Application oriented data cleaning For RFID middleware

Radio Frequency Identification (RFID) technology is broadly used in many applications just like Supply Chain and Retail trade. However, owing to the working mechanism of RFID reader and RFID tag, raw RFID readings are not always reliable. False negative induced by low identifying rate in the reader´s minor detect region and duplication owing to detection region´s spatial overlap are the two main problems that cause unexpected consequence in upper applications of RFID Middleware. In this paper we discriminate the application scenarios by defining the concepts FN-S and “FN-NS” and propose an improved adapted smoothing method and an Bayesian method towards the two different scenarios. Both theoretically and by convincible experiment, we show the superiority of our algorithms.