中文摘要：

无线传感器网络是物联网（Internet of Things）的重要组成部分，利用其实现物联网中目标的定位技术已成为研究热点之一．由于受环境、障碍物、网络攻击和硬件错误等诸多因素的影响，传感器节点所采集的数据易产生较大误差，形成错误数据，从而对定位造成严重影响．尽管已发展出了众多定位算法和模型，但针对错误数据实现定位的研究还较罕见，尤其在国内，几乎是空白．文中针对上述问题，旨在利用网络（几何）拓扑结构信息，提出一种用局部信息刻画全局分布密度信息的新颖物联网定位模型：鲁棒的局部保持的典型相关分析定位模型LE—RLPCCA．与现有同类典型方法在真实环境中的实验结果相比，LE-RLPCCA具有更高的定位鲁棒性和稳定性．

英文摘要：

Wireless Sensor Network-based localization technique, as one fundamental component of the Internet of Things, has attracted more and more attentions. However, some of the collected data easily result in faulty data with large biases due to the network attacks, large obstacles, network environments and hardware errors, which brings a big challenge for localization. The localization research in such a scenario is still relatively few. The authors try to develop a novel localization algorithm to solve this problem, by characterizing the density information of data distribution with sufficiently exploiting topology structures in WSN, and name it as Location Estima- tion- Robust Locality Preserving Canonical Correlation Analysis （LE-RLPCCA）. Experimental results in realistic environments show that LE-RLPCCA achieves higher robust and stability for faulty data than other publicly available advanced localization algorithms.