| RFID（Radio Frequency Identification）has been widely used in the manufacture, transportation,logistics, asset management and other areas, with the advantages of non-contact recognition, andautomatic identification, and fast interaction. However, RFID data stream contains a large numberof unreliable data, means negative readings, positive readings, and duplicate readings. Therefore,RFID raw data stream processing is becoming a hot spot.The thesis, based on the characteristics of RFID data stream, puts forward a four layersarchitecture-4LDSDF（Four Layer Data Stream Dispose Frame）,dividing the data processing intofour levels: the Physical Reader Layer’s function is to reduce negative readings, through deployingmultiple readers in a monitoring area; the Logic Reader Layer maps one or more physical readersinto one logic reader according to the logical mapping relations, then eliminate positive readingsusing the HM-BPRF（HashMap-Based Positive Reading Filter）algorithm; the Related Reader Layerremoves redundant readings according to the Fixed Formula and the HM-BDRR（HashMap-BasedDuplicate Readings Removed）algorithm; at last, the cleaning data stream can be used by the upperapplication system, should be stored in the Convergence Layer.The HM-BPRF not only solves the traditional window filtering methods’ problem of repeatstatistics and input/output data stream temporal inconsistency, but adopts positive judgment andcheck mechanism to reduce effective tag misjudgment, which improves the reliability of thealgorithm; The HM-BDRR introduces the concept of the process, and expresses the label eventinformation and process information with a unified data format, so the movement of the label isdisplayed in a concise and accurate way.Finally, this thesis applies4LDSDF into an asset management system, and realizes the functionof position tracking, history track query, and other functions, which prove the practicability of thearchitecture and the effectiveness of the algorithms. |
PDF Full Text Request