RFID Data Cleaning Processing Strategies And Algorithms

Before RFID technologies were only used in the area of military. In recent years, with the emergence of Internet of Things, because of RFID products' small size, low cost, no contact reaction, RFID technologies have been widely used in supply chain management, production automation, entrance guard, highway toll, cargo tracking and other civilian areas. Some problem may occur when RFID reader identifies tags because of some environmental factors. These problems include false positive readings, false negative readings, tag duplicated readings and reader duplicates. These negative facts would affect development of RFID applications.To clean raw RFID readings with those problems some algorithms are described in this thesis. Because of the infinity of the RFID data stream, RFID data cleaning make use of sliding window techniques. Eliminating false positive readings is essentially removing the noise data from the raw data stream. The algorithm for this purpose determines whether tag was noise by counting the number of the tag in the sliding window. The algorithm for solving false negatives uses the time smoothing technique, which is essentially recompensing the lost tags in the time dimension. By keeping that the timestamp interval between two adjacent tags is longer a certain threshold, redundant data are removed. Removing the spatial redundancy is smoothing data in spatial dimension, which is merging multiple tag streams and keeping the tag timestamp order. The thesis proposes a technique based on the spatial deployment of readers to set window size. The advantage of the method is that the change rate of sliding window size is low and has little effect on performance of clean process. In the terms of RFID security protection, the system also apply an access control policy based on tag and reader level security. In order to improve the performance of cleaning system, a strategy using clustering and dynamic load balancing techniques is proposed.Simulation experiments show that the using dynamic load balancing strategy effectively improve the efficiency of data cleaning.