地理时空变化语义约束的传感器动态观测数据在线清洗方法

Recent breakthroughs in sensor networks and the Internet of Things have made it possible to collect and assemble huge amounts of real-time sensor data for disaster mitigation and emergency response. However, irregular environmental changes and occasional instrument malfunctions have made noises and exceptions in observational data prominent. Therefore, before processing real-time sensor data online, data cleaning is urgently needed to ensure data quality. Since traditional data filtering methods didn’t take the data change pattern into consideration, these method have encountered some severe problems, including the poor adaptability of filter model, the low estimation precision and prohibitively high calculation cost. To overcome these shortcomings, this project presents a geographical spatio-temporal change semantics constrained online filtering method for real-time sensor data. The major research contents include: 1) Establishing semantic description model of geographic spatial-temporal changes; 2) Creating dynamic semantic mapping between real-time data changing pattern and the rules of spatial-temporal geographic process evolution; and 3) Studying he semantic-aware change discovery method, and the change semantic constrained Kalman filtering method to support the adaptive parameter optimization. The research is expected to surmount the critical issues of real-time sensor data cleaning, to provide dynamic data quality control for major natural disaster and emergency response and decision-making.

受外界环境变化和仪器设备故障等影响，灾害监测传感器动态观测数据的噪声和高值异常问题突出，严重制约了其实时应用效能，动态观测数据在线清洗因此成为世界性热点与前沿难题。灾害监测传感器动态观测数据具有复杂时变特点，针对已有数据清洗方法适应性差，难以根据动态观测数据的变化特征进行动态调整问题，本课题系统研究地理时空变化语义约束的传感器动态观测数据在线清洗方法，主要内容包括：建立地理时空变化的显式语义描述模型，在动态观测数据变化特征与地理时空过程动态演变规律之间建立高层语义映射，提出语义感知的传感器动态观测数据复杂变化特征发现方法和地理时空变化语义约束的传感器观测数据流在线卡尔曼滤波方法，实现滤波参数自适应优选。旨在突破传统动态观测数据清洗的关键瓶颈问题，为实时观测数据的动态质量控制提供一条可行的途径。

受外界环境变化和仪器设备故障等影响，灾害监测传感器动态观测数据的噪声和高值异常问题突出，严重制约了其实时应用效能，动态观测数据在线清洗因此成为世界性热点与前沿难题。灾害监测传感器动态观测数据具有复杂时变特点，针对已有数据清洗方法适应性差，难以根据动态观测数据的变化特征进行动态调整问题，本项目系统研究了地理时空变化语义约束的传感器动态观测数据在线清洗方法：建立了地理时空变化的显式语义描述模型，在动态观测数据变化特征与地理时空过程动态演变规律之间建立高层语义映射，提出了语义感知的传感器动态观测数据复杂变化特征发现方法和地理时空变化语义约束的传感器观测数据流在线卡尔曼滤波方法，实现了滤波参数自适应优选。本项目的数据清洗方法，有望突破传统动态观测数据清洗的关键瓶颈问题，为实时观测数据的动态质量控制提供一条可行的途径。