Exploring differences of visual attention in pedestrian navigation when using 2D maps and 3D geo-browsers

ABSTRACT

Despite the now-ubiquitous two-dimensional (2D) electronic maps, three-dimensional (3D) globe viewers, or 3D geo-browsers such as Google Earth and NASA World Wind have gained much attention. However, the effect of such interactive 3D geo-browsers on spatial knowledge acquisition and decision-making is not well known. This study aims to explore the potential benefits of using interactive 3D geo-browsers in three processes of pedestrian navigation (self-localization, spatial knowledge acquisition, and decision-making) in digital environments. We employed eye tracking to show differences of visual attention in pedestrian navigation between a 2D map (Google Map) and a 3D geo-browser (Google Earth). The results indicated that benefits and drawbacks of 3D representations are task dependent. Participants using the 3D geo-browser had an extensively visual search resulting in significantly longer response time than the 2D participants for spatial knowledge acquisition, whereas 3D users performed a more efficient visual search and resulted in a better navigation performance at complex decision points. We speculate that the inefficient knowledge acquisition when using the 3D geo-browser was most probably due to information overload and obstructed views. Landmarks in photorealistic 3D models assisted recall of spatial knowledge from mental maps, which contributed to efficient decision-making at a complex turning point. These empirical results can be helpful to improve the usability of pedestrian navigation systems.     

室内移动机器人自定位方法

针对室内机器人移动导航定位存在的问题,该文研究了基于Kinect传感器的定位算法。该方法应用Kinect传感器采集了移动机器人运动过程中连续帧的彩色和深度信息;通过尺度不变特征变换算法匹配连续场景图像中的特征点对,经随机抽样一致性算法剔除点集中的误匹配点;利用Kinect的深度信息将可用点集的二维坐标转换到相机坐标系下的三维坐标;并使用绝对定向算法计算机器人在相邻位置上的姿态信息和平移量,从而提高了机器人的移动参数的计算精度。通过实验验证了该方法的可行性,且能够满足实时要求,可以同时进行定位与地图创建。