on Pervasive and ubiquitous computing, ACM.