Quantifying Relationship between Relative Position Error of Localization Algorithms and Object Identification
Quantifying Relationship between Relative Position Error of Localization Algorithms and Object Identification
复制标题
量化定位算法相对位置误差与物体识别之间的关系
DOI：
10.1007/s11276-012-0516-2
复制
发表时间：
2013
影响因子：
3
通讯作者：
Hirozumi Yamaguchi and Teruo Higashino
中科院分区：
文献类型：
--
作者：
Noboru Kiyama;Akira Uchiyama;Hirozumi Yamaguchi and Teruo Higashino
物、设备和人的定位是普适计算的基础技术。然而，很少有文献讨论了定位误差的影响，由于定位算法的性能，如测距噪声和部署锚对人的识别对象。由于相对距离、相对角度和物体分组等因素在这种识别中相互关联，因此研究其特征并不容易。在本文中，我们提出的标准来评估的“准确性”的估计位置，在识别对象。这些标准有助于设计、开发和评估用于告诉人们物体位置的定位算法。增强现实就是一个典型的例子，需要这样的定位算法。为了建立没有歧义的标准模型，我们证明了Delaunay三角剖分很好地捕捉了人类在估计位置和真实位置之间寻找相似性的自然行为。我们已经研究了不同的定位算法，以观察所提出的模型如何量化这些算法的属性。还使用问卷进行了主观测试，以证明我们的量化足以呈现人类的直觉。
Positioning of things, devices and people is the fundamental technology in ubiquitous computing. However, few literature has discussed the impact of positioning errors due to localization algorithm properties such as ranging noise and deployment of anchors on people’s identification of objects. Since several factors such as relative distance, relative angles and grouping of objects are intricately related with each other in such identification, it is not an easy task to investigate its characteristics. In this paper, we propose criteria to assess the “accuracy” of the estimated positions in identifying the objects. The criteria are helpful to design, develop and evaluate localization algorithms that are used to tell people the location of objects. Augmented reality is a typical example that needs such localization algorithms. To model the criteria without ambiguity, we prove that the Delaunay triangulation well-captures natural human behavior of finding similarity between estimated and true positions. We have examined different localization algorithms to observe how the proposed model quantifies the properties of those algorithms. Subjective testing has also been conducted using questionnaires to justify our quantification sufficiently renders human intuition.