Key Technology Research On RFID Data Cleaning

RFID technology, which originated during the Second World War, has been used in many domains, such as logistics, aviation, supply chain management, pharmaceuticals. RFID has the trend of replacing barcode, due to its advantages such as automatic, fast, batch processing and non line of sight. Furthermore, due to the unreliable of RFID data streams, such as false negative readings, false positive readings and duplicate readings, restricts the widespread in many applications, thus, it poses many interesting challenges and opportunities in the data management systems. Therefore, how to efficiently clean RFID data streams has becoming an urgent problem to solve.To solve the problem, existing solutions always focus on the raw RFID data level and utilize sliding window to smooth data, or consider the spatio-temporal correlation among many mornitored objects, or take account of the redundancy of RFID data in spatio/temporal aspect.These technologies will interpolate many unneeded data in period of data cleaning, and the cleaning results cannot reach the ideal effect.Firstly, the paper introduces the development progress of RFID technology, then, it analyses the differences between RFID data steams and tranditional data, further, it reviews existing solutions in RFID data management and points out their problems or disadvantages, finally, based on them, it does research on RFID data cleaning.To meet the needs of users in data level, the paper proposes two data interpolating strategies, the first one is deterministic method, which includes time interval based model, containment relationship based model and inertia based model, the other one is probabilistic method, namely normal distribution interpolating model.To solve the problem of existing solutions, the paper proposes a novel RFID data cleaning strategy based on communication information among readers. This model deals with dirty data from logic level, which can reduce the generation of duplicate data. Furthermore, the communication information among readers is highlighted firstly to solve the problem of data quality. Firstly, to formalize the communication, the paper devises a novel communication protocol for RFID readers and gives a dynamic probabilistic cell event model. Then, based on the protocol and model mentioned above, we give an active RFID data cleaning strategy, which includes duplicate data reducing method (D-DR), missed data interpolating method (Topk-PDI and M-PDI) and positive data reducing method (P-DR).To evaluate the performance and effectiveness strategies and algorithms proposed, the paper conducts many simulated experiments. Further, the experiments not only demonstrate the efficient and effectiveness of our strategies and algorithms, but also show our methods have the advantages of real-time and accuracy.