Research And Implementation Of Heterogeneous Data Fusion On Mobile Collaboration Sensing Platform

As the fast development of communication and hardware technology, mobile smart terminals are becoming increasingly popular. These smart terminals are integrated with a large number of advanced sensors and have powerful computing ability, the mobile collaborative sensing platforms emerged in this context.Mobile collaborative sensing system is becoming widely used by people due to the widespread use of smart phones, this system collects a lot of information provided by the smart phone users which can be used to extract useful information in order to meet the need of different applications. However, lots of challenging problems would occur in the process of deployment and operation in actual environments. Firstly, mobile nodes are random, and a large number of nodes are not uniform and provide services with dynamic uncertainty. Secondly, different nodes have different resources, the heterogeneous character of the system leads the diversity of the sensing services. Finally, participants with the background of different cultures and educations influence the precision and reliability of the collected data, therefore the collected data sets will produce gaps in time and space and will be redundant in a small range. Using these data sets to evaluate the environmental index in a particular area would become very difficult. In order to solve the above problem, this thesis will study the data fusion strategy in mobile collaborative sensing system.In this thesis, after learning data fusion technologies and the weighted average algorithm, we implement the algorithm on the mobile collaboration sensing platform using temporal and spatial kriging algorithm. First of all, after referring to quantities of data fusion algorithm and papers, we find that kriging method can solve this problem efficiently. Then, we introduce the design of data fusion module which includes database design, interface design and grid design. After that, by using GreenOrbs and GeoLife datasets, we generate the data sets of Beijing University of Posts and Telecommunications, and these generated data sets are used to simulate collected data on mobile sensing system. Finally, since the ordinary kriging algorithm just considers the space dimension and ignores the time factor, we propose temporal and spatial kriging algorithms, we test it using experiments and implement the algorithm. on our platform.